Below are my calculations of crude oil prices based on published flow and elasticity data. They are consistent with where oil prices are today, and forecast a large increase when temporary buffers go.
I am drafting a full article to explain the rationale, which will be forthcoming in the next day or two.

Flow calculation table.

Baseline
Pre-war Hormuz flows: 20.1 mb/d
= 14.2 mb/d crude and condensate
+ 5.9 mb/d petroleum products

Permanent offsets
Bypass capacity, attested demand destruction, Iranian selective passage:
4-7 mb/d.
Petroleum products: zero offset.

Temporary offsets
SPR and coordinated reserve releases:
1.5-2.5 mb/d.
Reduced Chinese imports:
3.5-4.0 mb/d.

Net deficit with all measures deployed: 1-5 mb/d

Elasticity literature
Fed consensus short-run elasticity: –0.13.
Recent 2001–06 estimates: –0.077 to –0.034.

Price range consistent with current market
At 1-5 mb/d net deficit, the calculation using consensus elasticity produces:
$74-95.
Current $85-100 consistent with that range.

When the temporary offsets go, the net deficit moves to:
7-10 mb/d crude
+ 5.9 mb/d products shortage, which has no offset and transmits directly into crude pricing

Durable net deficit = 13-16 mb/d

At that point, depending on elasticity used, the calculations give:
Consensus –0.13: $136-151
–0.077: $182-208
–0.034: $325-387

In a supply shock of this magnitude, the probability shifts toward the more inelastic estimates.

The market is correctly priced if the temporary buffers were durable. It is not priced for when those temporary buffers run out.

Full article forthcoming.

On Saturday the AFR’s Chanticleer featured observations by Viktor Shvets and Matt King, two serious analysts agreeing on the same uncomfortable diagnosis: markets are frothy, momentum has replaced fundamentals, and professional investors know the rally is fragile but keep buying anyway. However, the comfort they draw from this picture is where the analysis goes wrong.

Shvets argues that the financial system cushion – capital running at five to ten times global GDP – gives markets an enormous buffer against correction, while King warns that correlation risk makes a momentum reversal dangerous, concedes that a further ‘melt-up’ pulling in more late money is equally plausible on any given week, and leaves open what actually arrests the cycle. Both are diagnosing the same trap – the market knows the structure is fragile but cannot prudently sit out – yet neither asks the pertinent question of whether the cushion is what it appears to be.

A second Chanticleer piece, published yesterday, responds to Wall St’s “nasty fall” on Friday, identifying the two most likely bubble poppers as rising interest rate expectations and hubris around AI spending. Unexpectedly hot US jobs data forced markets to price a Fed rate rise before year end, while SpaceX’s equity raise – with pricing set for 11 June and trading beginning 12 June – Meta’s rumoured capital call, and Google’s $80 billion tap of equity markets all signalled that the hyperscalers can no longer fund AI capex from their own cash flow. The analysis is correct on the financial mechanics, but what it does not address is why the interest rate pressure is stickier than previous episodes and why the AI capex squeeze is harder to relieve. The answer to both lies outside the financial system, in a physical constraint that the usual rate cut solution cannot solve.

The five-to-ten-times figure Shvets cites appears to be the aggregate stock of financial assets – equities, bonds, derivatives, bank assets, and shadow banking vehicles – measured against annual global output, but the problem with it is that inflated asset prices are included in the numerator. The cushion therefore incorporates the very thing it is supposed to cushion against: when equity markets fall 40%, the financial capital ratio falls with them, and the buffer shrinks at the moment it is needed most. Shvets is using the bubble as evidence of the safety net, but it is undermined by the circularity running through it. The capital stack is a sand castle: bonds, derivatives, and private credit built on fifteen years of compressed yields and abundant foreign capital, both of which are now withdrawing.

What Shvets is pointing to is a stock of claims, yet the real backstop in a crisis is not a stock but a power – the authority central banks hold to manufacture settlement liquidity and lend against distressed assets. No pre-existing pile of capital is needed for that; what is needed is political permission, fiscal headroom, and tolerance for inflation, and all three are narrower now than they were in 2020.

Covid created that confusion. Markets broke in March 2020 as the ‘dash for cash’ reduced J.P. Morgan’s Treasury market depth – how much of a security you can buy or sell without moving the price – to only $12 million against a normal of hundreds of millions, and the repo market that funds leveraged positions using those Treasuries as collateral was at the edge of seizure. The Fed rescued the entire market-based finance system by backstopping it, while governments replaced lost household and corporate income at levels normally associated with wartime. The result was not merely recovery but a larger stock of financial claims priced on the assumption that future shocks would also be liquidity-bridgeable – shocks where underlying productive capacity remains intact and the constraint is purely temporal, bridgeable by income support and market liquidity until the real economy restarts. Asset prices recovered faster than the real economy, and the financial system emerged with a higher ratio of claims to underlying productive capacity and with the lesson embedded that shocks could be monetised. Shvets’s cushion is partly the residue of that lesson: the market has capitalised expected rescue capacity into current valuations, which means the cushion and the bubble are the same object viewed from different angles.

The rescue assumption has a load-bearing condition, however: the shock must be liquidity-bridgeable. The Strait of Hormuz has been effectively closed since late February, and the Houthi declaration of 8 June threatening a Red Sea blockade now puts the only significant maritime workaround under direct pressure – Saudi Arabia has been routing oil exports via the Red Sea because Hormuz is closed. Diplomatic talks are stalled, Iran and Israel exchanged missile fire on 7-8 June in the most serious test of the April ceasefire to date, and while both sides have conditionally stood down, the underlying conflict is unresolved. What is missing from this system is not liquidity but molecules – barrels that cannot move, feedstocks that cannot reach crackers, and LNG that cannot reach European and Asian industrial users. The standard objection that the world is less oil-dependent than in the 1970s mistakes where the dependence has migrated: from crude into gas and petrochemical feedstock reliance, while just-in-time logistics eliminated the inventory buffers that previously absorbed disruption across multiple production cycles. The current shock transmits not through the price channel but through a volume constraint propagating downstream through input-output chains, each exhausted buffer revealing the next, on no policy-determined timeline.

That physical constraint is also what makes the stagflation threat real rather than theoretical. The Survey of Professional Forecasters projects CPI at 6% for 2Q26, with May data due tomorrow and the FOMC meeting on 16-17 June – central banks facing energy-driven inflation cannot ease without risking re-anchoring inflation expectations upward, yet the real economy is slowing simultaneously. The fiscal side of the rescue is constrained too: global debt has reached a record $353 trillion, and the 2020-scale income replacement that converted a real-economy shutdown into a liquidity-supported financial order is not available at the same scale twice in 6-7 years.

The financial system the physical shock arrives into is already carrying the vulnerabilities that 2020 exposed and only partially resolved. Private credit sits on a large pool of assets without daily liquidity or transparent pricing, with the AI disruption of SaaS revenue streams now threatening the cash flows that justified the original lending. The AI hyperscaler story connects here because data centres run on the energy and feedstocks the Hormuz closure is restricting, and hyperscaler valuations are the primary weight in the current momentum trade. A correction there does not stay in equities – it travels through the collateral and leverage structures of structured credit markets, as a US housing correction travelled through mortgage-backed securities into the global banking system in 2008.

Three of the world’s largest sources of cross-border capital are withdrawing simultaneously through independent channels – Japanese repatriation as the Bank of Japan normalises policy, GCC sovereign wealth fund redirection as reduced hydrocarbon revenues compress the capital available for foreign deployment, and China’s structural reduction in long-end Treasury exposure – and the combination is tightening global financial conditions through channels that conventional monetary policy cannot offset.

The belief that this cycle is structurally different from the ones that ended badly is the standard argument at the top of every previous one. The bull case requires Iran and tariffs and China’s balance-sheet slowdown and the credit cycle and Japanese repatriation and GCC redirection and hyperscaler free cash flow to all turn at once across drivers that share no common factor. That is a wish list, not a scenario, and the burden of proof sits with those making it. The overdetermined downturn argument has not shortened – if anything it has lengthened.

King is right that timing is unknowable, and right that the momentum trade traps even those who see the fragility – one cannot prudently sit out while it continues. However, neither observation bears on whether there is a real cushion and the cushion Shvets identifies is nominal rather than real: a stock of claims contingent on the same conditions it is supposed to buffer against, backstopped by institutional rescue capacity that is narrower now than it was in 2020 and designed for a different kind of shock. The Chanticleer framing – the bubble ends through rate expectations and AI hubris – assumes a liquidity-bridgeable world. The question Friday’s session opened is whether this particular correction has that exit available to it. A 1970s comparison is a subject for another day, but the historical precedent for energy-driven stagflation meeting an over-leveraged financial system is not encouraging.

The railways transformed Britain’s economy for a century, yet the investors who funded the 1840s mania lost two-thirds of their money. Similarly, the dot.com boom saw the failures of many investors, but the internet did not fail. Worthwhile technology endures even though capital structures often do not.

The technology sceptics who called AI overblown after February’s sell-off were reading collapsing valuations as evidence of collapsing technology, the same error as calling the internet finished in 2001. The boosters who respond by citing real capabilities and genuine disruption are right about the technology but wrong to think that settles the valuation question. AI will diffuse through the global economy over the coming decades in ways that exceed current imagination, yet it is also entirely consistent with most current investors losing most of their money. The argument that AI disruption is real – and that the problem is not failure but a mismatch between who captures the gains and who holds the costs – was set out in AI Is Not Failing. That’s the Problem. This piece picks up where that one left off.

Carlota Perez laid out the explanation in 2002, at the moment dot-com sceptics thought history had vindicated them. She had spent two decades studying the five major technological revolutions of the previous two centuries – railways, steel and electricity, the motor car, mass production, and the internet – and found the same pattern in each. Speculative capital funds an installation phase that builds the infrastructure. A crash follows. Then a deployment phase, where institutions adapt and the technology diffuses far more widely and powerfully than the original frenzy implied. The bubble is how the infrastructure gets funded, not proof of a failure of the technology. The fibre-optic cables and data centres that made the post-2000 internet possible were paid for by investors who lost everything on the companies that built them.

Jeremy Grantham, co-founder of GMO and one of the most closely followed bubble analysts of the past three decades – correctly calling the dot-com bust and the 2008 housing crash – calls AI “maybe the most visibly impressive innovation of the last 100 years, perhaps of a magnitude equal to the railways of the 19th century.” He also thinks it is a bubble. He is not contradicting himself. A genuine technological revolution and a speculative mania are not mutually exclusive – historically, they often appear together.

The capex-revenue gap

The bullish response to this is that today’s money is different. The hyperscalers – Amazon, Alphabet, Meta, Microsoft – are profitable companies with genuine cash flows, not the hollowed-out balance sheets of Pets.com. This is partly right, but it is becoming less right at increasing speed.

The five largest hyperscalers – adding Oracle to the four above – will spend more than $700bn on capital expenditure in 2026, roughly double 2025 levels, with approximately three-quarters tied directly to AI infrastructure. The deterioration in their aggregate free cash flow is striking: the group was generating around $45bn per quarter in free cash flow in 2024; by 1Q26 that had fallen to around $4bn. After dividends and buybacks, aggregate capex now exceeds projected free cash flow, meaning hyperscalers are beginning to borrow to fund AI investment, having previously funded it from operations. Amazon is projected to run negative free cash flow of up to $28bn in 2026, and filed with the SEC indicating it may seek to raise additional equity and debt as its build-out continues. The prisoner’s dilemma is driving this: no individual hyperscaler can safely reduce spending without risking being overtaken, so collective over-investment is guaranteed even when each participant can see the problem.

Against that investment, the revenue required to justify the infrastructure spend remains far ahead of proven, durable end-user willingness to pay. In a June 2024 piece for Sequoia Capital, partner David Cahn posed what he called “AI’s $600 Billion Question” – an update to his earlier “AI’s $200 Billion Question” from September 2023. His calculation is straightforward: take Nvidia’s data centre revenue run rate, double it for total infrastructure cost of ownership, double again for the gross margin operators need to justify the spend, and the result is the annual revenue the AI industry must generate to make the capex rational. In late 2023 that implied requirement was $200bn, and by mid-2024 it was $600bn. In 2026, with Nvidia revenue running substantially higher, the implied number is larger still, and the gap that Cahn identified two years ago has widened. In his December 2025 outlook for Sequoia, Cahn observed that end-user AI revenue remains “on the order of tens of billions per year” against infrastructure investment “on the order of trillions over the coming five years” – the same gap, but at larger scale. A 2026 Bain & Co survey reinforces the same point from the demand side: nearly 40% of companies that have measured their AI cost savings report a figure below 10%, despite having targeted reductions of 11–20%. The technology is being adopted, but it is not yet generating the enterprise productivity gains at the magnitude the investment assumes.

The revenue problem is not just about adoption rates, but also the economics of how AI is priced. The Economist explained a year ago that traditional annual recurring revenue – based on durable subscription commitments with churn below 5% – has been replaced in AI by what one venture capitalist called “experimental run rate”: customers experimenting enthusiastically but not committing durably, with churn above 20% in many products. The revenue growth is there, but its durability is untested and its current valuations assume it is more durable than it is.

Whether this is a bubble in the statistical sense is more complicated

Amundi’s Investment Institute – the research arm of Europe’s largest asset manager – ran a formal test. They constructed AI and non-AI portfolios from the S&P 500 and applied the autoregressive bubble detection methodology used in the academic literature, testing whether P/E ratios show the explosive non-stationary dynamics that characterised dot-com valuations from 1997 to 2000, the mathematical signature of a true speculative bubble. Their conclusion, published in April 2026, was that the current AI episode does not exhibit those dynamics. AI stocks’ P/E ratios have been falling from 2023 to 2025 despite strong price appreciation, because earnings growth has outpaced price gains. During the dot-com period, P/E ratios in TMT (Technology, Media and Telecommunications) expanded continuously and the statistical test found the signature of a genuine bubble. For AI stocks through end-2025, the same test found no such pattern.

This piece calls it a bubble anyway, and that needs explaining. The Amundi finding applies to established AI stocks in the public market through 2025. What this piece focuses on is the capital structure – the gap between installation-phase investment and deployment-phase returns – not a claim that every AI stock is statistically overvalued right now. The risk Amundi identifies reinforces rather than contradicts this: concentration, not speculative excess in the classic sense. At peak, 45 stocks held a 38.2% index weight, with average individual stock weights nearly double those of the dot-com TMT basket. A narrow group of companies has driven a disproportionate share of returns, and Amundi’s own comparison of the AI portfolio’s performance with the 1997–99 TMT trajectory – the takeoff and exuberance phases of the dot-com cycle – is a warning about where the cycle may be heading even if it has not yet shown explosive characteristics in the public market.

What the Amundi analysis cannot assess is the private market, where The Economist’s “vibe valuing” operates – Mira Murati’s startup at $10bn before it had a strategy, Perplexity at 400 times revenue, OpenAI burning several billion dollars annually while valued at more than $800bn. The public market through 2025 was pricing real earnings. The private market has been operating on different assumptions entirely, and the 2026 IPO wave is what brings those assumptions into public markets at scale.

SpaceX and the Odlyzko problem

SpaceX, Anthropic, and OpenAI are all now moving toward public markets simultaneously, at combined implied valuations approaching $3.5tn. None is profitable. Goldman Sachs – as lead underwriter earning its fee when the SpaceX deal closes at the highest possible price – projects that the xAI division will grow revenue from $3.2bn to $322bn by 2030, while SpaceX total revenue rises from $18.7bn to $474bn over the same period. That is a bold growth assumption to say the least. It is more accurate to say it is a market-share reversal assumption: as Morningstar’s own analysis of the S-1 notes, to justify the Goldman projections Grok would need to surpass Anthropic, Google, and OpenAI in coding, cybersecurity, agents, and chatbots from a current position of competitive inferiority. Morningstar puts SpaceX’s fair value at $780bn, less than half the $1.75tn IPO target. A revealing analysis on X, applying comparable market multiples to each of SpaceX’s three business segments independently, arrives at $550bn, a different methodology converging on roughly the same territory. The gap between those two disciplined estimates and the IPO price is not a rounding error.

The TAM (Total Addressable Market – the total potential revenue available to the business) claim that anchors the valuation needs a brief mention. SpaceX’s prospectus identifies a TAM of $28.5tn – larger than any economy on earth except the United States. Less than $2tn of that figure is attributed to space and satellites, the business SpaceX actually operates. The remaining $26.5tn is AI. The word “AI” appears 1,251 times in the prospectus. “Multiplanetary” appears 18 times.

The S-1 filed on 20 May 2026 sets out the trajectory in detail. In 2024, SpaceX was profitable at $791m net income, but that was prior to the xAI acquisition which changed it overnight: xAI recorded a $6.35bn operating loss in 2025 on $3.2bn in revenue, an operating margin of minus 200%. In 1Q26 alone, xAI posted $818m in revenue and a $2.47bn operating loss — an annualised margin of minus 302%. Scaling appears to be making the economics worse, not better. As PitchBook senior research analyst Harrison Rolfes put it: “If you compare xAI to a traditional SaaS company – cloud-based software delivered by subscription – the financials look reckless. If that payoff doesn’t happen, we are looking at the largest venture-funded correction in history.”

Andrew Odlyzko, a mathematician and financial historian at the University of Minnesota, whose research on the British Railway Mania is the standard academic reference on the subject, found that a contemporary calculation showed the 8,000 miles of newly authorised British railways would require passenger traffic to increase fivefold within five years from an existing base of 34 million to deliver their projected 10% return. Odlyzko noted that this calculation “should have alarmed observers by itself”, but collective psychology distracted attention from it. The Goldman xAI projection is the equivalent calculation, stated boldly rather than hidden in the SpaceX prospectus, and the index inclusion mechanism now ensures it is priced by passive investors whether or not any individual investor has read it.

A further detail worth noting: Anthropic – simultaneously filing for its own IPO at a $965bn valuation – is SpaceX’s largest disclosed revenue customer, paying $1.25bn per month for compute capacity through May 2029 under a contract either party can terminate on 90 days’ notice. Goldman’s projections for xAI revenue are thus partly dependent on a contract held by a company that itself has no confirmed path to profitability. The Goldman projections require multiple things to go right simultaneously, in the same period that the broader macro environment is deteriorating. Readers familiar with the overdetermined downturn argument – that the optimist needs every causal chain to resolve at once – will recognise the structure.

The index rules and the conscription problem

Index investing is sold as neutral exposure to the market. When index providers alter entry rules for giant, low-float, newly listed companies, then passive investors become exit liquidity, the buyers who allow early investors to exit at IPO pricing, without having chosen to participate in the valuation themselves.

S&P Dow Jones announced on 5 June that it would not change its eligibility requirements, explicitly holding the four-quarter GAAP profitability requirement in place, stating that “exceptions to the financial viability, seasoning, and investable weight factor requirements should not be granted solely based on market capitalisation.” SpaceX posted a net loss of $4.94bn for full-year 2025, and a further $4.28bn in 1Q26 alone, leaving its earliest possible S&P 500 eligibility at mid-2027 at the earliest, and only if it achieves profitability. S&P held the line, but Nasdaq and FTSE Russell did not.

Nasdaq’s Fast Entry rule, effective May 2026, allows companies ranked in the top 40 by market cap to join the Nasdaq-100 within 15 trading days of listing, down from a prior three-month minimum. The minimum float requirement has been eliminated, though companies receive a lower index weighting until more shares are available for trading. FTSE Russell shortened its window to five trading days. It was reported that SpaceX advisers were in discussions with index providers about accelerated inclusion earlier in the year, and the rule changes at both Nasdaq and FTSE Russell followed.

The passive-buying story is therefore a Nasdaq-100, Russell 1000, and FTSE story – not an S&P 500 story, for now – but it is still substantial. Around $527bn tracked the Nasdaq-100 at end of 2025, and with SpaceX floating only 3-5% of its total shares, the mechanical demand concentrated into that thin float will be significant. Bloomberg Intelligence estimated that fast S&P 500 inclusion alone would have triggered $14bn in forced passive buying for SpaceX – Nasdaq-100 and Russell inclusion will produce their own comparable flows. Every Australian superannuation member with US equity index exposure, every UK pension holder with global equity allocation, every US 401(k) holder in a total-market fund is being enrolled in this valuation without having chosen it. Anthropic is reported to be targeting a $965bn valuation in a confidential IPO filing and OpenAI is expected to follow in Q4. The index conscription problem for SpaceX is the floor of what this IPO wave produces, not the ceiling.

The Amundi concentration warning and the index rule rewrites are the same problem crystallising from different angles. Through 2025, investors at least chose their AI exposure, but from 2026, much of it is assigned to them. The distinction between markets as measurement and markets as mechanism – explored in Equity Markets: The Thermometer That is Mistaken for a Thermostat – is what the index rule rewrites collapse. When passive buying is mechanically triggered by index weight, the market stops measuring and starts directing.

The private credit channel

The part of this story receiving the least attention carries the most direct implications for the broader economy, and it connects directly to the three withdrawals argument developed earlier.

Outstanding loans to SaaS firms – cloud-based software companies delivering services by subscription, which for a decade represented some of the most stable and predictable cash flows in the private credit market – rose from around $8bn in 2015 to more than $500bn, equal to 19% of all direct loans, by end-2025, according to the Bank for International Settlements. The thesis underwriting those loans was stability: sticky recurring revenue, high gross margins, predictable cash flows, and low churn. For a decade, it was sound, but AI is now eroding those specific characteristics. The BIS’s March 2026 Quarterly Review found that software stocks fell almost 30% between October 2025 and February 2026, and that BDCs (Business Development Companies – specialised investment vehicles that provide direct loans to small and mid-sized businesses) with higher SaaS exposure underperformed peers by around 5 percentage points over the same period. Discounts to NAV (Net Asset Value) deepened, signalling concerns that the book value of illiquid private loans no longer reflects realisable values.

UBS has estimated $75–120bn in fresh defaults across leveraged loans and private credit by year-end 2026, driven by AI disruption of software borrowers. TPG’s co-managing partner has stated where the exposure is concentrated: portfolios underwritten in 2018–2020, before generative AI, carry the highest vulnerability. Those loans were written against cash flows that AI is now repricing downward, meaning the capex that is building the AI infrastructure is also destroying the revenue streams that backed the debt that financed the software buyouts. Industry-wide redemption requests spiked in late 2025 and early 2026, with several large non-traded BDCs and interval funds capping or slowing withdrawals.

Oliver Wyman drew the 2008 parallel: in 2008, banks discovered they owned far more US housing risk than their internal reports suggested, because exposures were distributed across multiple book categories that did not appear as mortgage risk in standard credit reporting. AI-related credit exposure is distributed similarly across corporate, real estate, infrastructure, fund financing, and alternative credit books, and thus may not appear as technology risk until it fails.

Where the three liquidity withdrawals tighten the vice

The Oliver Wyman and BIS analyses both assume relatively normal global liquidity conditions. They do not price the three withdrawals running simultaneously: Japanese repatriation driven by rising domestic yields as the Bank of Japan normalises policy, GCC (Gulf Cooperation Council) sovereign wealth fund (SWF) redirection toward domestic priorities as reconstruction and defence spending pull capital inward, and China’s structural reduction in US Treasury holdings as a function of reserve diversification and sanctions risk management.

The withdrawals operate through both the Treasury market and the US private asset ecosystem – private equity, venture capital, real estate, and credit funds. Gulf SWF redirection does not mean only selling Treasuries; it also means reducing allocations to the US private capital pool that has been funding leveraged buyouts, growth equity, and the software sector buyouts described above. Remove that patient capital from the investor base of large alternative managers and the refinancing extension model that has been keeping the private credit slow cycle functioning becomes harder to sustain. A State Street survey of SWF allocations published in March 2026 confirmed that SWFs have been systematically reducing fixed income allocations and redeploying into private markets, digital infrastructure, and AI – meaning the same capital withdrawing from one part of the ecosystem is partly inflating valuations in another.

Each of the three global liquidity withdrawals independently compresses global dollar liquidity. Running together, they tighten the refinancing environment for stressed software borrowers at the time when AI disruption is hitting their cash flows and the maturity wall for 2018–2020 vintage software debt is approaching – an estimated $620bn due through 2027. If the refinancing window narrows through external liquidity compression at the same time that AI-driven cash flow deterioration is reducing borrowers’ debt-servicing capacity, the defaults UBS is projecting are a floor, not a ceiling.

Harvard economist Jason Furman calculated that AI-driven infrastructure investment accounted for 92% of US GDP growth in the first half of 2025, and that excluding these technology-related categories, annualised GDP growth would have been just 0.1%. This pattern has held into 2026, with software and IT equipment investment contributing 67% of 1Q26 GDP growth of 2.0%, meaning the non-AI economy was growing at roughly 0.7% annualised. A slowdown in AI capex does not stay inside the technology sector. It reverses a contribution that has been carrying a disproportionate share of headline growth, concurrently with tariff uncertainty, China’s structural drag, the credit cycle position, and the oil price and hydrocarbon shocks from the Strait of Hormuz. The overdetermined downturn argument must now incorporate the AI capex cycle as one of the causal chains, and if it turns, the optimist needs everything else to resolve simultaneously.

The technology will exceed current expectations over coming decades. The Perez framework, the historical precedent, and the genuine capabilities already being demonstrated all point the same way. What the railways built in the 1840s, what the internet built in the 1990s, and what AI is building now is infrastructure that the deployment phase will use long after the installation-phase investors are gone. The question over the medium term – say, three to seven years – is not whether the technology succeeds, but who holds the losses when the gap between the Goldman projections and the reachable revenue closes, and through which channels those losses travel into the parts of the financial system that have nothing to do with AI.

Postscript – 8 June 2026

The article above was finalised as SpaceX’s IPO roadshow was under way, with pricing expected after market close on 11 June. Friday 5 June’s session produced a useful preview of the argument in action.

The S&P 500 fell 2.6% on 5 June, its worst day since October, with the Nasdaq down 4.2%, its worst single session since April 2025. Among the heaviest falls were Nvidia down 6.2%, Broadcom nearly 8%, and Micron more than 13%. The VIX (the market’s volatility gauge) surged 40% to its highest level in two months. South Korea’s tech-heavy market fell 5.5%. The immediate trigger was a May jobs report that came in roughly double expectations – 172,000 against an 80,000 forecast – which pushed bond yields higher and revived expectations of Federal Reserve rate rises.

The apparent contradiction in the mechanism is worth noting. Good economic news produced a technology sell-off because strong employment data threatens the rate environment on which elevated AI valuations depend. The argument for current multiples requires, among other things, that monetary conditions remain loose enough to support long-duration growth assets at these prices. A labour market running well above expectations at this point of the cycle – with energy costs from the Hormuz disruption already feeding into inflation – makes that assumption harder to sustain. This is the monetary trap described in Markets and Economists Are Both Right About Rates: the supply constraint does not ease when demand weakens, so central banks face deteriorating growth and still-elevated prices simultaneously.

This is not a crash. Briefing.com’s Patrick O’Hare explained the market’s own reading: “Everyone’s realising that perhaps this rally off the March lows has run its course for the time being.” What Friday showed is that the AI valuation complex is simultaneously rate-sensitive and narrow enough that a single day’s selling in a handful of semiconductor names can produce the S&P 500’s worst week in ten. That is the concentration risk the Amundi paper identified, observed in real time.

The same argument, in the form of tables or other variants, circulates constantly on social media. Rows of ticks and crosses – Israel green, Palestinians red, every row, without exception. The table is attributed to sources including the UN, the Arab League, and the US State Department, as though the framing itself were archival rather than prosecutorial. Danny Ayalon, who says he participated as part of the Israeli delegation at Camp David in 2000, posts his version regularly, typically citing Abba Eban’s line that the Palestinians “never miss an opportunity to miss an opportunity”.

Most of the table’s entries are not Palestinian rejections of formal, implementable, sovereign-state offers. Before any entry can earn a red cross in the Palestinian column, it needs to answer four questions: was the proposal made to the Palestinians themselves; was it formal enough to be accepted; did the offering side have authority to implement it; and would acceptance have produced a sovereign state, or managed autonomy under continued Israeli control? Most entries fail at least one of these tests before the political analysis even begins.

None of that requires Palestinian blamelessness. Palestinian leaders have made mistakes. Yasser Arafat’s failure to convert the Clinton Parameters into an agreement in December 2000 was a genuine missed opportunity – the closest thing in this history to an authoritative statehood framework – and Mohammed Rashid, his own adviser, later described it as a strategic mistake. This article’s argument rests on a narrower claim than Palestinian blamelessness – that the diplomatic process repeatedly offered autonomy or conditional statehood rather than genuine sovereignty, and that Israeli refusals and non-engagements were omitted from the ledger entirely.

1937

Some versions of the list begin in 1937, with the Peel Commission’s partition proposal. What they do not record is that the British subsequently withdrew the proposal themselves, after the Woodhead Commission concluded it was unworkable. They also do not record that significant Zionist factions rejected Peel too – Revisionist Zionists and religious Zionists among them – on grounds that will sound familiar: the proposed Jewish state was too small, too exposed, economically constrained, and severed from places they regarded as part of their national inheritance. Rejecting the partition of land you regard as your own is not uniquely Palestinian irrationality – it is how national movements characteristically respond when outsiders propose to divide what they claim.

1947

The 1947 entry is more substantial. The UN partition plan was accepted by the Jewish Agency and rejected by the Arab leadership, and that is accurate as far as it goes. But the plan gave a Jewish population comprising roughly a third of the Mandate’s inhabitants – and owning around seven percent of its privately held land, with much of the remainder classified as public domain – a state covering 56 percent of the territory, including the fertile coastal plain. The Palestinian Arabs, the majority population, were being asked by an external body to accept the permanent partition of the land in which they lived. Where historical partitions have endured they have been settled through coercion, war, displacement, elite bargain, or exhausted resignation, not cheerful consent.

Even the Jewish acceptance was not the unified green tick the table implies. The Jewish Agency accepted partition, but most of its leaders did so, in Avi Shlaim’s words, “with a heavy heart”. They did not like the idea of an independent Palestinian state, they were disappointed by Jerusalem’s exclusion, and they had serious doubts about the viability of the Jewish state within the UN borders. The Irgun and the Revisionist Right rejected partition outright or treated it as a temporary tactical stage, with Menachem Begin – a future prime minister – declaring the Jewish people would never be bound by the UN vote and proclaiming the intention to fight for a state on both sides of the Jordan.

More significantly, David Ben-Gurion’s own acceptance was explicitly provisional – tactical and contingent on power, demography, and opportunity rather than a settled acceptance of the UN map as permanent. In a letter to his son Amos written in October 1937, when the Peel Commission had proposed an even smaller Jewish state, he had already set out his intentions: “Erect a Jewish State at once, even if it is not in the whole land. The rest will come in the course of time. It must come.” By 1947 he accepted the UN partition plan, but Shlaim documents that “he did not accept as final the borders it laid down for the Jewish state”. During the 1948 war he proposed to the cabinet a major military offensive to capture large chunks of the West Bank, defeated only by a tied cabinet vote, and called the failure to proceed “a cause for mourning for generations to come”, while simultaneously rejecting an Egyptian peace feeler offering de facto recognition without mentioning it to the cabinet.

The Bernadotte assassination belongs in this context as evidence of Ben-Gurion’s strategic patience and that Jewish acceptance of the 1947 partition was contested from within. Count Folke Bernadotte, the UN mediator, was killed in September 1948 by the Stern Gang – a triumvirate that included Yitzhak Shamir, another future prime minister – because Bernadotte‘s proposed modifications to the partition plan would have required Israeli territorial concessions. His assassins were never brought to justice, and Ben-Gurion subsequently spent months in a desert retreat at Sde Boker in the company of Yehoshua Cohen, one of the men who carried out the killing. Ben-Gurion absorbed the Stern Gang into the IDF for the same reason he had crushed the Irgun over the Altalena incident – uncontrolled paramilitaries threatened the international legitimacy his longer strategy depended on. A state first, recognised and armed, was a better platform for what came next than a maximalist grab that lost the support of the great powers. These episodes show Ben-Gurion controlling the pace, not accepting the UN map as permanent.

Nor did the war that followed simply enforce the UN map after Palestinian rejection. Plan D, the Haganah’s operational blueprint, authorised the capture of Arab villages and cities – something the Haganah had not previously attempted – and was part of the military process that produced mass expulsion, flight, and village destruction across the country. Israel expanded from the 55 percent allocated by the UN to 79 percent of mandatory Palestine through military and political choices made by the stronger party.

A related argument claims that Jordan controlled the West Bank and Egypt controlled Gaza from 1948 to 1967, yet neither created a Palestinian state, so why single out Israel? The question deflects rather than answers, and the answer has two parts. Jordan’s control of the West Bank was itself the product of a tacit Israeli-Jordanian understanding reached before the 1948 war. On 17 November 1947, twelve days before the UN vote, Golda Meir secretly met King Abdullah and reached a broad understanding: he would take the Arab part of Palestine and attach it to his kingdom; the Jews would establish their state; and after the dust settled, the two would make peace. Egypt’s administration of Gaza was similarly not on Palestinian behalf – it ran Gaza as a military zone without any move toward Palestinian sovereignty. Shlaim calls the outcome “the partition of Palestine at the expense of the Palestinians”. Palestinian attempts to assert sovereignty – including the All-Palestine Government established in Gaza in September 1948 – were suppressed by those same Arab states. Ben-Gurion’s own position was explicit: “We are not contractors for the construction of an independent Palestinian state”. Israel was thus a direct participant in the arrangement that prevented it. That Jordan and Egypt also denied Palestinians statehood for their own reasons for 19 years does not address whether Israel has done so for the 59 years since 1967. Rather, it establishes that Palestinian statehood has been consistently prevented by every powerful party in the region.

1967

The 1967 entry attributes to Palestinians a position taken by the Arab League at Khartoum. The PLO existed from 1964, but Palestinians had no independent negotiating seat at Khartoum, and no Palestinian leadership was being offered sovereign statehood by Israel. The ‘three no’s’ were an Arab state coalition position issued after a regional war, not a Palestinian response to a statehood offer. The PLO’s 1964 founding charter rejected any Jewish state, a maximalist position that was itself a form of rejectionism, though one that preceded the occupation and was formally revised in 1993. However, Khartoum was an Arab League document, not a PLO one, and recording it in the Palestinian column conflates two different actors and two different moments. Meanwhile Israel’s own post-war planning moved in the opposite direction: the Allon Plan proposed Israeli control of the Jordan Valley, parts of the Judean desert, and Greater Jerusalem, leaving the rest of the West Bank as an autonomous region economically linked to Israel, not as a Palestinian state. Recording Khartoum in the Palestinian column while omitting the Allon Plan from the Israeli one perfectly encapsulates the table’s method.

The Camp David 1978 Framework

The underlying problem with the table becomes clearest when you step back from the individual entries to ask what kind of diplomatic architecture was actually on offer across these decades. The answer requires going back to 1978, before Camp David 2000 entered the picture.

Menachem Begin’s formula, articulated clearly at Camp David 1978, was autonomy for people, not sovereignty over territory, meaning Palestinians could administer their daily affairs, but the land itself would remain under Israeli control. Begin told Carter directly: “Had we thought that out of autonomy a Palestinian state would arise we would never have suggested it”. His director-general Eliyahu Ben-Elissar later confirmed the outcome in straightforward terms: Egypt got Sinai, and Israel got “the exercise of single sovereignty, Israeli sovereignty, over the territory between the Mediterranean and Jordan.” From the river to the sea. The phrase has since become associated with Palestinian maximalism, but in 1977 it was Likud’s founding platform and the explicit basis of Begin’s autonomy formula: Israel claimed sovereignty over all of it, so Palestinians could have administration but not sovereignty over any of it. There was no tension between the autonomy offer and the territorial claim – they were the same policy. Begin said “we are speaking of autonomy, not sovereignty, not a state”. He was not simply anti-peace: he made a full peace with Egypt, including withdrawal from Sinai because Egypt was a sovereign state whose territory Israel had taken and whose recognition Israel wanted. Thus the autonomy plan that Israel introduced at Camp David 1978 – and that the US and Egypt ultimately accepted as a framework – was explicitly designed to preclude Palestinian statehood while providing the appearance of a political solution.

This architecture did not disappear. It became the template for the autonomy talks of 1979–82, and elements of it reappeared in Oslo. When Yitzhak Rabin described the intended permanent Palestinian entity in his Oslo II ratification speech as “less than a state”, he was working within a framework whose limits had been established fifteen years earlier.

The table records no Palestinian concessions, only Palestinian rejections. The largest Palestinian concession came when the PLO accepted a state on the 1967 lines, reducing its claim to roughly 22–23 percent of historic Palestine. That acceptance preceded the 2000s by a decade and does not appear in the table. Similarly, the PLO’s acceptance of Oslo was the Palestinian leadership accepting terms narrower than their own Washington delegation had sought. The standard criticism of Palestinian leadership over Oslo is that they were given statehood and fumbled it, but the more accurate criticism is that Arafat accepted a framework that explicitly deferred sovereignty, allowed settlement expansion to continue, and left Israel in control of the territory being negotiated – and signed it anyway.

Camp David 2000

Camp David 2000 is the centrepiece of the rejection narrative. Ahron Bregman’s account, drawing on transcripts and participant records, shows that Ehud Barak moved further than any previous Israeli prime minister – on territory, on Jerusalem, on the willingness to countenance Palestinian sovereignty in parts of the Old City. However, the process was closer to forcing a reckoning than making peace. Patrick Tyler documents Barak’s strategy as an attempt to push Arafat into a final negotiation and, if he did not accept, declare that Israel had no partner. However, Arafat was not in a position to provide a definitive response either way. The Palestinians had wanted more months of preparatory talks, but Barak had not allowed it and had not implemented confidence-building commitments before the summit convened.

Early in the proceedings Barak killed the American written text, telling Bill Clinton “the paper no longer exists”, after which all discussions were oral with total deniability on anything proposed. Barak largely avoided substantive face-to-face negotiation with Arafat, saying he did not want Arafat to have a record, and thus preferred indirect channels through which positions could be floated and disowned. The summit then collapsed over Israeli sovereignty on the Haram al-Sharif, the Temple Mount compound containing the Dome of the Rock and al-Aqsa mosque. Barak would not relinquish it and Arafat could not concede it without Arab backing that had not been secured in advance, partly because Barak’s secrecy meant Arab leaders could not be briefed.

When Arafat asked for time to consult his leadership, Clinton treated the request less as a normal diplomatic step than as evasion, and the pressure shifted rapidly toward a yes-or-no reckoning that undermined the process.

Robert Malley co-authored with Hussein Agha a detailed rebuttal in the New York Review of Books arguing that the ‘generous offer’ framing was a post-hoc political construction. Malley was Clinton’s special assistant, directly involved in the talks, and Agha was a Palestinian academic with close ties to the Palestinian leadership and long involvement in back-channel diplomacy. Shlomo Ben-Ami, who was Israel’s chief negotiator, said in a 2006 Haaretz interview: “If I were a Palestinian, I would have rejected Camp David as well”. When the Israeli chief negotiator contradicts the rejection narrative, the narrative has a problem that more vivid memoir accounts cannot resolve.

The Clinton Parameters

The Clinton Parameters of December 2000 are a different matter, and the table’s failure to distinguish them from Camp David is one of its more significant distortions. The Parameters went further than Camp David on every core issue – including Palestinian sovereignty over the Haram – and represent the closest thing to an authoritative statehood framework. They were read orally rather than provided as a written document, because Barak feared Arafat would pocket any written text as an opening position, and Clinton wanted deniability should the proposals fail. On Bregman’s reading, Arafat’s failure to give the unambiguous ‘yes’ that Clinton demanded was a genuine missed opportunity, and Rashid later called it a strategic mistake. Shlaim’s interpretation complicates that verdict, observing that both sides responded with reservations. Israel’s own acceptance was heavily qualified by a twenty-page document, while Arafat’s qualified response was subsequently recast as outright rejection – partly, Clinton later told Saeb Erekat, the chief Palestinian negotiator, to preserve Barak’s domestic position. The failure was genuine, but the “Arafat rejected statehood” version of it is too simple.

Taba 2001

Taba in January 2001 appears in the table as another Palestinian rejection. Miguel Moratinos, the EU’s special envoy to the Middle East, produced a summary of where the parties stood at Taba, an unofficial document that became the main public record of the talks. The Clinton Parameters fed directly into Taba, hence why Taba was closer to agreement than at any previous point, with the parties’ own closing statement saying they had never been closer. Taba collapsed under a dying mandate: Barak was days from an election he was expected to lose, the talks had no durable political backing, and any understanding reached there would almost certainly have depended on a Barak Government that was about to disappear. Taba ended on 27 January, nine days before Sharon won the election on 6 February. There would be no serious negotiations for another seven years.

2008

The Ehud Olmert entry requires the most imaginative reading of the word ‘rejection’. Months of talks culminated in a September 2008 meeting at which Olmert showed Mahmoud Abbas a map but would not leave it with him. Abbas asked to study it and consult his team – the response any negotiator would give before formally responding to a complex territorial proposal – and Olmert refused. Abbas said he would return with his negotiators for formal talks, but Olmert resigned weeks later under corruption investigations and the Netanyahu Government that followed repudiated the Olmert framework. Olmert himself has said publicly that Abbas did not reject the proposal but asked for time to consult.

The fair criticism of Abbas is that he neither responded with a counter-proposal nor made any effort to put Olmert’s offer on the record before the Olmert Government collapsed, thus leaving the most serious proposal he had received to disappear with the administration that made it. Although that is a legitimate failure of Palestinian diplomacy, it is not the same as rejecting a signed, written, implementable statehood agreement.

So, at the decisive moments in 2000, 2001, and 2008, there was no formal written Israeli proposal with agreed maps that Palestinians could have signed or refused. Positions were advanced orally, incrementally, and with deniability built in. The ‘generous offer’ narrative depends on that absence of a written record, so that subsequent recharacterisation of what was actually proposed could not be checked.

2020

The Trump plan was drafted entirely without Palestinian participation, after the Palestinian leadership had boycotted the process from 2017 following the US recognition of Jerusalem as Israel’s capital. It offered the vocabulary of statehood while leaving Israel in control of borders and airspace, treating settlement blocs as permanent, and offering areas such as Abu Dis – outside Israel’s expanded municipal core – as a putative capital rather than sovereign East Jerusalem. Its political function was less to secure Palestinian agreement than to provide diplomatic architecture for the Abraham Accords, which decoupled wider Arab – especially Gulf – normalisation from prior Palestinian statehood in a way that broke with the logic of the 2002 Arab Peace Initiative.

The table contains no Israeli rejections, which is the clearest indication of what kind of document it is. The Arab Peace Initiative of 2002 offered Israel full normalisation with the Arab world in exchange for withdrawal to the 1967 lines and a negotiated refugee settlement, yet Israel did not seriously engage it as a negotiating basis, and when it was reaffirmed in 2007 it was again declined. Sharon ignored or circumvented the Geneva Initiative, the Ayalon-Nusseibeh discussions, and the Saudi peace plan. The Palestinians accepted the 2003 Road Map without reservations while Israel announced fourteen reservations and then pursued unilateral disengagement as a substitute for negotiation rather than a complement to it – Sharon refused negotiations, acted unilaterally, and sought US backing for retaining major West Bank settlement blocs. None of this appears in the table.

The argument the rejection narrative eventually reaches is that something is better than nothing – Palestinians should have accepted imperfect offers because they now have less. This is hindsight presented as reason.

The fundamental question is whether the offers on the table would actually have produced a sovereign state with contiguous territory, control of borders and airspace, and real self-determination;

or

a variant of Begin’s 1978 formula: autonomy for inhabitants, sovereignty retained by Israel, and legitimised by Palestinian signature.

Abba Eban’s line – “Palestinians never miss an opportunity to miss an opportunity” – assumes a continuous Palestinian agency making rational choices to refuse achievable statehood across eight decades. Bregman, in his introduction to Cursed Victory, notes that what Eban said of the Arabs “could well apply to Israeli Governments”. The slogan begins as wit but hardens into a theory of history, surviving by flattening every distinction the record actually contains.

Previous articles in the series:

Part VII – Reflections in the Mirror

Part VI – The West Bank Divide

Part V – The Weapons of Faith

Part IV – The Long War Logic

Part III – Hamas and the Social Architecture of Resistance

Part II – Hamas and Anti-Colonial Insurgency

Part I – The Label Problem: Terrorists or Insurgents

In parallel there is also a series specifically focussed on Israel:

Israel Part X – Uti Possidetis Juris

Israel Part IX – Land Claims

Israel Part VIII – Belligerent Occupation

Israel Part VII – The “Most Moral Army?”

Israel Part VI – Long-Term Strategy: Labour vs Likud & Reckless Genocide

Israel Part V – Reckless Genocide, Expanding the Frame

Israel Part IV – October 7, Genocide in Reverse?

Israel Part III – The Unravelling Defence, the Talking Points Collapse

Israel Part II – The Genocide Question

Israel Part I – What Does Genocide Mean, and What About Gaza?

REFERENCES

Anziska, Seth. Preventing Palestine: A Political History from Camp David to Oslo. Princeton: Princeton University Press, 2018.

Bakkour, Samer. The End of the Middle East Peace Process: The Failure of US Diplomacy. London and New York: Routledge, 2022.

Bregman, Ahron. Cursed Victory: A History of Israel and the Occupied Territories, 1967 to the Present. New York: Pegasus Books, 2015.

LeVine, Mark. Impossible Peace: Israel/Palestine since 1989. London and New York: Zed Books, 2009.

Pedahzur, Ami. The Triumph of Israel’s Radical Right. New York: Oxford University Press, 2012.

Rabinovich, Itamar. Waging Peace: Israel and the Arabs, 1948–2003. Updated and revised edition. Princeton: Princeton University Press, 2004.

Ross, Dennis. Doomed to Succeed: The U.S.-Israel Relationship from Truman to Obama. New York: Farrar, Straus and Giroux, 2015.

Shindler, Colin. The Rise of the Israeli Right: From Odessa to Hebron. Cambridge: Cambridge University Press, 2015.

Shlaim, Avi. The Iron Wall: Israel and the Arab World. Updated and expanded edition. New York: W. W. Norton, 2014.

Slater, Jerome. Mythologies Without End: The US, Israel, and the Arab-Israeli Conflict, 1917–2020. New York: Oxford University Press, 2021.

Tyler, Patrick. Fortress Israel: The Inside Story of the Military Elite Who Run the Country – and Why They Can’t Make Peace. New York: Farrar, Straus and Giroux, 2012.

The KPMG audit scandal has generated the usual cycle of resignation, apology, and parliamentary inquiry. The CEO has gone, the COO has stepped down from her executive role, ASIC has commenced a formal investigation, and current and former partners face a parliamentary hearing on 19 June. The story is being told as one of misconduct, partners allegedly competing by misusing confidential Lendlease board papers to support audit bids for Westpac and Dexus, with related allegations involving other major clients. The more interesting question is what they were fighting over.

An audit opinion is a signal. Markets and regulators require the signal to exist, which gives it value regardless of what sits behind it. The fee – around $75 million a year in the case of the Macquarie mandate, around $32 million for Westpac – is payment for a credible signature, but the credibility is imputed by the system that requires the signature to exist, not demonstrated by users inspecting the work behind it. Investors, lenders, boards, and regulators do not assess audit quality directly. They rely on the fact that an authorised firm has performed the process. The product is the acceptability of the signature. KPMG’s partners were allegedly competing by misusing confidential information for the right to sell a service whose institutional value is clear but whose assurance value is much harder to demonstrate.

I spent a brief period early in my career at Price Waterhouse audit before moving into credit and risk work, and that distance is why I can say what follows. Current auditors cannot easily say it, because doing so risks undermining the credential on which their careers depend. Recent ex-auditors who moved into CFO or finance director roles built their reputations partly on the same foundation. The profession’s incentives make public candour about its limitations unlikely, which is why the testimony rarely surfaces. The public record, however, does not require anyone to speak. Enron, HIH, Wirecard, Carillion, and now KPMG say it plainly enough. What the insider experience adds is the mechanism, the explanation of how these outcomes are generated by ordinary practice rather than unusual failure.

The basic production unit of an audit is a team of people in their early- to late-twenties, most of whom have worked nowhere but audit, checking a company’s accounts under time pressure imposed by fee budgets. The work consists of selecting samples, requesting documentation, and applying procedural tests designed on the assumption that the population being sampled is broadly honest and that documents reflect the transactions they purport to record. The methodology can detect sloppiness reasonably well. It is less well equipped to detect intelligent concealment, and it is largely blind to the class of problem where the document is real but the underlying transaction is not. Audit standards themselves acknowledge the methodology offers no reliable protection against structured fraud.

At the end of this process sits a senior manager or partner whose experience is also predominantly audit. Audit expertise encompasses standards, sampling, materiality thresholds, documentation, internal controls review, file defence, and engagement management. However, these skills are not the same as knowing whether a bank’s loan book is deteriorating, whether an insurer’s reserves reflect commercial reality, whether a construction company’s revenue recognition is honest, or whether a payments company’s reported Asian business actually exists. The audit skills may be entirely genuine and still be badly matched to the epistemic task the opinion claims to perform.

The methodology compounds this. Reasonableness tests – back-of-envelope calculations used to check whether numbers fall within an expected range – are taught as an analytical tool and used in practice largely as a completion device. If the numbers can be made to look reasonable by adjusting the base assumptions, they are recorded as reasonable. A former colleague told me that when his reasonableness test actually identified a problem that was then investigated and confirmed, the partner on the audit job remarked that it was the first time he had ever seen one work. He was not embarrassed. He was impressed. The anomaly was that the tool had performed its nominal function.

Errors, when found, can be managed into immateriality rather than forcing correction of accounts that are wrong in parts. The audit standards that require material uncorrected misstatements to be communicated to those charged with governance leave significant room for judgement about what is material and what offsets what. I once found a multi-million dollar error at a major client. The audit manager and senior identified enough errors running the other way that the net difference fell below the threshold requiring the accounts to change. The process was managed on a whiteboard. The objective was not to determine whether the accounts were accurate, but to find a path to signing them.

Documentation practices add another layer. A prior-year workpaper creates an institutional record that carries the weight of prior professional judgement. A paper noting ‘contract sighted’ with a tick implies the contract was sighted. In one example, I went looking a subsequent year for the contract – as the process requires – and was told it had never existed. The arrangement had always been oral. The original tick was simply wrong, and it had carried that false assurance forward until someone actually looked.

The deepest problem is not any of these individual practices but the social dynamic they collectively produce. When a discrepancy needs resolution, the path of least resistance is to construct the most plausible innocent explanation available and present it to the client as a question – ‘does that sound right?’ Clients who understand the audit process – and experienced finance teams do – return a reasonable-sounding answer knowing the auditor needs one. The auditor typically lacks the operational knowledge to assess whether the explanation is genuinely plausible rather than merely defensible. The file closes. Neither party has been dishonest in any direct sense. Both have performed the ritual correctly.

This is not a story about bad auditors. The people I worked with or interacted with were not dishonest or careless in any ordinary sense. They were operating within a set of incentives that rewarded completion over accuracy and made findings that extended engagements beyond budget professionally awkward. Audit is not fake work. It is real work organised around producing a defensible signature, but the assurance it offers is overstated relative to what the work can actually deliver.

The historical record confirms the pattern. Arthur Andersen signed off HIH Insurance’s accounts showing net assets of around $900 million; within months the company had collapsed with an estimated deficiency of between $3.6 and $5.3 billion. The Royal Commission found that Andersen’s independence had been compromised through a long and close relationship with HIH management, with former Andersen partners occupying the roles of HIH chairman, chief financial officer, and non-executive director. Arthur Andersen subsequently gave Enron unqualified opinions until weeks before the largest corporate bankruptcy in American history to that point, and was so implicated in the failure that the firm itself ceased to exist within a year, ending an 89-year-old institution that had audited some of the largest companies in the world. Wirecard received unqualified opinions from EY for nearly a decade while fabricating large portions of its reported revenue and the €1.9 billion of cash that was central to its balance sheet. The German parliamentary investigation found that EY failed to detect clear fraud risk indicators, relied on verbal assurances from executives on key questions, and did not request account confirmations directly from the banks where Wirecard claimed to hold its deposits, instead accepting certification from an intermediary. Carillion received KPMG’s approval for nineteen consecutive years, during which KPMG collected £29 million in audit fees and never qualified the accounts, before the company collapsed into the largest trading liquidation in UK history.

These are outliers in scale, not in kind. The dynamics that produced them – file management over verification, client relationship over epistemic rigour, the pressure to find a path to signature – are the ordinary dynamics of the audit process. The large collapses are the cases where concealment eventually became impossible. The question of what the same dynamics produced elsewhere, at a size that never forced disclosure, is unanswerable from outside the files.

The standard defence – that sophisticated fraud can mislead even diligent auditors – has some force. Deliberate and well-resourced concealment genuinely can defeat a sampling-based review. But in the cases above the gaps were not always subtle. HIH’s insurance provisioning and capital position had become visibly troubling to market and industry participants well before the collapse. Wirecard’s Asian business, which existed on paper to generate half the company’s revenue, would have unravelled under direct bank confirmation. The question is not only whether the fraud was sophisticated but whether the audit, as actually conducted, was designed to find what it was claiming to look for.

The profession’s response to each failure follows a consistent, well-trodden path: commitment to higher standards, tighter independence rules, mandatory rotation requirements, and increased regulatory oversight – all operating within the same basic production process, leaving the underlying incentive structure untouched. The auditee pays the auditor. Every reform measure is an attempt to manage a conflict that the fee relationship continuously recreates.

The alternative is to treat audit as what it actually is – a public function – and organise its production accordingly, in the same way that financial regulation, food safety inspection, and environmental monitoring are not left to firms paid by the entities they oversee. If the audit opinion exists to protect investors, creditors, and the broader financial system, then producing it is a public responsibility. A statutory audit authority, funded by a levy on listed and larger private entities, could allocate audit teams by sector, employ specialists in banking, insurance, and complex financial instruments, rotate staff internally to prevent the long relationships that compromised Andersen at HIH, and report to Parliament or a dedicated regulator rather than to the company being audited. The Australian National Audit Office demonstrates that audit can be organised as a public accountability function rather than a commercial service. Extending that model to corporate audit is a larger undertaking, and a public authority would face its own pressures – bureaucratic inertia, resourcing constraints, and the revolving door between regulator and regulated. However, these are design problems, none of which is insurmountable, not reasons to prefer an arrangement that has now produced Enron, HIH, Wirecard, Carillion, and KPMG in the space of a generation.

The partner who was amazed that a reasonableness test had worked was not describing an isolated experience. He was describing the normal state of affairs, without embarrassment, because within the profession it required none.

Daniel Yergin is one of the most authoritative voices in global energy. His book The Prize remains the definitive history of the oil industry, and his Pulitzer reflects a career of serious analytical work. Experts are not infallible, however, and his piece in the Wall Street Journal this week gets the diagnosis half right and the conclusion wrong.

He is right that the world has more sources of oil than it did in the 1970s, and right that diversification has moderated the first-round shock. Where the argument fails is in treating moderation as resilience. They are not the same thing, and the difference between them is where the real analysis starts.

Yergin’s case rests on what he calls the Churchill axiom – “safety and certainty in oil lie in variety alone.” He traces the diversification gains of the past half-century: the shale revolution transforming the United States from the world’s largest oil importer to its largest producer, and Western Hemisphere output now exceeding what the Middle East produced before the crisis, with Canada, Brazil, Guyana, and Argentina all adding supply. The Saudi East-West pipeline moving 7 million barrels a day to the Red Sea. The IEA coordinating strategic reserve releases. The implication is that the architecture built after the 1970s crises is doing its job. In the narrow first-round sense, that is true. Diversification has moderated the shock, but it has not proved the system resilient.

What this misses is where the vulnerability went.

Rather than eliminating its hydrocarbon dependence after 1973, the world restructured it. Oil was partially displaced by gas, and gas became a critical industrial feedstock for Europe and Asia in a way oil never fully was. LNG is now one of the central fuels for electricity generation in Japan and South Korea, and a significant industrial input across continental Europe. LNG exports from Qatar and the UAE through Hormuz represent almost one-fifth of global LNG trade. That supply has no pipeline bypass, no strategic reserve equivalent, and – for buyers under long-term Qatari contracts – no immediate alternative. Yergin cites Japan’s pivot to LNG as one of the diversification successes of the post-1970s era, without noting that LNG itself has become exposed to the same chokepoint logic the pivot was meant to escape. Qatari LNG exports have been operating under severe constraint since early March, with force majeure declared from Ras Laffan. Wood Mackenzie – an S&P Global company – assessed the strikes on Ras Laffan as removing approximately 19% of global LNG supply from the market and described the damage as fundamentally reshaping the global gas outlook. Yergin is vice-chairman of S&P Global, yet its analysts reached the opposite conclusion.

There is a further problem with the pipeline diversification Yergin cites. The Saudi East-West pipeline moves oil west to Yanbu on the Red Sea, but barrels leaving Yanbu still face the southern Red Sea and Bab el-Mandeb risk environment, where Houthi attacks since late 2023 have raised war risk insurance costs, reduced vessel availability, and complicated routing. The bypass shifts the chokepoint from Hormuz to a different contested waterway rather than eliminating it.

LNG shows the immediate flaw in the diversification story: supply was diversified away from oil, but not away from chokepoints. The wider problem is that the same half-century also stripped out the industrial buffers that once slowed energy shocks before they reached final output.

The post-1970s industrial restructuring – outsourcing, just-in-time logistics, and offshoring of petrochemical-intensive manufacturing – reduced the energy line on corporate income statements while embedding hydrocarbon exposure more deeply into supply chains. Companies appear less oil-dependent because they consume less energy directly, yet their inputs, components, and logistics are often more exposed to energy price shocks than the headline figures suggest. Ammonia and fertiliser production, plastics, diesel freight, and shipping costs are hydrocarbon-linked channels through which an energy shock moves into food, manufacturing, and distribution long after the crude price itself has stabilised. The buffers that existed in a 1970s industrial economy – company inventory holdings, domestic manufacturing capacity, and slack in logistics networks – were engineered out as deliberate efficiency gains. A 1970s economy absorbed shocks through inventory drawdown before they reached the production line, but the 2026 economy transmits them almost immediately.

Yergin also invokes US shale production growth as a diversification success – and it is, in volumetric terms. But US shale produces predominantly light sweet crude, and refineries configured around medium and heavy sour Gulf grades cannot substitute one for the other without yield, margin, and logistics penalties. Senior industry executives, including Chevron and ExxonMobil’s, have flagged oil prices well above current levels as the inventory and recovery arithmetic becomes apparent. More barrels do not help if they are the wrong barrels for the refineries that need them.

Yergin’s treatment of strategic petroleum reserves (SPR) follows the same pattern – correct in institutional terms, incomplete in current application. The SPR was designed as a counter-cyclical instrument: release inventory during supply shocks, then refill gradually in well-supplied markets where additional government demand does not move prices. That cycle depends on disruptions being short and sharp, with a clean return to normal supply. What is happening now inverts the design logic entirely. Reserves have been drawn down at record rates during the disruption – the US SPR is back near levels last seen in the early-to-mid 1980s. When governments eventually refill, they will be buying into a market simultaneously absorbing three other demands: commercial inventory rebuilding, Asian strategic and commercial reserve restocking after months of severely reduced imports, and supply still recovering along the curve Sultan Al Jaber, CEO of ADNOC, described – 80% of pre-war Gulf flows at four to six months after reopening, full normalisation not until Q1 or Q2 2027. Strategic reserve replenishment adds a fourth source of demand competing for the same recovering barrels at precisely the moment political pressure to rebuild visibly is highest. Instead of dampening the price cycle as designed, SPR replenishment amplifies it. The instrument becomes pro-cyclical rather than counter-cyclical.

The recovery arithmetic undermines his price scenario. Yergin writes that if the strait opens soon, prices will come down – then, three paragraphs later, cites S&P Global’s own analysis that it would take up to six months to return to 80% of pre-crisis supply after reopening. He states the evidence but does not follow it to its conclusion. If the recovery ramp runs to six months or beyond, the inventory drawdown continues through the recovery period, commercial and strategic restocking compete for constrained supply, and prices do not normalise on reopening. They normalise only when the recovery ramp, commercial restocking, and strategic reserve replenishment are absorbed, which is not the scenario the current futures price is pricing, and not what “prices will come down” implies to a reader.

Yergin argues that markets contribute to energy security by adjusting faster than governments, citing the 1970s gas lines as a failure of price controls rather than of the underlying supply. That is partially correct as history, but the wrong lesson for a physical shortage of this scale. Market adjustment under severe supply constraint means price rationing – prices rise until demand falls to match available supply – and at this level of disruption, what the market clears is economic activity. Governments across Asia are already rationing, issuing work-from-home directives, and prioritising domestic consumers over industrial users not because they misread the 1970s but because the alternative is pricing ordinary households and manufacturers out of energy entirely. Demand destruction is not resilience.

The post-1973 system diversified supply, but it also reduced inventory, redundancy, and domestic industrial slack. It made the energy system broader, but not more shock-resistant. Diversification shifted the source of oil; it did not diversify away from chokepoints, did not restore the grade flexibility that heavy sour Gulf crude provided, did not create an LNG strategic reserve equivalent, and did not anticipate a recovery curve long enough to turn its own counter-cyclical instruments pro-cyclical. Yergin is right about diversification – and wrong about what diversification proves.

When oil futures prices drop, the financial press reaches for a familiar sentence: the market expects prices to fall. When the futures curve slopes downward – near-term contracts at $87, 2030 delivery at $65 – the implication feels obvious. Traders are betting on cheaper oil ahead.

The direction runs the other way, and getting it wrong has consequences in the current oil market.

What a futures contract actually is

A futures contract is an agreement to buy or sell a fixed quantity of something at a fixed price on a fixed future date. Many commodity futures are physically deliverable – crude oil contracts can, in principle, oblige the holder to take or make delivery of actual barrels – though most positions are closed out or rolled before delivery arrives, so most traders never exchange the physical commodity.

The exchange requires both sides to post margin – a deposit that absorbs daily gains and losses as the contract is marked to market. This is the mechanism that keeps the system solvent without anyone needing to trust their counterparty. Hull’s Options, Futures, and Other Derivatives uses gold futures to illustrate: an investor buying two December gold contracts at $400 per ounce posts $4,000 initial margin, and each day the account is adjusted by the difference between yesterday’s settlement price and today’s. The contract is, in effect, closed out and rewritten at the new price every day.

The investor who bought July corn futures in March can sell an offsetting contract in April. The physical corn never moves. Two parties have exchanged the price risk for a period – one taking the gain if prices rose, the other taking the loss. For physically deliverable contracts, delivery arbitrage helps enforce convergence – the futures price must eventually meet the physical price, because someone can always choose to take or make delivery if the gap is wide enough. For Brent, the link is indirect: expiry settlement is tied to the ICE Brent Index and the broader physical Brent complex, with EFP and cash-settlement mechanisms connecting futures to physical pricing.

What the curve shape means

At any given moment, futures contracts for different delivery dates trade at different prices. Plotting those prices against delivery date gives you the futures curve – the term structure of futures prices.

You need to know two shapes:

Contango – futures prices rise with delivery date, an upward-sloping futures curve. Near-term cheap, long-dated expensive. This shape is often associated with surplus conditions, easy access to storage, or weak prompt demand. The price difference between near and far delivery reflects the ‘cost of carry’: storage, financing, and insurance. If you buy oil today and store it for six months, you need the future price to cover those costs, otherwise there is no incentive to hold inventory.

Backwardation – futures prices fall with delivery date, a downward-sloping futures curve. Near-term expensive, long-dated cheap. This shape is often associated with tight prompt conditions and high convenience yield – the operational value of having physical barrels available now, to keep a refinery running, fulfil a cargo, or avoid a shutdown. The barrel in six months is cheaper because the scarcity signal does not extend that far.

The formula connecting these is:

Futures price ≈ Spot price + carry costs − convenience yield

When convenience yield is low and carry costs dominate, the curve tends toward contango. When convenience yield is high – when refineries need barrels today and cannot substitute a futures contract for physical supply – the curve tends toward backwardation. Neither shape is a pure forecast. Both also reflect hedging pressure, seasonality, producer selling patterns, and risk premia, but the convenience yield term is usually the decisive driver of large swings between the two shapes.

The inversion: futures converge to spot, not the other way around

Here is where the common misreading occurs.

If the Brent futures curve shows near-term contracts at $87 and 2030 delivery at $65, the intuitive reading is: the market expects oil at $65 by 2030. Prices are coming down. The curve is the forecast path.

That intuition assumes spot follows the curve – that the spot price travels down toward where deferred futures are priced today. However – and this is the misunderstanding – the direction runs the other way. Futures contracts converge toward spot at expiry. The deferred contract does not pull spot toward it; it rises (or falls) to meet wherever spot is when expiry arrives.

Consider this example. Spot crude is at $100 today, six-month futures at $90. The casual reading says: market expects oil at $90 in six months. But suppose six months pass and physical conditions have not changed – spot is still $100. The six-month futures contract, now the prompt contract, does not settle at $90. It settles at approximately $100, wherever physical is trading at expiry. The contract has risen toward spot. A trader long that futures contract has earned a positive return – positive roll yield – even though spot did not move.

This is why backwardation can be supportive for long futures holders: if spot conditions hold up, convergence creates a positive roll tailwind. The return comes from the contract rising toward physical, not from correctly forecasting a price increase, though whether a long position is actually profitable depends on whether spot holds – the roll tailwind is favourable carry, not a guarantee.

The convergence diagram follows from this. Draw a line for spot, roughly flat. Draw a second line for a deferred futures contract, starting below spot, curving upward (backwardation) to meet spot exactly at the contract’s expiry date. The futures contract does not predict where spot will be – it settles where spot is.

Applied: the current oil market

The following section moves from mechanics to a specific market view. Readers who want the mechanics without the application can stop here.

The Brent curve has moved from steep backwardation – near-term contracts at $140+ in March, long-dated at $80-90 – toward near-flat, with near-term contracts around $87 and long-dated around $65-70 (as at time of writing). The standard read is that the market has priced in resolution of the supply disruption, deal optimism having built through successive ceasefire extensions and Rubio’s declaration that Operation Epic Fury was “concluded”, and the paper market having repriced accordingly.

The physical market has not moved the same way. Dated Brent – the price of actual physical barrels for near-term loading – has remained at a substantial premium to futures throughout this period. The US SPR is drawing at the fastest weekly rate in its 40-year history, Japan’s strategic reserve is experiencing its largest ever drawdown, US Big Four storage is below every comparable prior year, and global onshore crude excluding China has dropped to multi-year lows. The inventory buffers that have been absorbing the disruption are nearly exhausted: the IEA coordinated release expires imminently, and a Morgan Stanley buffer analysis puts the 70% reduction in strategic draw support arriving in July.

The physical scarcity that creates backwardation – high convenience yield, low inventories, operational necessity – has not resolved. What has happened is that a large speculative position has pushed the paper curve toward a shape that reflects the US-Iran negotiations arriving at a deal on schedule, before it has arrived or delivered flows.

A backwardated market is not, by itself, telling you that spot will fall to where deferred futures trade today. The contract will converge to the spot price prevailing at expiry – if prompt scarcity persists, that means the deferred contract rises toward physical; if the physical market loosens before expiry, both may move lower. The question is which condition holds when the July contract expires, and the physical inventory data argues for the first – futures price increasing toward spot. The short calendar spread positions built on deal optimism face not just a directional loss but the carry cost of being short in a backwardated market where roll works against them if physical stays elevated.

Bill Winters did not mean to be revealing. When he described the Hong Kong core banking migration as replacing “lower value human capital” with financial and investment capital, he was simply being honest. That is how the hierarchy conceives of the people inside it. The phrase escaped because it is the operating assumption, not a slip.

The problem is not that executives are wrong to distinguish between levels of contribution. Of course organisations have people who contribute more and less. The problem is that the value hierarchy they have constructed typically bears little relationship to the capability hierarchy that actually exists, and remuneration has drifted so far from any grounded assessment of contribution that it now exists mainly to justify itself.

The Gap That Isn’t

Corporate hierarchies treat executive value as self-evidently superior, and not by a small margin. They have to because the compensation calculation only works if the distance is vast – the CEO is not merely better than the next person down but categorically different, operating on a plane that mere professionals cannot reach.

The actual capability and experience gap between executives and middle-level professionals is often much smaller than remuneration implies. The gap between those middle-level professionals and junior ranks, by contrast, is enormous. The people with the deepest operational knowledge and the most granular understanding of how things actually function, are clustered in a band of the organisation that is paid a fraction of what sits above them. Remuneration inverts the capability gradient rather than tracking it.

This is not a new observation, but it has acquired new urgency because the AI displacement argument has given executives a language for accelerating the inversion while describing it as modernisation.

The Pipeline Problem

Yes, AI can replace many tasks currently embedded in lower-paid roles, and in many cases do them better. However, the issue executives are not addressing is where experienced professionals come from.

Individual firms can hire experienced professionals, but the corporate sector as a whole cannot. Someone has to develop them – through junior roles, accumulating operational knowledge that only comes from doing work at the level where consequences are immediate and visible. The mid-level professional who understands how a system fails, who can read a client, and who knows which risks are real and which are theoretical, acquired that knowledge somewhere. They acquired it in the roles now being described as lower value and automated away.

Hollow out the junior and middle layers and you do not just cut costs. You destroy the apprenticeship pipeline. The capability loss is invisible for years, and then suddenly it isn’t – when the people who actually understood how things worked have gone, and the executives who replaced them with AI discover that synthesis and coordination require something to synthesise and coordinate from.

The Substitution Nobody Discusses

Here is the claim that does not get made in polite company: a large share of senior executive work – synthesis, coordination, governance process, stakeholder presentation, board reporting – could be absorbed by experienced middle-level professionals working with AI, at a fraction of the cost, with far less loss of organisational capacity than the executive class would care to admit.

The savings would be substantial, but more importantly, two things would likely improve.

The first is the capability pipeline. If the people doing the synthesis and coordination work are drawn from the operational layer rather than parachuted in from the executive class, they carry real knowledge of what they are synthesising. The abstraction problem – senior executives operating at such a remove from actual work that their judgements are essentially uninformed – does not disappear, but it shrinks.

The second is risk. Executive incentive structures are options-heavy, weighted toward short-term share price, and the asymmetric nature is intentional. The intended effect is to encourage risk-taking, because the upside flows to the executive and the cost of failed strategic bets spreads across shareholders, employees, and in systemic cases, everyone else. This can reward behaviour that damages the organisation’s long-term interests, while presenting it as shareholder alignment.

Replacing a layer of options-compensated executives with experienced professionals on salary does not eliminate risk appetite, but it removes the artificial incentive to manufacture it.

Why This Conversation Doesn’t Happen

The boards that would need to authorise this substitution are drawn from the same cohort applying the same logic to themselves. Independent directors are typically former executives. Remuneration committees benchmark against peer groups, which ratchets compensation upward regardless of performance – each company paying above median to signal quality, the median therefore rising, the cycle repeating. The circularity is institutional rather than merely social. It is not a conspiracy, just a closed system that has every incentive to remain closed.

Bill Winters is not unusual. He is representative. The “lower value human capital” framing was not a gaffe to be walked back with a transcript – it is the revealed preference of a structure that has been compounding its own logic for decades, and has now found in AI a vocabulary for continuing to do so while sounding like it is looking after everyone.

The question worth asking is not whether lower-value roles can be automated. It is who decided which roles were lower value, by what measure, and in whose interest that assessment was made.

FT Alphaville put the essential point in one sentence in March: most oil traders don’t trade oil. The benchmarks everyone cites – Brent crude, WTI – are futures prices. A futures contract is an agreement to buy or sell a commodity at a fixed price on a future date, and the traders who buy and sell these contracts mostly never take delivery of a barrel. The vast majority of futures volume settles in cash. What these markets are trading is not oil but a bet on the likelihood that oil will be available at a given price on a given date. Price volatility in these markets, as Alphaville noted, is “a measure of how difficult it is at any given moment to suppress the knowledge that no one knows anything”.

Understanding this is crucial right now, because the deal – the Iran-US memorandum of understanding (MOU) to end the war that has severely disrupted flows through the Strait of Hormuz (roughly one-fifth of global oil consumption) and begin negotiations on Iran’s nuclear programme and sanctions – is being priced into the futures market in a very specific way. Brent crude futures have fallen from approximately $117 to $87 over the past two months.

Robin Brooks has made a specific and apparently evidentially grounded claim: futures were right. Paper prices anticipated the ceasefire and the deal, the premium for immediate barrels fell, and Dated Brent – the real-world benchmark for physical North Sea crude – then fell toward futures rather than the reverse. The collapse in the spread between physical and paper prices proved, on his account, that the acute physical scarcity argument was overpriced. He has set a target of $85 for Brent on a peace deal and argues the market is rapidly approaching it. That interpretation gets the direction of the April price move correct, but it over-reads what that move proves.

The forward-looking logic behind that conclusion is correct in principle. A futures contract for July delivery is not priced for today’s shortage but for what traders expect conditions to be in July. If they expect the deal to reopen Hormuz before then, the July contract is priced for a post-deal world. That reasoning is sound, yet it rests entirely on the timing assumption being correct, and on correctly understanding what the shape of the futures curve is actually measuring.

Boston Consulting Group analyst Jamie Webster has set out the key distinction. Backwardation – when near-term delivery prices are higher than prices for future delivery – measures immediate physical scarcity, not optimism about future stability. As Webster explains, “the greater the need for barrels today, the stronger the downward slope in the curve becomes”, creating pressure to release oil from storage and deliver immediately. It does not mean the market expects prices to fall; it means physical buyers are paying a premium to secure supply now – each available barrel, in the formulation of Gianluca Benigno, professor at the University of Lausanne, “begins to trade less like a commodity and more like insurance”. When arguments suggest a “near-total collapse” of that backwardation, they show that the insurance premium has fallen, not that the physical shortage has resolved. Insurance premiums fall when confidence about future delivery rises. The deal optimism priced into the futures curve is the mechanism by which the premium compressed, and that is not the same thing as proof that the deal will deliver on the schedule the curve is pricing.

The question the curve cannot answer is whether that confidence is warranted. Martin Wolf, writing in the Financial Times on 20 May, put it directly: “Yes, the oil futures markets suggest that prices are set to fall and so all will be well. But the oil futures curve is not a crystal ball, as my colleagues Jonathan Vincent and Malcolm Moore have noted.” FT’s Vincent and Moore stated that lower prices for crude to be delivered at a later date do not mean traders are predicting a sharp fall. The academic literature is equally clear. Ron Alquist and Lutz Kilian, studying two decades of oil futures data, found that futures-based forecasts are no more accurate than simply assuming the current spot price will persist, at any horizon up to twelve months. BP’s chief economist stated the practitioner view simply: “We cannot forecast oil prices with any degree of accuracy over any period whether short or long”. The futures market’s record of pricing normalisation before it arrives – and being surprised when it doesn’t – is not reassuring.

The physical market tells a different story to the paper one. During the acute phase of the crisis, before the April ceasefire and the deal optimism that since compressed the backwardation, several refined products – diesel, jet fuel, naphtha – traded above $200 per barrel, according to Webster’s analysis. Clyde Russell, writing for Reuters, observed that Dubai cash-settled cargoes were trading at a wide premium to the equivalent futures swaps – anecdotally, up near $160-170 – and that paper traders appeared “more confident than oil consumers about a quick ceasefire and manageable supply disruptions” in a way that “seems untethered from the reality on the ground”.

US distillate inventories – diesel, heating oil, and jet fuel precursors – stand at their lowest level since at least 2003 in the Bloomberg/DOE series, 24% below the long-run average. Neil Chapman, Senior Vice President of ExxonMobil, said at a Bernstein conference this week that inventories are approaching “unheard of” levels and that once they hit their floor, “dated Brent, a model would say, will shoot up to $150, $160.” His timeline is two to three weeks.

A market pricing $87 oil is implicitly assuming that physical normalisation arrives before inventory floors become binding. The physical data does not disprove that assumption, but it makes the assumption a very large one. The insurance premium fell because confidence in future delivery rose, not because the physical shortage resolved. The distillate inventory level, the refined product premiums during the acute phase, the massive drawing of strategic reserves, the JPMorgan forecast of the approaching operational floor, and Chapman’s observation all suggest the inventory floor is reached before the normalisation the futures curve is pricing. If that sequencing is right, the confidence that drove the premium lower proves misplaced and the premium returns. The buffers that have been absorbing the supply shortfall – the Strategic Petroleum Reserve, coordinated IEA emergency releases, commercial inventory drawdowns, demand management in Asia – are approaching exhaustion. Once they cease, the premium for immediate barrels cannot continue to fall. Physical buyers who need actual barrels cannot pay with probability weights on a future deal.

Brent futures do not float free of the physical Brent complex. As contract expiry approaches, the paper price has to converge on the physical market through the settlement and expiry mechanism, and that convergence can run in either direction. In April it ran downward: Dated Brent, S&P Global Platts’ daily physical benchmark assessment for prompt North Sea crude, fell toward futures as deal confidence rose. If the deal has not delivered actual barrels by the next settlement window, that same convergence mechanism runs upward, toward wherever the physical benchmark is trading when contracts expire. The April convergence that Brooks correctly identified as closing from the top works symmetrically. If the confidence that compressed the insurance premium proves misplaced, the premium returns and convergence runs the other way.

Sultan Al Jaber, CEO of ADNOC stated that even an immediate opening of the Strait does not restore full pre-war flows until Q1 or Q2 2027, and that it would take at least four months to return oil flow to 80% of pre-war levels. He also told the Atlantic Council that the world drew down approximately 250 million barrels from storage in just two months and that 30 to 35 days of effective cover remains. That is a clear articulation of the specific interval between the current futures price and the physical reality.

Whether the deal delivers on the timeline the $87 futures price implies is a question the curve itself cannot answer. That answer comes from ship traffic through Hormuz – still near zero – from inventory levels, and from the operational timelines that Sultan Al Jaber and Chevron CEO Mike Wirth have described. All of those point in the same direction: the insurance premium fell faster than the physical reality justified.

Postscript

Yet again we have an announcement that the US and Iran are nearing an agreement to sign the MOU. At best that is a deal to agree to a deal to agree to defer the discussion about a deal after another ceasefire that extended the multiple previous ceasefire extensions.

After the announcement Trump backed off again, saying he wasn’t ready to sign yet, while Tehran stated that the terms announced by the US were nothing like the terms it had agreed to. Deja vu.

Fresh US strikes on southern Iran on Tuesday confirmed the pattern this series has been tracking: successive deal announcements, each one spending a little more of the credibility the next needs to move markets. Jeff Currie – former head of commodities research at Goldman Sachs – counted five announcements with zero closed. Appearing on CNBC Asia on 26 May as WTI July futures traded down 4.3% on deal optimism during the Memorial Day session, Currie named the dynamic: ‘sell the tweet, buy the molecule’. Within hours, the US strikes sent Brent back toward $100 and the drop reversed.

The molecule is the actual barrel awaiting passage – 240 ships were waiting for Iranian permission to transit the strait as of Saturday, and the IRGC Navy reported 33 ships passed in the preceding 24 hours with permission. That works out to roughly 2-4 mb/d of selectively managed flow rather than the 20 mb/d of pre-war transit the market periodically prices as imminent.

The deal

The MOU remains unsigned. Mojtaba Khamenei has reportedly endorsed the broad template – the first indication that the missing arbitration problem may be resolving itself, though it remains a US assessment rather than a confirmed Iranian commitment. Nevertheless, endorsement has not produced a document, and a senior US official confirmed on Sunday that the Iranian system “did not move fast enough”.

Beyond the unsigned document, the two sides continue to describe different agreements. The US version has the strait reopening with no tolls, Iran clearing the mines it deployed, and verbal commitments on enrichment suspension and removal of the highly enriched uranium stockpile during a 60-day period. The Iranian version has Iran continuing to manage the strait, the nuclear file absent from the current phase, frozen funds released upfront, and the naval blockade lifted first.

A new blocking issue has emerged since Saturday: Lebanon. Israel wants language in the MOU preserving its right to conduct military operations there against “any threat”, while Iran is insisting on a sustainable ceasefire with no such carveout and has told all mediators it will not sign unless every clause is agreed and guaranteed. Netanyahu confirmed to Trump he expects that right preserved and Ben Gvir’s statement that Israel will block any deal it dislikes adds a layer to the Israeli veto analysis. The veto is now operating inside Israel’s domestic politics as well as through Washington – Ben Gvir is telling Netanyahu explicitly that the coalition will not hold under these terms. The Lebanon carveout Israel is demanding is a coalition management necessity for Netanyahu.

Mines extend the ramp-up timeline beyond whatever date an MOU is signed. Iran deployed mines in the strait, and the deal would require Iran to clear them before full passage resumes – a process measured in weeks rather than days, pushing supply recovery further toward Sultan Al Jaber’s Q1/Q2 2027 estimate for the return of pre-war flow levels.

Iran has also been systematically degrading bypass alternatives throughout – SAMREF was struck on 19 March, the UAE’s Ruwais refinery shut by drone strike, and Habshan hit the same day – so the bypass capacity available when the deal eventually takes effect is lower and less reliable than it was at the outset.

The clock

A chart and a table published this week quantify the approaching moment that BCA Research puts at early July – when “the pain from the oil shock will really start to rip through Western economies”.

Bloomberg’s Javier Blas distributed a chart showing the US Strategic Petroleum Reserve (SPR) weekly drawdown from 1985 to the present. The current draw rate of approximately 10 million barrels per week is the largest in the reserve’s 40-year history, dwarfing every previous release including the Gulf War and the 2022 Biden response. The reserve is approaching 365 million barrels – approximately two weeks from its lowest level since 1983, when it was first being filled. The significance is not that the SPR is about to run out but that it is approaching the floor of its entire operational history during an active supply crisis, and drawing further below that floor has no precedent and no established policy guidance.

HFI Research published its regular EIA crude balance table, for the week ending 22 May: total US demand at 151.99 mb/w against total supply of 151.59 mb/w, producing a net deficit of 0.40 mb/w, with the SPR contributing 9.10 mb/w of that supply. The SPR draw is not restoring equilibrium but buying time while a residual deficit persists that current prices are not fully clearing. When the SPR contribution reduces to zero – as it eventually must – the weekly deficit widens to 9.50 from 0.40 mb/w, leaving only price to do all the clearing work.

The EIA table measures the US domestic crude position, but the global picture is considerably tighter. The full global headline shortfall with the strait closed is 13-15 mb/d. After offsets, such as SPR draws, stockpile releases, bypass, and floating storage, BCA Research analysis puts the global residual shortfall at 7-9 mb/d, but Brent around $100 is not fully pricing even that – it reflects a market that still has inventory drawdowns absorbing part of the gap. When the SPR contribution reaches zero at the cliff, that residual gap widens from 7-9 mb/d to approximately 8.3-10.3 mb/d at a foreseeable date, regardless of what happens diplomatically in the interim, and the step-change will not be gradual. This is what Currie means when he says Iran’s leverage increases with every day that passes and inventories decline.

After the deal

Santos chief executive Kevin Gallagher, presenting to investors on Tuesday as the latest strikes broke, offered the clearest corporate expression of where energy economics settle regardless of when the deal is signed. Santos operates at a breakeven of $45-50 per barrel, generating $550-600 million in free cashflow for every $10 above that, with 60% returned to shareholders. It is explicitly positioning its non-Hormuz routing as a competitive advantage while projecting robust oil demand through 2050, treating higher energy prices as a planning assumption rather than a crisis premium. Since a deal appears not to be imminent and the terms under negotiation preserve Iranian management of the strait, that assumption is probably well-founded.

The most favourable exit-path scenario from Pricing the Press Release ($120-160) assumed prompt and full reopening, with flows at 20 mb/d. However, the deal as it actually appears – gradual reopening, mine-clearing required, Iranian management continuity, nuclear question deferred, Lebanon unresolved – produces something closer to 6-8 mb/d of effective flow. That is roughly 2-4 mb/d of Iranian-managed selective passage plus 3.5-4 mb/d of bypass capacity, itself partially degraded by strikes, leaving a shortfall of 12-14 mb/d still to clear. That puts the settlement between the exit and middle paths, at approximately $140-180, before accounting for the inventory rebuild demand that will run against still-constrained supply through the second half of the year. The markets do not reflect this yet – generally oscillating between $90 and $110 – but the SPR drawdown absorbing much of the supply shock has four to five weeks of capacity remaining at current draw rates.

The market has a simple model of this conflict: a deal means Hormuz opens and Hormuz opening means oil prices fall toward pre-war levels. Both halves of that equation are wrong, and the events of 23-24 May demonstrated the error across 48 hours in compressed form.

On Saturday 23 May, Donald Trump announced that a deal had been “largely negotiated” and would be announced shortly. Oil prices fell. By late Saturday the terms had leaked, revealing that the two sides were describing different agreements – Iran’s version required $370 billion in upfront US concessions, full Iranian sovereignty over the Strait, and zero nuclear commitments before a 60-day negotiation period could begin, while the US version included nuclear commitments and deferred sanctions relief. An unnamed White House official said the deal was “95% done”. Yet, the remaining 5% turned out to be about 95% of the issues neither side can agree on. By Sunday, Trump was posting bomb imagery on Truth Social under reported pressure from Israel and domestic hawks, Iran declared the agreement “completely cancelled” due to US obstruction, and the deal that never existed was publicly collapsing.

The pattern is now on its eighth iteration since March. Each de-escalatory signal produces a price decline; the physical draw on global inventories continues regardless; the strait remains effectively closed to most commercial shipping; and within hours or days the price recovers. The market has learned nothing from the previous seven cycles, and the reason is that it keeps pricing the press release not the number of barrels in transit.

What it should be pricing is the arithmetic of physical shortage, and that arithmetic produces oil prices well above current levels under every scenario available, including the most optimistic one. The deal on the table, such as it was, placed us in the exit scenario at best, and the physical ramp back to pre-war oil flows takes months, not days, even under immediate resolution.

The physical picture first. Before the conflict, the Strait of Hormuz carried roughly 20 million barrels per day of crude oil and petroleum products – about 20% of global consumption which the IEA put at just over 103 mb/d in early 2026. There are bypass routes, but they are limited. Saudi Arabia’s East-West Petroline and the UAE’s Abu Dhabi Crude Oil Pipeline together offer an estimated 3.5 to 5.5 mb/d of capacity – operational infrastructure built specifically for this scenario, currently running below capacity, but with ramp-up constrained by Red Sea terminal logistics and tanker repositioning rather than pipeline readiness. Iraq’s Kirkuk-Ceyhan route adds some additional volume, though it carries uncertainty given its history of political and operational disruptions. Four countries have no bypass infrastructure at all: Iran, Kuwait, Qatar, and Bahrain. A generous estimate has bypass absorbing 5-7 mb/d of the gap, leaving a shortfall of roughly 13-15 mb/d, or 12-15% of global consumption. The IEA called it the largest supply disruption in the history of the global oil market.

The natural response is that higher prices will destroy enough demand to clear the market – the orthodox solution economists reach for when supply falls short. This is correct as far as it goes, but it goes less far than usually assumed when the supply shock is large enough, and oil demand is extraordinarily inelastic in the short run.

Price elasticity of demand measures how much quantity demanded falls for a given rise in price, with a value of -1.0 meaning a 10 per cent price rise reduces demand by 10%. Oil is nowhere near that. The price elasticity of US demand for oil is often estimated to be around -0.05 in the short run. The Federal Reserve’s survey of thirty academic studies puts the consensus short-run demand elasticity at -0.13, with estimates ranging from -0.03 to -0.30. James Hamilton’s widely cited work finds the implied elasticity in the 1979-81 episode at around -0.21, and notes that more recent studies suggest demand has become even less responsive to price, somewhere between -0.034 and -0.077 for the 2001-06 period.

The reason is that the bulk of oil demand is captive. Industrial feedstocks, freight, aviation, and power generation in markets without immediate alternatives cannot reduce consumption in response to price within any relevant short-run timeframe. The slice that actually responds – discretionary driving, some industrial substitution, efficiency switching – is a small fraction of total demand, so price has to rise very far to destroy enough of it to balance the market.

The required price increase depends on which elasticity estimate applies. At the upper bound of -0.21 – the most responsive demand scenario – a 12-15% shortfall requires a price rise of 57-71% from the pre-war $69, producing a clearing price of approximately $108-118. At the consensus of -0.13, the required rise is 92-115%, producing approximately $133-149. At the lower estimate of -0.05 – the most inelastic scenario – the required rise is 240-300%, producing approximately $235-276. The arithmetic therefore establishes a clearing price range of roughly $108 to $276, with the consensus pointing to approximately $133-150.

Demand destruction works as a market-clearing mechanism when the shortfall is modest. A 2-3% supply reduction requires a proportionally modest price increase, demand falls by the required amount, and the economy adjusts to a new stable price level. The adjustment is painful but contained. A 12-15% shortfall requires something categorically different. The clearing prices that the arithmetic produces are theoretical under standard assumptions – they model what price would balance supply and demand if demand could fall smoothly down the elasticity curve. In practice, at these price levels, the demand being destroyed is not discretionary spending that households choose to reduce but economic activity that ceases to be viable: industrial processes cannot absorb energy costs that have doubled or tripled, airlines cannot cover fuel costs at fares passengers will pay, and agricultural supply chains dependent on diesel and fertiliser become uneconomic. Supply does not recover during this process – full pre-war flows remain many months away even under immediate resolution, and while demand destruction is eliminating economic infrastructure month by month, the physical shortage persists.

The analogy is with the difference between a common virus and Ebola. In a mild infection, the immune response contains the pathogen before organ damage becomes irreversible. In Ebola, the virus destroys organs faster than the immune system can respond – eventually so many organs are destroyed that the virus has nowhere left to replicate and dies, but by that time the damage is catastrophic and much of it permanent. A 12-15% supply shock is not a mild infection that the demand-destruction mechanism resolves cleanly. The price levels that would theoretically clear the market are price levels at which the economy is already in serious difficulty. The cure and the disease become indistinguishable.

The historical record illustrates that relationship between oil price and demand. In October 1973, the Arab oil embargo removed approximately 4.5 mb/d – around 7% of world consumption at the time. The price of oil went from $2.90 to $11.65 by January 1974, a near-quadrupling. In 1979, the Iranian Revolution removed a slightly lower share of global supply, and prices more than doubled over 1979-80, from around $13 to over $35 per barrel. Both episodes resulted in dire economic conditions and stagflation lasting nearly a decade, yet both removed less supply, as a share of global consumption, than the current Hormuz closure removes after bypass. They generated price responses that look extreme relative to the percentage of supply removed because inelastic demand means small shortfalls produce large price moves. The difference today is the existence of strategic reserves and bypass routes, which moderate the immediate price response relative to those episodes, but do not change the underlying clearing price implied by a supply removal that is proportionally larger.

There are short-term offsets that are currently suppressing the price below the arithmetic range. The IEA’s coordinated release of 400 million barrels from strategic reserves amounts to roughly 6-7 mb/d of additional supply spread over two months – meaningful, but finite and non-renewable at scale. OPEC+ pledged 206,000 barrels per day of additional output, less than 2% of the shortfall. US shale can respond but on a 6-12 month lag. Strategic reserves exist to cover weeks to months of disruption, but not to change the fundamental clearing price if the physical shortage persists – they buy time but do not restore equilibrium.

The scenarios already set out distribute across the arithmetic range based on duration and the degree to which supply recovery supplements demand destruction as the market clears.

The most favourable exit path, with Hormuz reopening by June or July and supply beginning to recover, sits at the optimistic end of the arithmetic range at $120-160 – well above where the market trades when it prices a deal. The resolution premium that pushed prices down to $99 on Saturday was pricing pre-war conditions, not exit-scenario conditions. Sultan Al Jaber, CEO of ADNOC, confirmed that even this path does not restore full pre-war flows until Q1 or Q2 2027, and that it would take at least four months to return oil flow to 80% of pre-war levels. A June announcement produces approximately 20 mb/d of flow by November, still 5 mb/d short of the pre-war baseline, with the global inventory rebuild adding demand against still-constrained supply through the second half of the year.

The middle path, with no full opening until late July or August, and resumption gradual rather than clean, falls above the consensus at $180-240 – consistent with elasticities at the pessimistic end of the range as demand destruction carries more of the clearing burden for longer, and financial amplification adds a risk premium on top of physical scarcity. The middle path remains the base case as of 25 May: the MOU is unsigned, the core disputes are unchanged, and the 60-day framework, even if agreed, defers rather than resolves the questions on which the physical constraint depends.

The worst path – closure persisting into the second half of the year, with Houthi activation of Bab el-Mandeb, attacks on submarine internet cables, and sustained strikes against Gulf energy infrastructure – reaches the lower bound of the arithmetic range and beyond it at $260-320, consistent with the lower -0.05 elasticity estimate plus financial amplification on top of physical scarcity.

The deal on the table, such as it is, places us in the exit scenario at best, and only if signed – which as of today it has not been. Sultan Al Jaber’s timeline runs to Q1/Q2 2027 from whenever a deal is actually signed. The 60-day nuclear negotiation period defers the question both sides describe as non-negotiable. The resolution premium that drove prices down on Saturday was pricing a restoration of pre-war conditions that no available deal actually delivers, which is why the market has been consistently wrong in both directions throughout – falling too far on deal signals and recovering too little when those signals dissolve.

The United States and Israel launched Operation Epic Fury on 28 February without a settled definition of victory.

Trump later declared it had already occurred “in the first hour”, said he would know when it was over when he “felt it in my bones”, and assured the public that “any time I want it to end, it will end”.

The stated objectives shifted almost daily: dismantling Iran’s nuclear programme, preventing uranium enrichment, obliterating its ballistic missiles, freeing the Iranian people, and regime change, announced variously as the point and then not the point. None of it resolved, and by 6 May US Secretary of State Marco Rubio told reporters “at this stage, we do not have nuclear negotiations” – an admission that the war’s stated justifications had been set aside entirely.

Iran’s response has been more effective than almost anyone anticipated. By closing the Strait of Hormuz and sustaining that closure under sustained military and economic pressure, Tehran has converted a defensive posture into an offensive one and in doing so has introduced a demand that did not exist in February: a de facto toll right over the strait modelled on Egypt’s Suez Canal. The Suez comparison is Iran’s own framing, though Suez is a canal rather than an international strait, and UNCLOS guarantees the right of transit passage through the latter regardless of coastal state wishes. Iran’s demand is not a sovereignty claim in international law but a jurisdictional toll backed by coercive force, with no precedent in the post-war order.

Iran had always held the geographic capability to disrupt the strait – the fast boats, coastal missiles, and mines were never in doubt, and the deterrent value was always there. What the war gave Iran was the leverage and the incentive to convert that capability into a permanent institutional claim. The Persian Gulf Strait Authority (PGSA) did not exist in February. It exists now because the war created the conditions for it: a closure that had to be sustained, infrastructure that had to be administered, and a jurisdictional precedent that, once established in legislation and published maps, cannot be easily set aside.

Mojtaba Khamenei, who assumed the Supreme Leadership following his father’s death in the opening strikes, used his first public statement to say that the leverage of blocking the waterway “must continue to be used”. Iran’s parliament has approved a tolling plan, the PGSA has begun operating and issuing payment requests of up to $2 million per vessel, and Iran has published a map extending its claims into UAE and Omani coastal waters. The deputy speaker of Iran’s parliament has described the Strait of Hormuz as “our nuclear weapon” – the explicit institutional confirmation that Hormuz has replaced the nuclear programme as Iran’s primary deterrence instrument. Iran has also signalled since late March its preparedness to disable submarine internet cables running through the strait, and if US strikes resume, to activate Houthi control of Bab el-Mandeb, which would close the second critical passage for global energy flows. The war that began over centrifuges has become a contest over who controls the world’s most critical energy and communications infrastructure.

The exchange of proposals through May confirmed the distance between the two positions rather than closing it. The US presented a 14-point memorandum of understanding in early May. Iran’s counterproposal was rejected by Trump as “totally unacceptable”. Iran’s chief negotiator issued an ultimatum – accept Tehran’s terms or face “failure”. As of 22 May, Iran said no agreement had been reached and vowed it would “never bow”. Both proposals are on the table and both have been rejected by the other side.

Seven de-escalatory interventions between late February and mid-May each produced a price decline that the reality on the ground then reversed. On 18 April, Iran’s Foreign Minister Araghchi declared the strait “completely open” for commercial vessels – oil dropped 12% and stocks surged before markets understood that Trump had simultaneously declared the US naval blockade of Iranian ports would “remain in full force”. Iran had declared its own closure lifted on the assumption the US would reciprocate, but with no such intention on the US side, the strait remained effectively closed and Brent reached $126 a week later. On 22 May, Al Arabiya published breaking news of an imminent agreement, oil prices fell, and Al Arabiya then deleted the tweets and denied they were authentic.

The legal architecture around the war had undergone its own parallel deterioration. On 1 May, Trump declared hostilities “terminated” in a letter to House Speaker Johnson – not because they were, but to reset the War Powers Act 60-day clock – while making clear he remained “open to future military actions” as “separate engagements.” On 19 May he cancelled a set of planned strikes citing “serious negotiations” and received a military options briefing from national security advisers the same afternoon. The threat, the legal fiction of peace, and the contingency planning had become simultaneous events, and the credibility of all three had been exhausted together.

Brent opened the crisis at $69 and sits around $105 at time of writing, 50% above pre-crisis levels despite the full weight of White House reassurance. That is no longer a panic spike – it is a market beginning to price the possibility that disruption has become the equilibrium condition. And yet, it remains too optimistic.

The gap is too wide

Rather than narrowing, the gap between American and Iranian negotiating positions has expanded and migrated to a different issue entirely.

As recently as 21 May, the sequencing logic was still intact: Iran treating Hormuz as Phase 1 – a preliminary concession that could give Washington something to call a win – while deferring the nuclear question to a later track. The PGSA has since closed that path. Iran is no longer offering Hormuz reopening as a tradeable concession; it has built permanent institutional infrastructure that converts what was previously a temporary military closure into a jurisdictional claim. A Phase 1 concession is something you can give back, but a legislature-backed toll authority with operating staff, issued invoices, and published jurisdictional maps is not. The war demoted the problem it was launched to solve and produced a harder one in front of it.

The US’s original minimum was zero enrichment and dismantlement of Iran’s nuclear capability. That minimum has not moved. What has changed is that Iran’s minimum has expanded to include the de facto toll right over Hormuz – a demand without historical precedent, since no coastal state has successfully imposed mandatory tolls on an international strait in the post-war order. Rubio’s position was unambiguous:

“That’s just not acceptable. It can’t happen. If that were to happen in the Straits of Hormuz, it will happen in five other places around the world”.

Iran’s parliament has legislated it. The G7 foreign ministers have insisted on safe and toll-free navigation. Both positions are in the institutional record and neither can retreat from it.

The nuclear question remains ever present beneath the Hormuz question, unresolved and deferred to a future negotiating track that does not yet exist. Iran requires enrichment rights as a minimum. The US requires zero enrichment. These have been the respective positions through two decades of negotiations and they have not moved. Iran’s enrichment programme is not primarily a weapons programme but a prestige project and a deterrence posture, maintaining latency sufficient to deter attack. Surrendering it means surrendering both the regional standing the programme confers and the deterrence posture that latency provides. Any successor leadership, regardless of factional alignment, faces the same dilemma – agreeing means giving up what the programme was built for.

The US side is bound by different parameters. The operation was to be a demonstration of dominance, but a settlement that preserves Iran’s enrichment rights, Hormuz leverage, toll revenues, and sanctions relief would be strategically inconsistent with the war’s justification, however imprecisely that justification was stated from day to day. The domestic political constraint is now visible in Congressional action as well as polling – on 21 May Republican leadership pulled a war resolution from the floor because the votes were not there, while 60% of US voters oppose further military action. The Trump-Netanyahu relationship has fractured over exactly this point, as Israel applies the same veto ‘The Deal That Keeps Not Happening’ identified – its direct interest in preventing any Iran deal that falls short of the original objectives – now against a deal Trump himself is trying to make.

Trump is now reported to be considering a “decisive” final military operation after which he could declare victory and end the war. This would be the third such declaration – Hegseth declared “a decisive military victory” on 8 April and Trump declared hostilities “terminated” on 1 May – and it carries the same problem as the first two. A strike, however large, does not open Hormuz. The IRGC maintains the closure passively, requiring only that it sustain its current position rather than take any active step. The morning after a “decisive” operation, the PGSA will continue issuing toll demands and the physical constraint on global oil supply will remain exactly where it was. Victory declarations change the US domestic narrative but not the physical constraint the market has yet to fully price.

No one can sign

The second obstacle is independent of the first and operates even if the gap somehow narrowed. The first round of talks in Islamabad collapsed because the Iranian delegation had to return to Tehran to get approval to sign anything. American officials described Iranian leaders as “of two minds” – a phrase that sounds like a hard negotiating position but is more accurately a description of an institutional problem. The arbitration function that converted competing Iranian pressures into binding decisions – exercised by Khamenei for four decades – is missing. Mojtaba Khamenei holds the title but not the accumulated institutional authority, and there is a further physical dimension: reported severe injuries mean real-time arbitration is not merely politically difficult but operationally impossible, the courier system – handwritten letters carried by human chain to his location – too slow to keep pace with the IRGC’s operational tempo. By the time a reply returns, the operational window has closed. The IRGC, the presidency, the clerical establishment, and the hardliner factions retain autonomy that no single actor can currently override.

However the IRGC’s tactical control of Hormuz does not require a functioning arbitration hierarchy – the capacity to maintain the closure is unified and has proved decisive across nearly three months. The fragmentation problem is specific to the commitment side: no actor can currently bind the IRGC to a concession it has not determined for itself, and without that commitment, any agreement that reopens the strait is not a deal but a statement of intent. Rubio’s formulation – that Washington needed Iran to show “the extent of the concessions they are willing to make at the front end” – was a description of the problem rather than a solution to it, and nothing in the external environment changes that.

Why ceasefires don’t hold

The ceasefire in effect since 8 April has not produced a stable equilibrium. Military skirmishes in the Gulf have continued – tanker seizures, IRGC interdictions, contested transit claims – and the PGSA has continued operating during the ceasefire, building the institutional infrastructure for permanent Iranian control of Hormuz rather than suspending it pending a deal.

The asymmetry that made the closure effective in the first place has intensified. Iran’s doctrine, which its commanders describe as “survive and exhaust”, requires nothing more than maintaining what it already holds – and when storage capacity fills, Iran shuts in production rather than opening the strait, absorbing the loss of export receipts, and waiting. With more than 200 billion barrels of proven reserves, Iran’s oil industry is not under long-term threat from storage constraints and is thus not an effective lever against the regime. The storage depletion deadline often cited as a forcing function for Iran falls in the wrong direction – it is a deadline for consuming countries where refineries run short of feedstock and industrial output contracts, not for the country that can simply leave its oil in the ground. The US position requires Iran to dismantle the PGSA, abandon its toll claim, defer its jurisdictional ambitions, accept no reparations, and surrender its most effective strategic instrument – in exchange for an agreement that defers the nuclear question to future negotiations, offers sanctions relief that previous experience suggests is reversible at the next change of US administration, and provides no guarantee against future military action.

If US strikes resume, Iran has signalled a prepared escalation package confirmed across eight weeks of IRGC statements, parliamentary declarations, and stated doctrine. This includes Houthi activation of Bab el-Mandeb, which would close the second critical passage for global energy flows and add approximately another 5-6 mb/d to the supply disruption already in place, alongside attacks on submarine internet cables running through the strait, and sustained missile and drone strikes against Gulf energy infrastructure. The IRGC warned of the cable threat on 28 March, Bab el-Mandeb has been signalled since late March, and Gulf infrastructure targeting has been the IRGC’s stated retaliatory doctrine throughout. Were it activated, the supply disruption would approach 25-30% of global consumption rather than the 12-15% the preceding scenarios have been modelling, with price dynamics in the $260-320 range established as the worst-path scenario.

As of publication, Trump’s self-imposed deadline – “two or three days, maybe Friday, Saturday, Sunday” – given when he cancelled the last set of planned strikes on 19 May, has arrived. Israel is reported to be preparing to rejoin US strikes. Iran has threatened wider war if conflict resumes. Congressional support for escalation has fractured. The ceasefire is in its most fragile state since 8 April, and the reason is the same as it has been throughout – neither side has the incentive to grant the concessions the other requires, and neither currently has the institutional architecture to make a binding commitment even if the incentive existed.

Brent crude is trading around $105 per barrel – a resolution-premium price reflecting a market assigning high probability to early diplomatic resolution and discounting the physical data almost entirely. The three paths from here differ not in direction but in duration, and duration is what determines whether the damage is severe or historic.

The favourable path has Hormuz reopening in June or early July, which is what current pricing implies, but reopening and restoration are different things. Sultan Al Jaber, CEO of ADNOC, stated publicly that Gulf oil flows would take at least four months to recover to 80% of pre-war levels even under immediate resolution – meaning a July reopening produces approximately 20 mb/d of flow by November, still 5 mb/d short of the pre-war baseline, and cumulative forfeited Middle Eastern production reaches approximately two billion barrels before normalisation, a volume the global inventory system cannot readily absorb.

The JPMorgan inventory analysis puts total visible global inventories hitting operational stress level – 7.6 billion barrels – in June, with the operational floor, below which pipelines and refineries begin to encounter feedstock difficulty, arriving in September at 6.8 billion barrels. A July reopening slows the draw but does not stop it, and Q2 corporate earnings, reported from July, will be the first illustration of the feedstock constraint appearing in industrial results rather than energy sector data alone.

The AI investment cycle bifurcates in this path in a way that determines how much of the damage the advanced economies absorb. Pre-war, the hyperscaler capex boom was adding an estimated 0.3 percentage points to annualised US growth – and if financial conditions stay loose enough to sustain it, that investment provides a partial offset, not enough to cancel the energy shock but enough to prevent the US tipping into outright recession. If the energy shock and the bond market sell-off tighten conditions sufficiently to interrupt the circular financing arrangements that underpin hyperscaler debt issuance, that offset disappears and the growth picture darkens considerably even under early resolution.

Price in this path: Brent in the $120-160 range, peaking as the inventory draw continues into June and July before declining through Q4 as flows recover. The market is currently below this floor, meaning even the best case involves prices rising from here before they fall.

The middle path has Hormuz not fully reopening until late July or August, with resumption gradual rather than clean, as tanker operators, insurers, and cargo owners will not return to normal liftings on the day a ceasefire is announced and mine clearance, cargo insurance reactivation, and infrastructure certification each add weeks to the ramp-up. The two billion barrel cumulative forfeit is reached and exceeded before flows recover, and the inventory floor approaches in September.

The price arithmetic – worked through briefly here and in detail in a subsequent piece – puts the market-clearing price for a sustained 12-15% supply shortfall in the $180-240 range, with the floor set by the optimistic end of the demand elasticity estimates and the ceiling reflecting financial amplification adding a risk premium on top of physical scarcity as the closure extends. The market at $105 is priced below even the most favourable scenario, and the gap to the middle path’s floor is nearly double the current price. That gap will not close gradually as sentiment shifts – the oil price will increase significantly and rapidly when the physical shortage can no longer be deferred by inventory draws.

The central bank bind tightens here independently of the energy price – inflation stays elevated on energy and feedstock costs while growth contracts, closing the standard policy exits simultaneously, since rate cuts add to inflationary pressure while rate holds deepen the recession. The stagflation configuration turns from a tail risk to a base case, and the policy timeline that would otherwise resolve it extends with every week the strait stays closed.

The same AI bifurcation applies in this path, but at a worse baseline, because financial conditions tightening from extended closure is the environment in which the circular financing arrangements the IMF’s April 2026 Global Financial Stability Report identified as fragile are least likely to survive. The hyperscaler capex boom, which pre-war was absorbing surplus global capital and driving technology investment across the sector, becomes a vulnerability rather than an offset. The IMF’s own scenario analysis suggests that if AI investment falters it removes one of the two remaining pillars – the AI investment tailwind and the market’s assumption of early permanent ceasefire – supporting current equity market levels.

Price in this path: Brent in the $180-240 range for an extended period. The Dubai cash market – which prices physical crude for immediate Gulf delivery – briefly hit $166-170 on 19 March, suggesting the lower boundary of this range has already been touched in physical spot markets while the forward market has not caught up.

The worst path has closure persisting into the second half of the year. This will result in a breach of the inventory floor and refinery feedstock stress appearing in industrial output. It will manifest not as a consumer energy price effect, but as a physical availability constraint on production, with price dynamics turning non-linear as inventories fall below working minimums and buyers compete for available barrels regardless of cost.

Price in this path: The $260-320 range follows from the arithmetic of a sustained physical shortage – at the lower estimates of how much demand falls when price rises, clearing a 12-15% supply shortfall requires a price increase of 240-300% from pre-war levels, and financial amplification adds a further premium once the inventory floor gives way.

The reason this path is not merely theoretical is that it does not depend on a negotiating gap – these close when the price is right, but a deal also requires someone with the authority to make it stick. That person does not currently exist in Tehran – the internal arbitration mechanism to translate any agreement into binding action is missing. This mechanism had converted competing institutional pressures from the IRGC, the presidency, the clerical establishment, and the remaining hardliner faction into unified decisions. It was exercised by Khamenei for four decades, but was removed when he was killed in the opening days of the conflict, nothing has replaced it, and it does not respond directly to external pressure or economic pain. The negotiating gap is also too wide to bridge regardless of institutional capacity – disagreements over nuclear and enrichment issues – but the arbitration problem means that even a narrower gap would not currently produce a deal that binds anyone.

The AI question is moot in this path – the shock overwhelms any potential offset.

The current data puts the market at a price below the favourable path while the physical reality runs on the trajectory of the middle one. Kpler shipping data shows OPEC+ crude exports at approximately 17.8m bbls/d in May against a pre-war baseline near 25.0m, with the 2026 export line falling entirely outside the 2022-2025 range. US total crude inventories – commercial and strategic combined – drew 17.8 million barrels in a single week, the largest weekly draw in more than forty years of recorded data. Mohamed El-Erian noted on 17 May that for almost three months the true scale of the energy supply deficit had been masked by an extraordinary drawdown of global crude stockpiles.

On 20 May, Brent dropped nearly 5% on a single Trump statement about negotiations being in their “final stages”, even as Hormuz remained physically closed and the weekly draw continued at record pace – the market updated on the signal rather than the condition, which is the resolution premium at work. That gap between financial pricing and physical reality closes in one of two ways: resolution arrives before the inventory floor forces it, or the inventory floor forces it regardless, and the JPMorgan timeline puts the forcing window between August and October.

I have spent considerable time looking for a reason not to make the case of $200+ oil price and a global downturn. I cannot find a compelling one.

Over several months I have been publishing a body of work arguing that the global economy was heading for something severe and prolonged, through mechanisms that standard policy responses cannot address. That is not a comfortable position. The default view is that the catastrophists are usually wrong and things will work out. I am aware of the analyst who has predicted eight of the past two downturns. Yet the argument I am making does not rest on temperament – it rests on mechanisms that can be examined, challenged, and tested against the data. And the data is not sympathetic to the optimistic view.

The structure of the argument is this. Several independent causal chains are running simultaneously, through different channels and with different transmission lags. A war that has closed the world’s most important energy chokepoint and damaged infrastructure that will take years to restore, not weeks. A financial system that amplifies shocks rather than absorbing them, through a global liquidity withdrawal transmitting through three simultaneous channels, leveraged positions in non-bank lending sectors, and a central bank bind that leaves rates too high for the slowing economy and too low for still-elevated inflation. A coercion theory – the misapplied US blockade – that does not apply to Iran’s fiscal and monetary architecture, which means the war ends when it ends rather than when Washington needs it to. A pre-existing global debt position stretched to the point where the IMF now projects government debt breaching 100% of GDP globally by 2029 leaving materially less room to respond than in any prior cycle. These are different problems that converge on the same outcome through independent routes, which is why it is categorised as overdetermined – resolving one does not rescue the overall case.

The supply calculation is not complicated. The Strait carries roughly 20% of global oil consumption. Bypass routes and demand reduction absorb perhaps 5–10% of that gap, leaving a structural shortfall of 10–15%. That does not translate into just a 10–15% price increase – oil demand is highly inelastic. In 1973, a disruption of 4-7% of global supply produced a near-quadrupling of price; in 1979, a disruption of 3-4% partially offset by compensation from other producers produced a doubling. A shortfall twice as large points to $200 as a floor, not a ceiling.

The range of outcomes this could theoretically produce runs from uncomfortable to historically unusual and hence economically severe. My base case is the middle path, whereby the Strait of Hormuz does not fully reopen until late July or August, the uncertainty makes resumption gradual rather than clean, and the price arithmetic puts Brent over $200 for an extended period with the economic consequences that implies.

In the most favourable path, Hormuz reopens by June, but the physical ramp back to pre-war flows takes months rather than days – Sultan Al Jaber, CEO of ADNOC, told The National that Gulf flows would take at least four months to recover to 80% of pre-war levels even under immediate resolution, leaving a residual shortfall of 5 mb/d beyond that. The global inventory rebuild will add demand against still-constrained supply, and the economic damage will accumulate longer than the diplomatic timeline implies, while AI investment provides a bifurcated contribution – either a partial offset or itself breaking under tightening financial conditions.

In the worst path, closure persists well into the second half of the year, non-linear price dynamics activate, financial amplification adds a risk premium on top of physical scarcity, and $300 Brent becomes possible rather than an abstraction.

The inventory arithmetic is less agnostic than the market-pricing question, and it does produce a date range. US total crude inventories – commercial and strategic combined – are drawing at 17.8 million barrels in a single week, the largest weekly draw in more than forty years of recorded data. Global observed draws ran at 120-130 million barrels per month through March and April, and the buffer above minimum working inventory levels exhausts in an August to October window at current rates. Even on the assumption that Hormuz reopens in July, cumulative forfeited Middle Eastern production reaches approximately two billion barrels before normalisation – a volume that cannot be absorbed without price consequences. Corporate Q2 earnings, reported from July, will be the first time the feedstock constraint appears in industrial results rather than energy sector data alone.

What this analysis cannot give you is the precise date on which financial markets acknowledge what the physical data has been increasingly showing for weeks. That gap can persist far longer than physical conditions alone would imply – Brent dropped nearly 5% yesterday on a single statement about negotiations, despite Hormuz remaining physically closed and the weekly draw continuing at record pace. But inventory data gives that gap a mathematical end point, and sets a window within which the physical constraint will operate directly on prices regardless of how markets are interpreting diplomatic signals.

The next part of this series will set out the scenario analysis in more detail – if you want the forecast without the detailed arguments and workings, that is the piece to read. The pieces that follow will work through each thread of the argument and the evidence behind it. The final piece synthesises them and will be updated as events develop.

Donald Trump said America was in “the final stages” of a deal with Iran. Oil prices slid. Iran’s president, Masoud Pezeshkian, said his country would not be forced into surrender and called for mutual respect in diplomacy. The IRGC threatened to extend the war beyond the region if America attacked again. Trump added that strikes remained on the table without a deal, then told reporters he was in “no hurry”. All of this happened within roughly 48 hours.

The market reacted to “final stages”. That shows you how this conflict is being read – and how it is being misread.

The pattern has been in place for weeks: deadline, retreat, optimism, repeat. It would be easy to read this as pure Trump – the theatrical dealmaker performing urgency he does not feel. Although that reading is not entirely wrong, it misses a more important development, which is on the Iranian side.

Tehran has separated the negotiation into two distinct tracks. One is the Strait of Hormuz: Iran is willing to reopen it as a preliminary concession, a gesture of good faith. The other is the nuclear programme, which Tehran is treating as a separate, slower, later negotiation. That sequencing is deliberate, and it is what the optimism of the past 48 hours obscures.

From Iran’s position, the logic is clear. Reopening the Strait relieves the economic pressure the blockade has imposed and gives Washington something to call a win – or at least a pause – without requiring Iran to make the concessions that are most valuable to it. Nuclear capability, including the enrichment levels that preserve what I have elsewhere called the Prestige Bomb option, stays off the table pending a fuller agreement that may never arrive.

This creates an asymmetric problem for Washington. Three gaps make a comprehensive agreement nowhere near close: the distance between the two sides on acceptable enrichment levels, Trump’s insistence on terms visibly superior to the 2015 Obama deal, and Tehran’s formal separation of the two tracks. A partial agreement – one that reopens the Strait while leaving the nuclear question unresolved – is the more likely outcome, and it suits Iran considerably more than it suits Washington, even as it gives Trump enough to declare victory, which is likely how an agreement gets reached at all.

The 1962 Cuban Missile Crisis offers a relevant precedent. The public deal – Soviet missiles out of Cuba, US non-invasion pledge – concealed a quiet side agreement to remove Jupiter missiles from Turkey months later, deniable as connected. Both sides got what they needed without having to say so. Whether the Iran context can sustain that kind of construction is a question for the series that follows – it requires principals who can commit and deliver, and Iran’s leadership architecture complicates that considerably.

The Strait is the hinge for everything that follows. Whether oil flows through it, and on what terms, drives the scenarios for markets, energy prices, and global credit conditions. The nuclear question is consequential over a longer horizon, but it transmits on a different timescale, and what is moving now is the Strait.

With gasoline approaching $4.50 a gallon and Brent above $100, the Trump Administration is signalling that other ideas are on the table. The idea most people are reading into that is an export ban – restrict or outright prohibit crude exports to keep more supply at home. It has circulated since the Hormuz closure began, and although the White House officially killed it in March after Vance met with oil executives at the American Petroleum Institute, it keeps reviving because it sounds intuitive. The United States is exporting roughly 6 mb/d of crude while prices are rising. Why not keep it?

Because the marginal exported barrel is not the barrel US refineries are designed to use.

US production is dominated by light sweet crude from the Permian – high API gravity, low sulphur, and relatively easy to process. That is the grade leaving the country at 6.28 mb/d. American refineries, however, were built around medium sour and heavy sour grades: the thicker, more sulphur-rich blends from Canada, the Middle East, and Latin America that require more complex processing but yield better volumes of gasoline and diesel. The refinery estate was built around the assumption that heavy crude would be available cheaply and that surplus light crude could be exported at a premium. That configuration cannot be changed in weeks or months, and US refiners running more light crude than they are optimised for face worse yields and compressed margins, not expanded production.

An export ban would therefore create a glut of light sweet crude that most refineries cannot efficiently absorb, while disrupting the economics that support continued imports of the heavier grades those refineries have actually been built to process. The predictable response from refiners is not to expand throughput but to cut it, because running sub-optimal feedstock at scale destroys margins. Less refinery throughput means less gasoline, which is the opposite of the stated objective.

There is a second mechanism working against the same goal. American consumers do not buy crude oil. They buy refined products priced in a global fuel market. Removing 6 mb/d of light sweet crude from seaborne international markets would tighten global supply and push Brent higher, and that increase would feed directly back into US pump prices. The ban would lower the domestic crude price while raising the global benchmark on which US fuel is ultimately priced, and the two effects would largely cancel, leaving consumers modest relief at best and probably temporary, while weakening domestic production incentives in the medium run.

The White House understood this in March. Oil executives made the same argument to Vance, and the administration came out of that meeting with a clear public statement: export restrictions are not under consideration. What appears to have changed is not the economics but the political temperature. At $4.50 a gallon, the pressure to be seen doing something is intense, and an export ban has the rhetorical advantage of sounding like America-first energy policy – keeping US oil for US consumers – even though the physical reality is that the US refinery system cannot efficiently use it.

This is the Goodfellas dynamic that has characterised much of this administration’s response to the Iran conflict: demonstration substituting for strategy. The demonstration has a target audience – voters who see cheap petrol as a right and blame whoever is in office when it disappears – but the mechanism does not deliver the outcome. The Administration has already exhausted most of the levers that do work: the SPR is being drawn down at record pace, shipping restrictions have been waived, and sanctions on Russia and Venezuela have been quietly eased. What remains is the one thing none of these measures can substitute for, which is reopening the Strait of Hormuz. Every other option – gas tax holidays, export bans, SPR releases – is a rounding error against the volume of oil not reaching the market because Hormuz is closed.

The conventional case for fiscal restraint rests on two claims. The first is that bond markets assess government creditworthiness, and rising yields signal deteriorating fiscal sustainability – “bond vigilantes” punishing profligate governments by demanding higher returns. The second is that deficit spending is inflationary and surpluses disinflationary. Both claims carry the same variable: the deficit number. Both are pointing at the wrong thing.

Look at the mechanism of the first claim and the bond vigilante story becomes a simple supply-demand process obscured by the faux language of discipline. When a government runs larger deficits, it issues more bonds. More supply reaching the same or thinner demand means lower prices and therefore higher yields. Commentary systematically overattributes these yield movements to solvency concerns when supply-demand mechanics explain far more. Calling this “market discipline” attributes intentionality and judgment to a clearing process. New issuance also degrades the value of existing holdings – when a new line becomes the on-the-run benchmark, existing bonds lose liquidity and price accordingly, so holders sell – not because their view of sovereign creditworthiness has changed but because the supply mechanics of the market have moved against them.

The Truss episode – UK PM Liz Truss in 2022 – is the standard evidence for the vigilante view, but it runs the opposite way to what the commentary claims. Yields on 30-year gilts spiked during her 49 days – proof, apparently, that markets punished fiscal recklessness. What actually happened was that an unfunded supply announcement hit a demand side already fragile from liability-driven investment funds running leveraged duration exposure. A supply shock coincided with a demand shock. The Bank of England intervened on the demand side, not on the fiscal position, and yields normalised within days. The speed of that normalisation is itself the tell: solvency repricing does not unwind in 72 hours. A supply-demand imbalance can be resolved by a buyer of last resort stepping in, whereas a genuine reassessment of fiscal sustainability cannot.

Japan is the reductio. Debt-to-GDP has been above 200% for years. The Bank of Japan has at points held over half of outstanding government bonds; domestic institutions hold most of the rest, structurally incentivised by regulation and duration-matching that keeps them domestically aligned. Supply is absorbed and yields have stayed suppressed for decades, albeit at a cost, as the yen’s long depreciation and eventual inflation pressure in 2022 confirmed. A yen-issuing sovereign faces consequences from sustained deficit financing, but they run through currency weakness and real purchasing power, not through a default event that was never plausible.

A recent Financial Times piece on the “Italianisation” of British politics attributed rising gilt yields to political instability and fiscal deterioration. Every mechanism it identified – foreign investors being flighty versus domestic institutions being sticky, the defined-benefit sector’s shrinkage suppressing gilt demand, quantitative tightening reducing the base of stable holders – is a demand phenomenon. Every solution proposed – diluting capital regulations to encourage bank holdings, retail bond programmes, incentivising pension fund re-entry – is demand management. The analysis points directly at supply and demand throughout while insisting the story is about creditworthiness.

The distinction between eurozone members and monetary sovereigns is necessary here. Italy cannot print euros, and so when the populist coalition government threatened large unfunded spending in 2018, it was a plausible solvency signal for holders of paper denominated in a currency Italy does not issue, and spread widening reflected that risk. For sterling, for the Australian dollar, or for the yen, that specific mechanism does not apply, and treating the two cases as equivalent is where most bond vigilante commentary goes wrong.

Monetary sovereignty is not immunity from inflation, currency weakness, or import compression. Rather, it changes the binding condition from solvency to real-resource and external-balance conditions. Governments carrying significant external debt or heavily reliant on imported goods priced in foreign currency face additional transmission channels that domestic fiscal arithmetic does not account for, and political credibility deficits can accelerate those channels considerably. Turkey and Argentina demonstrate what happens when those conditions are absent: the transmission runs through FX collapse and import compression, not through a bond market verdict on fiscal discipline.

The second claim – that deficits drive inflation – is correct in a narrow sense, but that narrowness is the nuance most miss. Deficit spending injects demand, and if that demand hits an economy running below capacity, it flows into real output rather than prices. Inflation requires competition for scarce resources, and capacity surplus removes the scarcity. Deficit spending is inflationary only to the extent it intersects with capacity that is actually tight.

This is the argument’s centre of gravity. Fiscal policy operates as a vector rather than a scalar. The direction of the spending – what it does to productive capacity in already-pressured sectors – determines the inflationary outcome. The deficit number indicates the magnitude of the injection but says nothing about which part of the economy receives it. A government spending heavily in sectors with abundant slack generates little price pressure regardless of deficit size. The same spending directed at already-tight sectors generates significant pressure even at modest scale. The headline figure cannot distinguish between them.

This means the aggregate deficit figure tells you very little about inflationary impact without knowing composition. What determines the inflationary impact is whether each component of spending competes with a part of the economy that is already under pressure. Defence procurement weighted toward imported capital equipment connects only loosely to domestic labour market tensions – the spending leaks offshore rather than bidding up domestic wages. Hospital and pharmaceutical funding is labour-intensive, but the relevant labour markets face workforce shortages in nursing and allied health that predate the current fiscal position and persist regardless of the spending decision. The inflationary multiplier on these categories is lower than on broad cash transfers flowing immediately into general consumption – the velocity is slower, the labour-market overlap narrower, and more of the spending flows into imported inputs or sector-specific workforces rather than general domestic demand.

The post-COVID episode confirmed the point by demonstrating its limits. Aggregate capacity surplus existed – unemployed workers and depressed demand – but supply chains were simultaneously fractured. Deficit spending into that environment ran into sectoral bottlenecks before anything like full aggregate capacity utilisation was reached. The first-round price spikes were sectoral, not aggregate – specific supply chains, specific commodities, specific labour markets. Second-round effects followed as expectations of sustained accommodation fed into wage-setting and forward pricing, but the primary transmission mechanism was composition, not scale. Standard demand-management tools were poorly targeted as a result.

The vector logic produces a conclusion the orthodox debate struggles to accommodate: a surplus can be more inflationary than a deficit. Spending directed at expanding supply-side capacity in tight sectors is disinflationary even if deficit-funded – it adds productive capacity and reduces cost pressure over time. A surplus achieved by cutting that investment degrades the supply position as capacity goes unbuilt, shortages persist, and cost pressure continues without the spending that would have relieved it. The fiscal position tightens on paper while the inflationary mechanism worsens.

The National Disability Insurance Scheme (NDIS) spending cuts in Australia’s 2026 budget illustrate the problem. The Government is counting $37.8 billion in reduced NDIS spending over four years as fiscal consolidation and implicitly as an anti-inflationary measure. NDIS funding flows to disability support workers and care providers. Cutting it does reduce aggregate household income at the margin – the demand channel does exist – but it is weak, indirect, and poorly targeted relative to the dominant source of current price pressure, which is a combination of existing housing supply constraints and imported energy costs from the Hormuz closure. There is no meaningful pathway from cutting care payments to lower oil prices or increased housing supply. The anti-inflation rationale fails because the spending and the price pressure occupy entirely different sectors of the economy.

Much of the anti-austerity literature circles this argument without quite finding it. Mark Blyth, in Austerity: The History of a Dangerous Idea, demolishes the empirical record – austerity has repeatedly failed, the fallacy of composition means every government cannot simultaneously run surpluses while someone else absorbs demand, and the ideology serves distributional interests rather than macroeconomic ones. But Blyth accepts the orthodox view of bond markets at base, treating government debt as a liability that must be serviced against market expectations. He attacks austerity’s track record without dismantling the theoretical architecture that generates it.

Stephanie Kelton and the Modern Monetary Theory (MMT) tradition come considerably closer on the bond market point. The distinction between currency issuers and currency users is correct and essential: a government issuing its own currency cannot default in that currency, the household budget analogy fails, and the solvency ceiling for a monetary sovereign is largely illusory. MMT identifies the correct limit on spending – inflation rather than solvency – and is clear that what checks fiscal expansion is whether the economy has spare capacity to absorb it without price pressure. But MMT’s capacity condition is aggregate. The insight that inflationary impact depends on where spending intersects with specific tight sectors – that fiscal policy is a vector operating on a differentiated economy rather than a scalar operating on a uniform one – remains underdeveloped in that literature.

Austerity applied as a cyclical tool – cutting in downturns, withdrawing demand when capacity surplus is largest – is almost perfectly misconceived under this analysis. It reduces spending broadly, including in loose sectors where additional spending would be non-inflationary and potentially supply-expanding, while making no contact with the tight sectors actually generating price pressure. The fiscal position improves on paper, but the inflation problem, where it exists, does not.

The legitimate concern about sustained high deficits runs on a longer horizon. Accumulating debt does not generate default risk for a monetary sovereign, but rising interest payments progressively narrow fiscal flexibility – the room to respond when the next shock arrives. That is a political economy argument about preserving future optionality and it has some force. It is not an argument about present danger, and it does not require cutting the spending categories that could expand productive capacity and reduce inflationary pressure from the supply side.

The question that fiscal analysis needs to answer is not whether the deficit is large. It is whether the spending competes with sectors that are already tight, who is absorbing the bond supply and at what price, and what the supply-side effects of each spending category actually turn out to be. Those are harder questions than a deficit ceiling, yet they are also the ones that determine the outcome.

Every field has its conversation-stopper. In economics it is the peer-reviewed paper. In medicine it is the clinical trial. In foreign policy and military affairs it is the combination of institutional affiliation and operational experience – “former ambassador”, “West Point professor”, “combat veteran” – that functions less as evidence than as a verdict already handed down. Challenge is not invited because the authority has spoken, and to challenge it is viewed as challenging the authority, not the argument.

The credential and the relevant expertise are often not the same thing. A former ambassador knows how embassies function and how diplomatic protocol operates, but not necessarily how sovereign risk accumulates, how currency crises propagate, or how states actually behave when external financing collapses. An urban warfare expert knows how armies fight in cities, how commanders make targeting decisions under pressure, and the practical logistics of clearing operations in dense terrain. However, that knowledge does not automatically extend to how casualty ratios should be interpreted across different conflicts, or whether precautionary warnings satisfy the effectiveness standard under international humanitarian law, or what the statistical baseline for civilian deaths in modern urban warfare actually is.

The same confusion runs across most public discourse about other specialised topics. During COVID, epidemiologists whose expertise was in disease transmission and reproduction numbers were regularly asked – and regularly offered – judgements about the economic effects of lockdowns, the psychological consequences of school closures, or the political sustainability of compliance. Their expertise in one domain was converted into authority over adjacent domains where their credentials did not follow them. Economists made the same error in the other direction, treating macroeconomic models as evidence about public health outcomes. Everyone was an expert on everything, provided they had the right institutional address.

Corporate governance has used the same mechanism for decades, in a form that is, if anything, more entrenched. From the 1970s onward, executive pay was redesigned around market signals – share price as the authoritative measure of performance – on the theory that markets aggregate information better than boards can. Once price became proof, the question of whether a CEO was actually creating value was displaced by the question of whether the market thought so, and the title pre-certified the answer before the work was examined. The media reinforced this by asking CEOs about every topic imaginable from geopolitics to civil rights. Challenging that verdict looked like ignorance of how things work. The credential – here the combination of executive position and public arena endorsement – made the argument feel unnecessary.

International relations and economics both have specific versions of this. The fields are large, theoretically contested, and their records of anticipating major discontinuities – the Soviet collapse, the Arab Spring, the 2008 financial contagion – are not ones that invite great confidence. Yet the “IR scholar” or “professional economist” label operates as a reliable conversation-stopper in media ecosystems that have neither the time nor the incentive to distinguish between institutional affiliation and demonstrated analytical accuracy. On X in particular, the bio has become the shorthand. “Fellow at [Institute]” or “Professor of [Subject]” functions as a pre-emptive response to challenge, and the “who are you?” rejoinder – directed not at the argument but at the absence of credentials – is a status assertion substituting for a methodological objection.

The practical consequences are visible in the track record. Consider the Iran collapse thesis – the argument, repeated with confidence by credentialed commentators from 2022 onward, that US sanctions combined with oil export disruption would produce fiscal crisis, currency collapse, and regime instability within months. It was wrong, and wrong in a specific and traceable way. The error was a misunderstanding of how oil exports actually flow through a sanctioned economy. Iranian oil exports generate foreign currency – dollars and yuan – flowing through the central bank. They do not generate rials, yet the Iranian Government collects taxes and spends in rials. A sustained fall in oil export volumes thus reduces the country’s foreign currency receipts, but it does not directly empty the treasury. The collapse thesis assumed those two things were the same. The rial had already depreciated sharply to absorb the foreign currency pressure, the parallel rate had incorporated the sanctions discount years earlier, and the economy had adapted through inflation, import compression, and bilateral trade arrangements rather than collapsing. Sanctions have produced real and cumulative damage, but damage operating through those channels looks nothing like the acute fiscal and political crises the thesis required, and an analyst familiar with how sovereigns and governments actually behave would have seen that the adjustment was already well underway – even if it was politically inconvenient to say.

What historical study, structured analytical wargaming, and decades of assessing and pricing sovereign and financial risk brings to that question is a different mechanism map. This background draws on experience of how states and militaries actually behave under pressure rather than how they are expected to respond under a theoretical model – the gap between the two being where credentialed analysis often fails. Sovereign risk analysis shows repeatedly that sanctioned and cornered states adapt in ways their adversaries’ models don’t anticipate: capital controls, bilateral currency arrangements, tolerance for parallel exchange rates, and the substitution of domestic production for imports that were assumed to be irreplaceable. History and wargaming train a similar intuition on the military side – how forces under sustained pressure improvise logistics, compress doctrine, and sustain operations well beyond their nominal capacity, in patterns that appear to be imminent collapse from the outside but are functioning adaptation from within. Theory, like a plan, tends to fail on contact with how things actually work. Iran was no exception.

The same applies to the Hormuz disruption. Much of the early analysis treated it as a straightforward oil price shock – the standard commodity model of supply down and price up. The more relevant question was what kind of economic event a supply disruption of this duration and geography actually is. Oil is not simply a commodity input to be repriced to create equilibrium for supply and demand models. Rather, it is the feedstock for petrochemical production across Asia, the basis for long-term industrial supply contracts, and an input into logistics chains already under strain. A price shock and a supply shock have different transmission mechanisms and different downstream effects, and conflating them produces incorrect predictions. The credential for making that distinction is not an IR or general economics qualification but familiarity with how physical commodity markets, refinery economics, and industrial supply chains actually function – and the willingness to think beyond what orthodoxy is telling you. It is applying a different mechanism map – not a better one, but the right one for that question.

None of this is an argument against expertise. It is an argument for precision about what specific forms of expertise speak to which specific questions, and for scepticism about the institutional shorthand that substitutes credential for argument. It is also an argument that the lack of imagination sustains over-reliance on orthodox models and the unwillingness to engage with analysis that falls outside recognised credentials. The person who has spent a career in urban warfare doctrine has something valuable to say about how commanders make targeting decisions, but that knowledge does not automatically confer authority over whether a given civilian casualty ratio is historically anomalous, whether the legal standard for effective warning was met, or whether the moral and political conclusions he draws actually follow from the operational facts he observed. Those require different mechanism maps, and the credential does not stretch to cover them simply because the holder is confident that it does.

Underlying all of this is a democratic point that the epistemological argument tends to obscure. Civilian oversight of military force is foundational in liberal democracies – it is not optional – and if meaningful critique of military conduct requires operational experience that is by definition unavailable to civilians, then civilian oversight becomes intellectually paralysed, often at the moment of greatest consequence. The “you weren’t there” response to civilian analysis of military operations is not a methodological observation but a political claim about who is permitted to judge – one that sits uneasily with the basic premises of democratic accountability. Analytical standing comes from the fit between the kind of knowledge deployed and the kind of question being asked, and from whether the predictions that follow turn out to be right.

The dominant critique of the 2026 Australian federal budget is straightforward: the government is spending too much at a time of elevated inflation, making the RBA’s job harder and increasing the probability of further rate rises. That critique has the right concern but the wrong mechanism. Applied to this particular inflation, in these particular conditions, it reaches for a textbook response that doesn’t fit the problem.

Inflation in Australia right now has two layers. The first existed before the Hormuz closure – capacity pressures building through 2025, private demand growing faster than expected, and the labour market tighter than the RBA had modelled. That component is demand-pull inflation: more money chasing a limited supply of goods and services, which rate hikes address by making borrowing more expensive and slowing spending. The RBA has already deployed them: three increases this year, the cash rate at 4.35%, with markets pricing a further 60bp before year-end.

The second layer is the Iran war. The Strait of Hormuz carries roughly one-fifth of global oil supply. Its effective closure has pushed oil to more than $US100 a barrel, with second-round effects spreading into transport, construction inputs, and food. That component is supply-side cost-push: prices rising because the cost of producing and moving goods has increased, not because Australians are spending more freely than the economy can absorb. The cash rate is a single lever; it cannot be calibrated separately for each inflation source. It reduces demand across the board while doing nothing to the cost of a barrel of crude, because that cost is set by a geopolitical closure outside the domestic economy entirely.

The RBA knows this. Its May Statement on Monetary Policy presents the two drivers separately and acknowledges that higher fuel prices will erode household real incomes – compressing demand – independently of what the cash rate does. Treasury’s downside scenario, in which oil reaches $US200 and the conflict persists, has GDP going negative in the September quarter. Treasury models this because it recognises the supply-shock channel, while critics who argue the budget is fuelling inflation are implicitly assuming a demand-shock channel. That is the hinge of the disagreement, and it is rarely stated explicitly in the commentary. The same inflationary pattern is visible in the United States, where inflation rose to 3.8% in April – a three-year high – with energy and transport costs accelerating following the Hormuz disruption, despite no equivalent new fiscal expansion occurring simultaneously.

The standard rebuttal is that rate hikes are not intended to lower oil prices; rather, they are intended to prevent higher oil prices from embedding in domestic wages and expectations, which is how a temporary supply shock becomes a persistent inflation problem. That concern is legitimate, but the income compression that fuel prices generate is already doing significant dampening work on its own: households absorbing a 40% increase in fuel costs spend less on everything else, slowing consumption without the RBA needing to add further pressure. Treasury forecasts a fall in real living standards through 2025-26 precisely because this channel is already operating. The RBA’s own analysis shows the primary effect of the conflict on Australian GDP is through household consumption being squeezed by higher fuel costs – demand is being compressed, just not through the interest rate mechanism.

A prolonged supply shock produces a growth contraction that monetary tightening cannot offset and may worsen. That is a stagflation path – the combination of rising prices and slowing growth that is particularly difficult to address because the standard instruments contradict each other. Inflation calls for rate rises to suppress demand and a contracting economy calls for cuts to support it. Raising rates aggressively into a supply shock accelerates the demand destruction without reducing the cost pressure generating the inflation.

Against that backdrop, the argument that the “budget’s $18 billion in additional spending in 2026-27 makes inflation meaningfully worse” requires a further step: it requires showing that the spending competes for the same scarce resources that are actually generating the price pressure. On an annual basis, $18 billion represents roughly 0.7% of GDP, a modest injection into an economy where the dominant price pressure is imported energy, not domestic demand.

The spending breaks down into increased hospital funding under the state agreement, defence procurement, the Pharmaceutical Benefits Scheme (PBS), and Services Australia (welfare and social support payments). Most of it does not meaningfully interact with the binding constraint generating the current inflation. The inflationary multiplier on these categories – the degree to which a dollar of spending generates additional spending and thereby adds to price pressure – is lower than on broad household transfers, albeit not negligible.

Government spending of any kind sustains aggregate income and supports wage persistence, meaning wages that have risen tend to stay risen rather than falling back, which feeds ongoing services inflation even after the original trigger has passed. This effect is not trivial in an economy where services inflation was already sticky before Hormuz closed – the RBA’s own analysis shows capacity pressures were building through 2025. The question is how much the multiplier effect is amplified or moderated by the composition of what is being spent. Defence procurement is heavily weighted toward imported capital equipment sourced offshore, where the connection to domestic labour market pressures feeding services inflation is considerably less direct. Hospital and PBS funding is labour-intensive, and those labour markets are themselves tight, but that tightness reflects sector-specific workforce shortages in nursing and allied health that existed well before the Iran war and will persist regardless of whether this particular budget passes.

RBA Governor Bullock’s formulation that “the extent to which government make up the shortfalls for households by giving them more money, it makes it harder to dampen demand” applies most directly to broad cash transfers that flow immediately into consumption. It applies with considerably less force to infrastructure capacity and public service delivery, where the inflationary impulse is slower and more diffuse.

The NDIS savings are the most politically charged element of the fiscal equation and also the most analytically questionable as an anti-inflation measure. The government is counting on $37.8 billion in reduced NDIS spending over four years – and multiple commentators, including former Treasury deputy secretary Luke Yeaman, have flagged these targets as highly optimistic. Setting aside whether the savings will materialise, cutting the NDIS is a poor instrument for the inflation Australia is experiencing, and the reason is transmission. To address inflation, you need to act on the mechanism generating it. The current price pressure comes primarily from a hydrocarbon supply disruption operating through global commodity markets. NDIS funding flows to disability support workers and care providers – reducing it does not reduce the cost of LNG or crude oil. There is no pathway from cutting care payments to lower energy prices, which means the anti-inflation rationale for the cuts is largely illusory.

The legitimate fiscal concern is different, and it runs on a longer time horizon. Australia’s structural deficit is not a temporary consequence of the Iran war or the COVID hangover. Government spending is forecast to remain near 27% of GDP, the highest outside the pandemic since the 1980s. Gross debt reaches $1.249 trillion by 2029-30 and interest payments are the fastest-growing line in the budget, approaching $47 billion annually by the end of the decade. That is an accumulating burden – not because Australia cannot issue the debt, but because rising interest costs progressively narrow political willingness and fiscal flexibility, reducing the room to respond when the next shock arrives. The structural deficit critique has merit, but the inflation critique – as deployed in this budget debate – overstates the fiscal contribution relative to an exogenous supply shock that domestic policy cannot affect.

If the consensus view prevails – that the budget is insufficiently restrictive and the RBA must therefore tighten further – the rate path goes to 4.7% or higher, layering borrowing cost increases onto households already absorbing fuel price rises, declining real wages, and the RBA’s own projected unemployment drift toward just under 5% by mid-2028. The misdiagnosis leads to unnecessary tightening, which deepens the demand destruction. And deeper demand destruction still does not touch the supply constraint. None of it uncorks the Strait of Hormuz.

The Strait of Hormuz is patient zero. Disruption propagates outward from it through the global supply system in a similar way to how an infectious disease moves through a population. It moves along input-output chains and in branching waves, with each disrupted node becoming a new transmission point into the industries that depend on its outputs. Ten weeks in, primary infections are visible in the data. Secondary infections are incubating behind inventory buffers that are drawing down faster than the consensus has forecast. The market, meanwhile, is asking when patient zero recovers, which is a reasonable question, but not the most important one.

The Wrong Collapse addresses the recovery timeline directly: Iran’s economy deteriorates on a long slope rather than collapsing through the mechanisms the collapse predictions invoke, which means the Strait stays closed longer than markets have priced. But the companion problem is what happens to the global supply system during that interval. An epidemic model explains why the duration of patient zero’s infectiousness matters far more than a standard supply disruption analysis would suggest. Epidemics have their own momentum once the branching has begun, and treating patient zero does not stop infections that have already moved past the primary stage.

Why an epidemic model

The standard economic treatment of Hormuz closure is a supply disruption: a quantity reduction and a price adjustment, and eventually a new equilibrium through demand destruction and supply substitution. That model works adequately for shocks that are bounded and immediately apparent – a brief port closure, a pipeline outage in a single region, a weather event – but it fails for a propagation event, where the initial shock enters the supply chain through feedstock-dependent industries and each disrupted industry then passes the disruption on to those that depend on its outputs.

Inventory buffers absorb the primary shock initially, keeping downstream operations apparently normal while stocks draw down, but when the buffers exhaust, those supply chains break. Each stage is viewed by markets as isolated and temporary while cumulative damage builds across the economy. Different feedstocks have different buffer depths, hence different incubation periods. The chains with thin inventories go symptomatic first and will be apparent now. The chains with deeper buffers look unaffected – operations continuing with normal prices – and will keep looking unaffected until their stocks exhaust. When that happens, they go symptomatic at the same time as the chains that depend on them. By the time the aggregate data captures the full disruption, it has already progressed through multiple production cycles. The cracker margin signal is a leading indicator of this: cracker margins tighten before resin spot prices move, resin moves before earnings revisions, and earnings revise before industrial output data registers the disruption. The lag at each stage runs to three to six months. The cracker margin has been widening since the first days of the closure.

The standard substitution argument depends on three conditions holding at once: (1) alternative supply exists at sufficient volume, (2) it can reach existing processing infrastructure within the relevant time window, and (3) it is chemically compatible with that infrastructure. On volume, the 20:1 ratio of lost to replacement daily output answers the first. On routing, Cape of Good Hope rerouting adds weeks to delivery times and significant freight cost, and cannot restore the throughput or timing that just-in-time supply chains require. The harder limit is chemistry: refineries and crackers are configured for particular crude and intermediate grades, and specific hydrocarbon streams are not globally fungible. A cracker configured for Arab Light naphtha cannot simply switch to West African condensate without the yield falling, the catalyst degrading, and the output failing specification. Downstream polymer, pharmaceutical, and agricultural chemical chains face the same constraint – alternative sources exist, but they are the wrong grade and on the wrong schedule.

The calibration

This is a dialled-back version of two historical comparisons. The point is not that this is of the magnitude of Covid plus the GFC, but rather it is that specific mechanisms from each are continuing, at lower intensity, and in combination.

Covid disrupted approximately 100% of global supply chains roughly simultaneously, making the shock immediately apparent in its full scale even when the duration was uncertain. Closing Hormuz has disrupted approximately 20% of global hydrocarbon flows, sequentially, with staggered incubation periods that allow each wave to be rationalised individually. The sequential nature is what the Covid comparison initially obscures: each wave can be absorbed by the consensus as a temporary and manageable event – which it may be, considered in isolation – while the cumulative damage builds across the economy. Covid was immediately legible because the shock essentially struck all at once, but this situation is less legible because it appears in instalments.

The GFC comparison is narrower – the subprime daisy-chain has no equivalent here, but the load-bearing assumption failure does. Goldman’s scenario set for Hormuz contains three recovery scenarios and no persistence scenarios. Their critical assumption is that the Strait reopens before the inventory floor is reached – similar to the recovery assumptions that rating agency models were using in 2006 and 2007, though the parallel runs only that far and not to what followed. When a single assumption is doing that much work, the question is what happens to everything priced on top of it when the assumption fails.

The primary infections

The primary data is already unambiguous. Goldman’s own production figures show a 14.5mb/d loss against a 0.7mb/d non-Gulf supply response – a 20:1 ratio of lost to replacement output in daily production terms, with the recovery assumption carrying the remaining weight. Bloomberg’s latest monthly production survey has Iraq at -63% and Kuwait at -69% from pre-war levels. Saudi Arabia and the UAE, partially shielded by pipeline bypass capacity, are at -29% and -40%, respectively.

The gasoline crack spread – the refining margin – is tracking 2022’s peak trajectory, well above the upper bound of the five-year range. WTI spot, meanwhile, fell from $102 to $95 in the week ending 8 May. Crude falling while refined product margins hit records demonstrates the simultaneous effects of feedstock volume constraint and demand destruction in different parts of the system – the distinction between a hydrocarbon shock and an oil price shock, as explained in an earlier piece. Price signals are working on demand, but they cannot work on a volume problem.

Three maritime intelligence trackers – Windward, HormuzTracker, and Hormuz Strait Monitor – provide a daily picture of conditions at the Strait. The signal across all three has been consistent: commercial throughput remains effectively at zero against a normal baseline near 60 vessels per day, while war-risk premiums remain 8-15x normal. Where transit numbers have ticked up on individual days, the marginal traffic has been operating covertly, with transponders suppressed. The shadow fleet is expanding, but commercial shipping is still frozen.

The incubation timer

JPMorgan’s inventory analysis – drawing on Kpler, the IEA, the EIA, and JODI – provides the clearest timeline for when the primary infection moves from visible to critical. World oil inventories will hit the operational stress level of 7.6 billion barrels in June under a no-resolution scenario with 5.6mb/d demand reduction. The operational floor – minimum inventory required to keep pipelines functioning and refineries operating – is reached at 6.8 billion barrels in September.

Below that floor the system stops being a “chronic shortage, slower growth” scenario and becomes a physical inability-to-operate scenario. Energy analyst Art Berman described the current configuration last week as markets “adapting to a quieter future of chronic shortage and demand destruction”. That is an accurate description of the path above the floor, but it does not describe what happens below it where chronic shortage is no longer applicable.

Asian refiners are compounding this. They are cutting runs – not for lack of crude, but because committing to WTI cargoes at $110/bbl for July delivery is irrational when Iraqi barrels might be available at $85 if the Strait reopens first. This is game theory – the prisoner’s dilemma. Waiting is each refiner’s dominant choice, and the aggregate of those individually rational decisions produces the collective outcome each was trying to avoid – the bidding war for crude that they were trying to defer. HFI Research described this as a “comedy sketch” – Asian refineries waiting until product storage is gone before buying crude. Joseph Heller would have recognised it.

Secondary infections

The chains that take primary outputs as inputs are moving from incubation toward symptomatic. The sulphur chain illustrates how this works. Sulphur is a byproduct of Gulf oil and gas, and has effectively been trapped in the Gulf as crude since the conflict began. It is a critical input for copper mining through solvent extraction and electrowinning (SX-EW) processes and for nickel production in high-pressure acid leaching (HPAL) plants – about one-fifth of global primary refined copper comes from SX-EW operations. The Democratic Republic of Congo, the world’s second-largest copper producer, relies on the Gulf for most of its sulphur imports, and Indonesia, the world’s largest nickel producer, sources approximately 75% of its sulphur requirements from the Middle East.

In mid-April as acid prices surged, Macquarie Research estimated that Indonesian HPAL production costs had risen by 24% to $US18,000 per tonne, while the LME nickel price had moved to an 11-week high. Natixis calculated that sulphur accounts for 20% of cash production costs for Congo’s SX-EW copper producers, with every $US100 per tonne rise in sulphur prices translating into a 4% rise in operating costs.

These are secondary infections, originating in the Gulf, and arriving in metals markets in Kinshasa, Santiago, and Sulawesi weeks later, through supply chains that most energy-market models do not connect to Hormuz at all. When the buffers in agricultural chemicals, polymer manufacturing, and pharmaceutical intermediates exhaust in the same way, the symptomatic phase begins simultaneously with the early incubation of the chains downstream of them, each wave covering more of the economy than the last.

The ~20% of global hydrocarbon flows understates the eventual footprint as the propagation through secondary and tertiary chains means the eventual economic reach is substantially larger – though how much larger depends on how far the incubation proceeds before the Strait reopens. A short closure keeps the disruption primarily in the energy layer. A closure that runs through the September inventory floor allows the secondary infections to complete their own cycles. At that point the question is no longer about hydrocarbon supply; instead, it is about the downstream industries that have been running on buffers that are now exhausted.

The financial transmission

The GFC parallel is deeper than just the load-bearing assumption point addressed in the calibration section. The physical shock collides with a financial system already carrying multiple independent vulnerabilities, as outlined in the overdetermination argument. Each is capable of extending the damage beyond what the supply disruption alone would produce. They are a tariff shock still working through global trade, a credit cycle turning after fifteen years of near-zero rates, and private credit carrying concentrated exposure to the leveraged corporate sector while earnings face downward revisions. In addition, global markets are looking at three separate liquidity withdrawals: Japan’s yield curve normalisation is pulling capital home as the Bank of Japan raises rates; GCC sovereign wealth funds are reallocating domestically as the war creates competing demands – fiscal, economic, and political – that are pulling Gulf capital homeward; and China’s slow, deliberate withdrawal from US dollar holdings has accelerated in the post-tariff environment. None of these individually constitutes a crisis. Together they represent a withdrawal of marginal demand from US debt at the moment the US fiscal position and corporate leverage are most exposed to rising interest costs.

Private credit is the closest analogue to the GFC’s daisy-chain, though the instrument is different and the gross leverage smaller. Nevertheless, the principle is the same: risk accumulated during easy conditions, held in instruments less liquid and less transparent than regulated bank loans, now confronted by real-economy deterioration that will surface defaults in a market where cross-holdings and concentrations are not visible until something fails. The GFC’s damage was amplified by opacity and the inability to price risk before failures cascaded, and private credit has both properties.

The hyperscaler free cash flow (FCF) data adds a connecting dimension because the AI capex cycle has played the role of primary support for equity sentiment throughout the disruption, hence why the market has been comfortable to price a return to normal despite the physical data suggesting otherwise. As that support erodes, though, the equity market’s acceptance of the recovery scenario will shrink, and the repricing of positions built on that assumption will accelerate. FCF across the major hyperscalers has fallen from roughly $45bn per quarter to approximately $4bn, with Meta burning cash in the second half of last year and Microsoft going negative for at least one quarter. Data centre infrastructure draws from the same Hormuz-exposed supply chains – energy inputs, polymer-based components, industrial materials – and the financial cushion to absorb cost escalation has narrowed.

The Hormuz shock does not need to cause a financial crisis on its own. It only needs to tip a system already under that weight.

No herd immunity

Reopening the Strait does not stop an epidemic that has already branched. Supply chains that have restructured around scarcity – alternative routing, reduced throughput, substitute sourcing at higher cost – do not revert to pre-closure configurations on the day the Strait reopens. Insurance markets, now at 8-15x normal war-risk premiums, reprice on demonstrated safety over time rather than on the announcement of a resolution. Inventory rebuilding from the operational floor takes months. Secondary and tertiary infections already running will continue on their own timelines, at rates that depend on the adaptive responses available to each, but independent of what has happened upstream. By the time the later infections appear in the macro data, the primary infection will have already run through the September floor. That is what the cracker margin signal has been pointing to since the first days of the closure.

The September operational floor is the date by which Goldman’s load-bearing assumption must prove correct. If the Strait is still closed in September, then everything priced on the recovery assumption reprices – and the repricing will be fast because positioning has normalised around an assumption that the physical data has been contradicting for ten weeks.