The Strait of Hormuz has supplied this publication's 1997 parallel since Issue 23 — reserve drawdown, fuel rationing, a current account shock defended until it cannot be. The mechanism is real and the comparison holds. But it is no longer the only currency under stress, and it may not be the more informative one. While attention stayed fixed on the Strait, the yen broke through 160, then 161, then touched 162.20 against the dollar on 30 June — its weakest level since December 1986. It did this despite the Bank of Japan ending negative rates and raising its benchmark to 1%, the highest since 1995, and despite Tokyo spending approximately ¥11.73 trillion — roughly $72.5 billion — defending the currency between late April and late May. The textbook says higher rates and direct intervention support a currency. The yen fell anyway. When the textbook stops working, that is usually the more important data point than the headline it failed to explain.

The conventional account blamed the energy shock — Japan imports the overwhelming majority of its energy in dollars, and a Hormuz-driven oil spike mechanically raises dollar demand. That story had the advantage of being true as far as it went, and the disadvantage of an expiry date. WTI has fallen back near $70, well off the $110 Hormuz-closure peak. The yen has not recovered. A currency that keeps falling after its stated cause has faded is not being driven by that cause. It is being driven by something the stated cause was obscuring — and the candidate explanation is the Eurodollar system, the offshore market in dollar-denominated bank credit that finances global trade and answers to no central bank. When that system tightens, the symptom is a scramble for dollars wherever they can be found, expressed first in the funding markets least able to defend themselves. A BOJ rate hike addresses the interest rate differential. It does nothing for a dollar shortage rooted in the plumbing rather than the spread.

China's bond market is telling the complementary half of the story. While the yen has been falling, Chinese sovereign yields have ground toward multi-year lows — the 10-year near 1.73–1.79% in late June — with the curve bull-steepening as capital crowds into safety and asset managers rotate out of corporate and bank credit ahead of an anticipated downgrade wave. Read in isolation, Japan's weakness looks like an external funding story and China's rally looks like a domestic growth story. Read together, they are the same signal observed from two ends of one pipe: external dollar funding tight enough to break a currency that policy is actively defending, domestic Chinese demand weak enough that capital is fleeing toward duration rather than risk. A genuinely healthy global economy does not produce both at once — and it places gold's recent fall in sharper context. Gold, the yen, and Chinese sovereign debt are pricing the same underlying dollar scarcity from three vantage points that do not normally move together for the same reason. That convergence is more persuasive than any one signal alone.

The UK sits inside the same system it is not yet visibly defending. Index-linked Gilt breakevens are running 3.10–3.40% across the curve, well above the 2.20% implied by US long-dated inflation pricing, while the 10-year Gilt trades near 4.78% and 30-year paper has edged toward 5.50–6.00%. The Bank of England held its base rate at 3.75% in June, with the forward curve pricing a possible insurance hike toward 4% before a gradual easing as growth slows below 1%. None of that is yet a crisis signal. All of it is consistent with a market pricing global dollar scarcity into a sovereign curve that is not — yet — the one experiencing the most acute symptom. The Strait reopened. The yen kept falling. Oil fell with it — and the yen kept falling anyway. That last detail is the one this publication considers most important: if yen weakness were an inflation story, falling oil should have arrested it. It did not. What is being priced, across the yen, the Chinese bond market, and gold's 25% decline, may not be inflation at all. It may be its opposite.

Q2 earnings season opens this week with the major US banks — JPMorgan, Wells Fargo, Citigroup — and runs through July. It is the first reporting cycle in which three simultaneous conditions apply: Hormuz closure costs appear in margin data; AI capex commitments meet actual quarterly cash returns; and forward guidance is issued by management teams who have seen the June NFP print and cannot pretend the consumer is holding. The equity market has answered every piece of uncomfortable data this year with the same verdict: up. Earnings season is the first event in 2026 that the market cannot price in advance of the evidence.

The June Non-Farm Payrolls print — 57,000 jobs against a consensus of 110,000 — missed by a margin the headline conceals. Prior months were revised down by a combined 74,000. The unemployment rate fell to 4.2% because 720,000 people left the labour force, not because hiring improved. Real wages are negative: earnings up 3.5%, inflation at 4.2%. Leisure and hospitality shed 61,000 jobs during a World Cup co-hosted on home soil. This is not a labour market that monetary tightening has successfully cooled. It is a consumer who has run out of capacity. The policy response to inflation and the policy response to demand destruction are opposite in direction. A Fed tightening into demand destruction is compounding the problem it believes it is solving. The Atlanta Fed's GDPNow model printed Q2 2026 growth at 1.2% on 1 July — down from 2.5% on 25 June. That revision, in six days, is the velocity of deterioration the NFP headline does not convey.

The week's most consequential development was not a price move but a business decision. Meta is building a cloud infrastructure business — selling excess compute capacity in direct competition with AWS, Google Cloud, and Azure, following SpaceX's xAI division which has already signed compute-lease deals with Anthropic, Google, and Reflection AI. Both companies built more capacity than their core businesses can absorb. Apple did the opposite: it is paying approximately $1 billion per year to license a custom 1.2 trillion parameter Gemini model from Alphabet — accessing the output of $90 billion in annual AI investment at roughly 1% of its cost — while reducing its own capital expenditure 19% year-on-year. Apple is not losing the AI race. It decided the race was for someone else to run, and is collecting a 30% distribution cut on every AI subscription across two billion iOS devices while its competitors build data centres they now need to rent out.

The counter-movement has a name: intelligence sovereignty. Palantir and Nvidia this week announced a Sovereign AI Operating System for the US government — open-weights Nemotron models, on-premises hardware, absolute ownership of data and model weights. Karp's description of the alternative — feeding proprietary data to closed-source frontier labs — was "effing insane." The arithmetic supports him: running an open-source model in a private control plane costs approximately 16 times less than equivalent closed-model workloads. Life sciences and biotech companies are already declining to share proprietary data with closed providers. Nvidia's motive is structural: as a chip maker, a closed-model duopoly that controls AI purchasing is its worst outcome. Open-source Nemotron is the weapon; Palantir is the delivery mechanism into the customer that needs sovereignty most and is least willing to be told it cannot have it. The rotation away from AI-exposed names has found its explanation. The market is repricing who benefits — and demand destruction, not inflation, is the force that will determine whether any of it matters. The yen knew it in March. The NFP and GDPNow confirmed it in July. The results season may well price it.