Our forecasting track record

What changed, in plain words

• Government borrowing: the US 30-year Treasury yield hit 5.33%, last seen in 2007, with Germany, France and Britain also at multi-year highs. — When a government has to promise a higher return to borrow for 30 years, it usually drags up the cost of long-term borrowing for everyone else too.
• Japan: the Nikkei 225 fell 2.54% as Japan's 10-year government bond yield hit a three-decade high near 2.95% on an unfunded tax cut. — Japan spent 30 years as the world's cheapest place to borrow; that is now unwinding, and it is being read as a fiscal problem, not a growth story.
• Oil and Iran: Brent rose above 91 dollars a barrel for a third session after the interim US-Iran deal expired Monday. Iran and Oman are negotiating a shipping arrangement without US involvement. — Dearer oil is the link between the Middle East and your energy bill, and it is why bond investors are worried about inflation again.
• The Strait of Hormuz: UK maritime authorities reported a ship struck by a projectile, with engine-room damage and one crew casualty. Crossings remain in the single digits a day against more than 130 before the war. — The world's main oil route is still barely functioning, and attacks on ships have resumed rather than stopped.
• Company borrowing: the extra interest risky companies pay, our cleanest early-warning gauge, sits at 2.67 points — historically cheap. — The part of the market that normally cracks first in a genuine crisis is not cracking, which is the strongest argument that today is a repricing rather than a rupture.
• The fear gauge: the VIX, Wall Street's measure of expected share-price swings, dipped to 14.2 on Friday, its lowest of 2026, and one strategist called that complacency heading into a historically choppy September. — Unusual calm in one channel while another channel reprices sharply is a divergence worth noting, not proof either channel is right.
• The AI buildout, now inside the bond story: named commentary today put AI-related bond issuance among the reasons long-term borrowing costs are rising. Separately, Nvidia has arranged with six large investment firms to raise upwards of 500 billion dollars that AI companies can borrow against, keeping the spending off their own balance sheets. — The AI building boom we have tracked for weeks is now big enough to show up as a factor in the government bond market, which is a new kind of connection.
• Europe's central bank: investors moved to price the ECB's key rate reaching 2.76% by March 2027 from 2.25% now, with roughly 90% odds attached to a September increase on the data cited. — Europe is being priced for rate rises, not cuts, which is a change from most of the past two years.
• Shares: US futures pointed lower before the open, with Nasdaq-100 futures down 1.1% against Dow futures down just 0.1%. Home Depot beat expectations but left its full-year outlook unchanged, and the stock rose about 1%. — The fall is concentrated in the long-horizon technology names most sensitive to interest rates, rather than spread across the whole market.
• Trade policy: the three Section 338 proclamations imposing an extra 50% tariff on certain Canadian dairy, vehicle and alcoholic-drink imports take effect at 12:01am Eastern tomorrow, covering close to 20 billion dollars of goods, about 5.2% of what the US imported from Canada in 2025. — A tariff is a tax on imports, usually passed on in prices, and this one starts tomorrow with no deferral announced.
• On the calendar: minutes of the Federal Reserve's 28-29 July meeting are published Wednesday, and Walmart reports results on Thursday. — Both are dated facts, not forecasts — the minutes show what officials actually argued about, and Walmart is the widest read on what Americans are spending.

Headline: a synchronized long-end repricing that credit has not confirmed

The dominant signal today is a simultaneous global long-end government bond selloff. In the US the 30-year reached 5.33% (a 2007 high) and the 10-year 4.75% (highest since January 2025) while the 2-year sat at 4.21% — bear steepening, with 2s10s at +0.53pp. The 10-year TIPS real yield is up 11.6bp on the month to 2.45%, so this is not purely an inflation-expectations move; the real cost of long money is rising too.

Critically, the move is not American. Germany's 10-year Bund cleared 3.25% (highest since March 2011) and its 30-year hit 3.78%, a 15-year high; France's 10-year OAT reached 4.10% (highest since June 2009) and its 30-year an 18-year high into 2027 budget negotiations and a presidential election; the UK 30-year gilt traded at 5.85%; and Japan's 10-year JGB hit a three-decade high near 2.95% on an explicitly fiscal catalyst — a food consumption-tax cut to 1% for two years with no identified replacement revenue. Synchronization across five sovereign curves with idiosyncratic domestic triggers is the signature of a common factor (term premium and supply), not of five local stories.

The named drivers are threefold: elevated energy prices after the US-Iran memorandum expired Monday, inflation persistence and the hike risk it creates, and issuance volume. On the last point, Quilter Cheviot's Richard Carter cited borrowing scale in the UK, France and Japan and added that significant volumes of AI-related bond issuance have added to supply. That is the first time our AI-buildout financing leg has appeared as a driver INSIDE the rates theme rather than adjacent to it. S&P Global Ratings has Italy refinancing debt equal to 17% of GDP in 2026, against France 12% and Germany and the UK 7% each — the supply calendar is not symmetric.

Against all of that, the corroboration is missing where it matters most. The deterministic FRED gauge reads calm on both components: HY OAS at 2.67pp with 21-observation velocity of −0.04pp (narrowing, not widening) and no curve inversion. The VIX printed a 2026 low of 14.2 on 14 August. And the equity response is duration-selective rather than broad: Nasdaq-100 futures −1.1% against Dow futures −0.1%, which is what a discount-rate shock looks like, not what a solvency shock looks like. Three of four cross-checks say absorbed, one says repricing. We are calling that a watch on the rates channel specifically, and refusing to promote it to a systemic reading the evidence does not support.

On the running Hormuz theme: the escalation continued rather than resolving. Trump said he was not interested in extending the interim deal; UKMTO reported a vessel struck outbound with engine-room damage and a crew casualty; transits remain single-digit against 130+ pre-war. Standing prediction P-0074 (a US-Iran interim agreement reopening the strait on or before 2026-08-20, p=0.55) now looks poised to resolve against us — the deadline is Friday and the counterparty has publicly stepped back. We are not adjusting it; it will be scored as logged.

Readings

Fact-check log

Standing theses — re-scored

• #5 Sovereign-debt / rates stress — STRONGLY CONFIRMED on its own terms, and today's dominant theme. Every confirm signal the thesis names fired at once: rising term premia (US 30y to a 2007 high with the 2y anchored, i.e. bear steepening, 2s10s +0.53pp), synchronized multi-year highs across five sovereign curves with independent domestic catalysts, and an explicit supply argument (S&P Global Ratings: Italy refinancing 17% of GDP in 2026 vs France 12%, Germany and UK 7%). The disconfirm signal — term premium contained, duration rallying cleanly — did not fire. Escalated as P-0090 (p=0.25) so the bear case for our own read is scored by reality rather than asserted. — The thesis that governments' long-term borrowing costs would reprice upward had its clearest confirming day yet.
  For your money This is the thesis with the most direct line to household money — it runs through mortgage rates, loan costs and the value of bonds inside pensions. It moves over quarters, not days, so it is not a reason to make a big move; it is a reason to keep a cash buffer for anything needed soon and to stay spread across different kinds of holdings rather than concentrated in one.

  What to watch Watch whether the synchronization persists across several countries for weeks rather than days, and whether any government's debt auction is reported as poorly covered. A clean reversal in long-term rates would be the signal this thesis is wrong.
• #1 AI-credit fragility — CONFIRMED on structure, DISCONFIRMED on price, and now cross-wired into #5. The structure keeps getting more elaborate: 500bn USD of off-balance-sheet compute financing, Nvidia guaranteeing up to a quarter of deals while lenders hold the rest, and GPUs reclassified from fast-depreciating equipment to long-lived infrastructure — a reclassification, not a fact about the hardware. Moody's warning and Alphabet's negative free cash flow on 44.9bn USD of quarterly capex are the same pressure. But the thesis's own disconfirm test — spreads stay tight — keeps passing: HY OAS 2.67pp and narrowing. The genuinely new development is the link: AI-related issuance is now named as a duration-supply driver in the sovereign bond market, which is the first time this thesis has shown up INSIDE #5 rather than beside it. P-0088 already tests whether AI paper keeps clearing cheaply. — The financing engineering keeps growing more complex while the market keeps funding it cheaply — both halves are true at once.
  For your money Almost anyone with an index fund or workplace pension owns a slice of this without having chosen it, so it matters even to people with no view on AI. Debt moved off one balance sheet still exists somewhere. This isn't a reason to make a big move — it is a reason to understand how much of your savings already sits in a handful of very large companies.

  What to watch Watch for AI-linked debt starting to price wider than comparable paper, or a large AI financing reported as hard to place. Broad corporate spreads staying this tight is the ongoing argument against the thesis.
• #6 Energy & critical-commodity supply — CONFIRMED, and the trigger for today's rates move. Brent above 91 USD on a third straight gain, the June memorandum expired, attacks on shipping resumed with a confirmed casualty, and transits still single-digit against 130+ per day pre-war. This is the supply-driven price path the thesis names, not demand-driven softness. The causal chain into #5 is explicit in the sourced commentary: elevated energy prices raise the odds inflation stays high, which raises the odds central banks tighten, which lifts long yields. — The energy-supply thesis is doing exactly what it predicted, and today it was the thing that moved the bond market.
  For your money Fuel and heating are where this reaches a household budget first, then everything that has to be shipped. Conflict premiums can drain as fast as they build, so this is not a reason to stockpile or to chase energy-linked holdings; keeping near-term money in cash and holdings spread out is the standard defence.

  What to watch Watch whether a named producer reports actual lost supply rather than rerouted cargoes, and whether the Iran-Oman talks produce a workable passage arrangement.
• #7 Equity concentration / breadth — CONFIRMING WEAKLY. The pre-open dispersion is stark: Nasdaq-100 futures −1.1% against Dow futures −0.1%, roughly a tenfold difference, and Japan's fall was led by technology and components (Kioxia −7.6%, Taiyo Yuden −11.5%, Murata −9.6%). That is concentrated leadership taking a concentrated hit from a rates shock. But a single pre-open session is thin evidence and the dollar hinge the thesis names did not move much (DXY 99.64). P-0070 already tests whether breadth genuinely broadens. — The narrow group that led the market up is taking the biggest hit, but one pre-open session proves very little.
• #2 De-dollarization — NEUTRAL, and honestly unhelpful today. The dollar index sat around 99.6 and gold was too choppy intraday to cite a level for, so the cleanest expression of this thesis produced no reading we trust. The one qualitative note worth keeping: gold did not break out on a day of confirmed maritime escalation and expired diplomacy, which is mildly at odds with a pure safe-haven story. — No usable signal today; we would rather say that than manufacture one.
• #9 Labor / employment regime — NO NEW DATA, and structurally blind for the next fortnight. The weekly reviewer flagged on 2026-08-16 that the hand-maintained macro calendar has no jobs release seeded after 2026-08-07, so the August Employment Situation (expected around 2026-09-04) is not on our week-ahead radar. That is a known, human-action gap, not an inference we should paper over. Nothing today bears on it either way. — We have no fresh read on jobs, and we're saying so rather than filling the gap with commentary.

Research, not financial advice. Portfolio results are mock capital.
Project snapshot — changes waiting for review: 10 · tracked forecasts open: 54 (checking now: 0) · track record so far: 0.128245 · practice portfolio updated: 2026-08-17.
Past briefs & dashboards: market-sentinel-lyart.vercel.app/reports