Grant Bishop

Tuesday, 10 November 2026 · Markets closed · Edition 20

The record · revised 2026-11-01

Claims

Every claim we have made, the date we made it, what would prove it wrong and when we will check. Nothing is removed; a refuted claim stays on the ledger.

Claims made
7
Resolved
0
Confirmed
0
Refuted
0
Withdrawn
0
As of
ClaimMadeWe check onStatusOutcome
This inflation comes from broken supply, not too much money
in 20 days
Open
The market is paying less for earnings, not doubting the businesses
in 10 days
Open
Spare computing capacity does not stop new power construction
in 5 days
Open
Gulf oil money may be paying for the government-backed AI build
in 110 days
Open
The Fed under Warsh keeps raising rates into this inflation
in 30 days
Open
Spending on the AI build is running ahead of its revenue
in 110 days
Open
AI's economy-wide productivity lift does not arrive before 2028
in 416 days
Open

Each claim in full

This inflation is coming from broken supply, not from too much money. While it lasts, owning the broad market is not the hedge our earlier views assumed.

What would prove us wrong

Core inflation (excluding food and energy) above 3 percent for two quarters after oil falls back below $85 — that would mean the inflation has become demand-driven and the old hedge is back. Or broad-market multiples expanding while oil stays above $100 — the market disagreeing with the 1970s.

layer 1 · review 2026-11-30 · source: BEA core PCE, monthly; EIA Brent

The summer sell-offs on beat-and-raise reports are the market paying less for earnings, not the businesses failing. They reverse when the price of money stops rising.

What would prove us wrong

At the late-October to mid-November prints, companies that beat and raise keep falling while companies that miss hold. That would say the market is pricing the businesses, not the money.

layer 1 · review 2026-11-20 · source: 10-Q filings and earnings releases

Spare rented computing capacity moves work around; it does not cancel new power and cooling construction, which follows capital spending, not chip scarcity.

What would prove us wrong

Hyperscaler capital-spending guidance cut at the late-October reports; or a data center of 100 megawatts or more publicly deferred citing surplus; or Vertiv cutting its growth guidance. (Vertiv stopped reporting quarterly orders and backlog in February; we watch deferred revenue and growth guidance instead.)

layer 1 · review 2026-11-15 · source: hyperscaler 10-Qs; Vertiv's earnings release

Candidate claim: windfall oil revenue in the Gulf is flowing into government-backed AI infrastructure, which is why that demand has separated from the rented-compute market.

What would prove us wrong

A Gulf sovereign fund pausing or cutting an AI commitment while oil stays above $90; or those commitments failing to grow over two quarters despite the windfall.

layer 2 · review 2027-02-28 · source: dated fund and partner announcements

The Warsh Fed tightens into this inflation rather than looking through it, and the pressure on stock prices from the rising price of money continues at least through year-end.

On 16 September the Fed raised its policy rate — the range it targets for overnight lending between banks — by a quarter point, to 3.75–4 percent, on a unanimous vote 25. The Committee's own projections put that rate at 4.1 percent at the end of this year and still 4.1 percent at the end of next: one more quarter-point from here, then a hold. In June the same projections had said 3.8 and 3.6 26. In August we said we would not carry the reported hawkish turn until it was checked against the Fed's own releases. It has been. A chair installed to bring rates down has begun a hiking cycle into a supply-driven inflation, which is the 1970s pattern, and it is the reason the price of money is rising under every stock — the mechanism behind The market is paying less for earnings, not doubting the businesses.

What would prove us wrong

No second increase by the 8–9 December meeting, or a cut, while the statement still calls inflation elevated.

layer 1 · review 2026-12-10 · source: FOMC statements of 27–28 October and 8–9 December; the December projections

Spending on the AI build is running ahead of the revenue it produces.

What would prove us wrong

Cloud revenue keeps pace with contracted backlog through the next two rounds of reports, and no other big buyer starts renting out spare capacity.

layer 1 · review 2027-02-28 · source: cloud companies' 10-Qs, 10-Ks and earnings releases; SpaceX filings on the Google lease

Businesses keep adopting AI, but the economy-wide productivity lift does not arrive before 2028.

What would prove us wrong

Productivity jumps much sooner than we expect, or adoption stops spreading.

layer 2 · review 2027-12-31 · progress note each December · source: BLS Productivity and Costs (quarterly) 27; Census Bureau Business Trends and Outlook Survey 28

Changelog

2026-11-01v3.0.1 — Copy correction: seven section headings reworded in plain language. No claim changed.
2026-10-19v3.0 — Theory restored. The spine with its case against, compute as the meta-resource, the waterfall, the three layers and three orders, what the frame implies, and where it leads — compressed out between v2.0 and v2.2.3 — are back, in the current register. The current regime and its variables are unchanged. Near-term conviction lowered: Spending on the AI build is running ahead of its revenue (the deployment gap) and AI's economy-wide productivity lift does not arrive before 2028 (the lift broadens slowly) added. Watchpoints sorted by layer. The v1.0 and v1.1 rows below are restored to what those versions said; v1.1's dated claims, retired in v2.0 without resolution, are recorded in the June record.
2026-10-13v2.2.3 — Copy correction: variable 5 now says the grid-queue figure measures power projects, not data centers. No claim changed.
2026-10-12v2.2.2 — Source pass: every figure traced to a primary release or removed; Brent restated to the EIA spot series; market moves stated without figures; adjusted basis labeled. Removed as unverified: the dissent-count comparison and the September futures odds. No claim changed.
2026-10-12v2.2.1 — Copy corrections: source 8 closed, Hormuz duration restated; no claim changed.
2026-09-30v2.2 — The Fed under Warsh keeps raising rates into this inflation stated. The hawkish turn flagged unverified in v2.0 is confirmed against the FOMC statement of 16 Sep 2026 (12–0, +0.25 to 3.75–4.00 percent) and the September projections (median 4.1 percent at end-2026 and end-2027, from 3.8 and 3.6 in June). Claim, disconfirmer and 10 Dec review stated. A Fed chair hired to cut is not cutting updated: the hike came in September, not the December the June view priced. Frame unchanged: the revision sharpens This inflation comes from broken supply, not too much money and The market is paying less for earnings, not doubting the businesses, it does not move them. Sources 17–18 added.
2026-08-23v2.1b — Register rewrite under the approved standard: business-pages reader, mechanisms not nouns, one sentence with teeth per section. No figure or source changed. About a third shorter.
2026-08-23v2.1 — Source pass. Brent corrected from "above $100" to $94.39 (21 Aug), May peak and July level dated. Bab-al-Mandeb and Russian-diesel claims removed pending verification. July FOMC hold, 9–3, and September odds sourced; The Fed under Warsh keeps raising rates into this inflation upgraded to a stated claim. Meta corrected from announcement to unconfirmed report. MGX date corrected to 1 July; characterized as a private fund with sovereign anchors. Vertiv traced to the 29 July release; the 47x–36x multiple and TSM/NFLX examples cut for lack of primary source.
2026-08-22v2.0 — Full rewrite. Spine revised: demand-pull vs supply-shock distinguished; the equity hedge made conditional. Regime named. New variables: good news unpaid; compute split into merchant and sovereign legs; oil-funds-sovereign-compute as candidate. Fed pivot flagged unverified. This inflation comes from broken supply, not too much money to The Fed under Warsh keeps raising rates into this inflation opened. Internal draft.
2026-06-10v1.1 — The spine held through an exogenous war shock; the yen's defense and the Warsh Fed read as its texture. Both structural legs deepened: compute's floor moved from announced spending to contracted revenue, and power was named the binding constraint; programmable settlement was upgraded across regulator, institution and sovereign, with the rails maturing into a falling market, stated plainly. Six watchpoints reviewed, six candidates decided, the deployment gap added. Five coverage areas claimed: space, metabolic, robotics, critical minerals, monetary hedge. A reading of the SpaceX listing registered the day before it priced. Archived.
2026-05-12v1.0 — First publication. The spine (inflation as the political way out of debt) with its case against; compute as the meta-resource; the waterfall (productivity, programmable settlement, hyperdiversification); three confidence layers and three orders of effects; six watchpoints; what the frame implies; and a standing position, retired in v2.0 with directional language. Archived.