Part 1 ended with a countdown. 27 April 2026 is the moment the 4-year yen-intervention loop stops being theoretical. This is the mechanism it tests, the data that sits behind each of its three sides, and the destination the capital has already chosen.

The conventional metric for de-dollarisation is the USD share of global reserves, which now sits at approximately 46% — a 26-year low, down 15 percentage points since 2017. That is the summary number, but it is also the most misleading one, because it frames the question as is the dollar losing share? when the analytically important question is where has the capital gone?

Three destinations are visible in the data.

First, gold. Gold has surpassed US Treasuries as the largest component of global central bank reserves for the first time since 1996, constituting approximately 24% of reserves (~$4 trillion) versus 21% for Treasuries (~$3.9 trillion). The People's Bank of China extended its purchasing streak to 16 consecutive months as of February — holdings of approximately 2,309 tonnes, with analyst estimates of true reserves running as high as 5,411 tonnes, Brazil divested $61 billion in Treasuries across 2025 while doubling its gold. Annual central bank gold purchases have consistently exceeded 1,000 tonnes since 2022 — the year the US and allies froze $300 billion of Russian central bank assets, and 95% of central bank reserve managers surveyed by the World Gold Council expect gold reserves to continue rising through 2026, with 0% indicating an intention to reduce holdings.

For the first time in the survey's recorded history, there is no named seller.

Second, Chinese government bonds. Northbound Bond Connect turnover hit 1.22 trillion yuan ($179 billion) in March — an all-time monthly record, with average daily volumes reaching 55.6 billion yuan. China's onshore bond market is valued at over $25 trillion, the second-largest globally; foreign ownership sits below 3% and the PBoC has estimated it could reach 15%. If foreign ownership merely reached the 10% weight China holds in the JPM GBI-EM index, incremental inflows would be measured in trillions, not billions.

Third — and this is where the phrase "Panda bond" stops being a curiosity and starts being a data point — yuan-denominated debt issued by foreign borrowers. Total panda bond financing hit 218 billion yuan in the opening weeks of 2026, already exceeding the full-year 2025 total. Deutsche Bank issued the largest single panda bond ever by a foreign bank (5.5 billion yuan, oversubscribed), Indonesia sold 9.25 billion yuan at approximately one percentage point below its euro-denominated debt, and Morgan Stanley, Barclays, Hungary, and Kazakhstan's sovereign wealth fund all joined the issuer list. Panda bond issuance was up 91.81% year-on-year in 2025. These are not treasury operations, they are sovereign and institutional decisions to denominate borrowing in yuan — which is a different kind of statement than simply buying yuan-denominated assets.

The stability arbitrage is what drove the flows through March. US Treasury yields spiked to 4.4055% during the global bond sell-off while China's 10-year yield moved from 1.80% to 1.84%. A four-basis-point move against a 40-basis-point one is the kind of relative stability that structural allocators pay attention to, and it compounds: onshore CNY bonds returned +3% in USD terms during the 2021–2023 sell-off while US Treasuries fell 12%.

The connection to the Iran war is the tell the consensus model has not integrated. Iran charges the Hormuz toll in yuan, and at least two vessels paid in yuan before the April ceasefire. The same currency absorbing global investment and borrowing flows is the currency collecting rents on the world's most critical energy chokepoint. Capital, rent, and geopolitical leverage are aligning in the same monetary system. The de-dollarisation thesis is no longer about the dollar losing share abstractly — it is about what a rival system looks like when its parts begin to function together.

Japan holds $1.13 trillion in US Treasuries — the largest foreign position in the market. The US–Japan policy rate differential sits at approximately 275 basis points, which is the mechanism by which capital flows out of yen-denominated assets into dollar-denominated ones, and the process that corrects it is yen intervention — which mechanically requires selling dollar assets to buy yen. The Ministry of Finance has spent approximately $150 billion on interventions since 2022, and the 160 USD/JPY level has been the informal red line, breached on 7 April 2026 at 160.46 for the first time since July 2024. Vice Finance Minister Mimura issued language that markets interpreted as a final warning.

The circular problem is that the only way out of the differential — from the Japanese side — is a BoJ hike. Every other lever either requires Treasury sales or accepts currency weakness. Weak Treasury demand pushes US yields higher, which widens the differential, which weakens the yen, which increases intervention pressure, which requires selling more Treasuries, which weakens demand further. The loop has run for four years; the 160 breach is the signal that the buffer against it has compressed.

But a hike is no longer the clean lever it was when the 71% probability was priced.

Japanese industrial production contracted 2.0% month-on-month in March 2026, against +4.3% in February — a 6.3-point swing in a single month. The PPI reading over the same period accelerated to 2.6% year-on-year from 2.1%, with oil and coal product prices up 7.7% month-on-month. These two lines moving in opposite directions at the same time is the textbook stagflation signature: a hike into contracting industrial production is recessionary policy; a hold into accelerating imported inflation is corrosive to household purchasing power and to the yen itself.

That is the domestic trap, and it is the part of the argument the 71% hike probability does not fully price. The probability reads the rate-differential problem, which a hike solves — it does not read the industrial production problem, which a hike worsens. The BoJ cannot optimise for both. It has to choose, and the choice it makes on 27–28 April determines which loop runs next.

Assemble the components.

Hike into recession. A hike narrows the differential from the Japanese side and removes the need for Treasury sales, but it also raises borrowing costs into an industrial base that just printed -2.0%. The hike solves the Treasury problem and creates a Japanese recession. Japanese corporate earnings, a weak automotive export cycle, and semiconductor sector exposure to the Korean supply crisis all compound into a domestic growth shock at precisely the moment the yen needs capital inflows to stabilise. Recession weakens the yen further through expected-growth channels, and the lever aimed at the currency problem ends up firing on it.

Hold into imported inflation. A hold preserves the industrial base in the short run at the cost of structural yen weakness and accelerating consumer inflation. Japanese households hold approximately 54% of financial assets in cash and deposits, which earn nominal yields well below headline inflation, so every month the hold persists is a month of real wealth transfer from household balance sheets to exporters and to the government.

Intervene without hiking. The Ministry of Finance can sell dollar assets to buy yen without BoJ involvement — this was the 2022–2024 playbook. The constraint now is the bid side of the Treasury market: foreign creditor share of Treasuries has fallen from 23% to 6% in thirteen years, gold has overtaken Treasuries in central bank reserves, and recent Treasury auctions have cleared at elevated yields with primary dealers absorbing abnormally high shares — the 2-year auction in late March required primary dealer absorption of nearly 24%, more than double the prior six-month average of 11%. Japanese MoF sales into this environment do not just defend the yen; they add supply to a market already short of buyers, which widens the differential they are trying to close.

The three sides reinforce each other: Intervention requires Treasury sales; Treasury sales widen the differential; a wider differential increases intervention pressure. A hike relieves the differential from the Japanese side, contracting industrial production makes the hike recessionary, a recession weakens the yen, and a weaker yen reopens the intervention imperative. Hold preserves industrial production, imported inflation erodes the yen, a weaker yen forces intervention, and intervention returns to the Treasury-sales problem.

There is no lever that does not feed one of the other two.

The BoJ meeting is priced at a 71% hike probability, but that number aggregates across two separate questions: the first is whether the BoJ has the policy space to hike — the answer, pre-March IP data, was unambiguously yes, and the 71% is largely a residue of that earlier framing, and the second is whether the BoJ is willing to hike into contracting industrial production, which is a different question with a different answer.

Governor Ueda's language over the past three weeks has been carefully non-committal. The difference between the 71% prior and the actual outcome is information the market has not fully repriced, and a hold is not a neutral outcome. A hold confirms that the industrial production constraint binds — and the trap tightens without a release valve, because the intervention path that stays open is the Treasury sales path.

In the event of a hold, the observable sequence over the following weeks is: USD/JPY re-breaches 160 within days as the differential argument reasserts, MoF intervenes by selling Treasuries, Treasury auction tails widen, the 10-year yield pushes toward 4.5%, the Fed's rate-trap (no cuts priced through December, JPMorgan forecasting zero cuts for 2026) becomes harder to escape, the yen weakens through the intervention level, and the loop restarts.

A hike is the only clean exit, and it is the one that makes the Japanese domestic economy the adjustment variable. The BoJ is not being asked whether to hike, but which part of its mandate to sacrifice.

The Bank of Japan, 27–28 April. Last edition named this meeting as the structural inflection point for the yen vortex and everything downstream of it, and 4 days closer, the call holds — and the evidence under it has hardened. Japan's March industrial production print (-2.0% MoM, a 6.3-point swing from February's +4.3%) arrived after Edition 5 published, and it is the data that turns the 71% hike probability into a question about which part of the BoJ's mandate it is willing to sacrifice.

The rate decision is the headline, but Ueda's language is the signal. A hike with dovish framing is a different signal from a hike with hawkish framing; a hold with hawkish forward guidance is a different signal from a hold that acknowledges the industrial production floor, with each combination implying a different loop running next.

Two secondary reads worth watching alongside it: USD/JPY behaviour in the 48 hours after the decision — a re-breach of 160 within days signals the differential argument has reasserted regardless of what the BoJ did, and any comment from the Ministry of Finance on intervention readiness — when BoJ policy is constrained, MoF intervention is the lever that lands directly on the Treasury market.

Thirteen active disruption vectors across three tiers:

The structural signature is three-way concentration, 5 are China-gated (rare earth magnets, tungsten, sulphur via the acid ban, and the naphtha feedstock cascade running through Asian crackers), 6 are Hormuz-gated (fertiliser, helium, LNG, crude oil, neon, and the sulphur source constraint), and the remaining 2 — bromine and MEG — are Israel-theatre-gated through Neot Hovav and Jubail respectively.

No single ceasefire unblocks the index, and no single trade deal unblocks it either. The vectors at the top of Tier 1 (rare earths 90%, tungsten 85%, sulphur 50%) have no substitution pathway at scale; the vectors in Tier 2 have partial substitution pathways that run multi-quarter to multi-year.

Shipping and insurance sit outside the percentage framework because they gate every other commodity's recovery timeline.

The vector that moved most between Editions 5 and 6 is fertiliser — not by rising further, but by having the biological clock it was attached to pass, and the vector that will move most between Edition 6 and Edition 7 is bromine, if any of the Tier 2 vectors tip into structural damage; otherwise the index holds.