Invest1 distinct publisher3 min readPublished
More than half of roughly 9% business capex growth is AI infrastructure, by his Jackson Hole account, which puts at least 4.5 points of measured spending behind a productivity payoff the data will only confirm years later.
The Investor · Invest desk

Compiled by The InvestorSomething wrong?How this is made
If AI infrastructure absorbed more than half of the roughly 9% business capital expenditure growth of the trailing four quarters [3][4], then data centers, chips and the computing backbone contributed at least 4.5 percentage points, and everything else firms bought grew at less than 4.5% [10]. Half of nine is four and a half. The finding is about composition: what the growth is made of, not how large it is.
The token-sale figure behaves the same way under division. Leading AI labs reportedly topped $100B, a 500% increase on the prior year [5], which places last year's base near $16.7B and this year's increment near $83.3B [11]. Note the word the publisher chose, reportedly, and note that what we have is one outlet's account of the keynote [12], so these digits are claims about what was said rather than an audited series.
For a rate-setter the difficulty is timing. Capex is countable this quarter; the return on it shows up in productivity data that the same account concedes is notoriously hard to measure in real time and harder still to attribute to a single cause, with error in either direction producing a policy mistake [14]. The two branches then run in opposite directions off identical spending. Productivity delivered means faster growth without inflation and room to hold rates lower for longer [8]; bubbles, concentrated market power and displacement outrunning absorption mean labor-driven inflation sitting on top of overvalued AI assets [9]. The same investment series feeds both paths. Markets can price the capex flow because it is observable, but they cannot price whether those dollars become output, and that is the gap between a rate path being priced and a rate path being earned.
The allocation signal is the task force appointed earlier in 2026 on productivity and jobs, chartered to assess how AI is reshaping the labor market and the economy's output potential [7]. Research capacity inside a central bank is finite, and pointing it at potential output is a decision to treat this as a supply-side estimation problem rather than only the asset-price problem the second half of the speech describes.
The eighteen-plus AI references in recent FOMC minutes [6] read as a reaction function taking on a new variable rather than a committee narrating the news, in a body that does not ordinarily spend its summers on neural networks [13]. Minutes describe business investment because business investment happened, and the committee still moves on wages and prices, which is a fair point but not the whole picture. What settles it is placement rather than count, so if AI turns up only where staff review financial conditions and never where participants argue the decision, eighteen is a word frequency. And if the next four-quarter print holds near 9% with AI still over half of it [3][4] while the committee's language reverts to labor-market boilerplate, then the "hinge point in history" line [2] was a speech, and I overread it.
Ranked by verification strength, evidence, and original report placement.
Kevin Warsh, identified as Federal Reserve chair, used his keynote at the Jackson Hole Economic Policy Symposium on August 28 to say that AI progress has outrun even the most optimistic forecasts of prior years.
Warsh described the current moment as a "hinge point in history".
Business capital expenditure growth hit roughly 9% over the trailing four quarters.
More than half of that roughly 9% capex growth went directly into AI infrastructure, meaning data centers, chips and the computing backbone that powers large language models.
AI references have appeared at least 18 times in recent FOMC minutes, with policymakers weighing productivity gains on one side and financial stability risks on the other.
Earlier in 2026 the central bank appointed a dedicated task force focused on productivity and jobs, specifically to assess how AI is reshaping the US labor market and output potential.
Distinct publishers with included, body-backed reporting in this cluster.
cryptobriefing.com
1 article · August 29, 2026
Follow any of these and your For You feed starts watching them — no settings page required.
invest
Fed disunity, not the Fed's path, is the risk in duration before Warsh speaks1 distinct publisher
invest
Two-to-one on a hold: the dollar's three-month low reprices every July forecast1 distinct publisher
invest
Two judges, 42 hours: Nvidia's print and Warsh's first keynote price the same trade1 distinct publisher
invest
Gold near $4,379 is a bet on Fed paralysis, not an inflation scare1 distinct publisher
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
One outlet, no documents
Every quantity that makes this story matter — the 9% capex growth, the more-than-half AI share, the $100B in token sales, the eighteen mentions in the minutes, the existence of a Fed task force — reaches the reader through a single Crypto Briefing write-up that links to no transcript, no minutes and no statistical release. The word "reportedly" is doing the job a citation should.
Real money, unaudited arithmetic
What is supposedly being adopted here is capital: data centers, chips and compute soaking up more than half of a 9% capex expansion, plus a claimed $100B of lab revenue. The direction is consistent with everything else visible in the buildout, and the residual is genuinely striking — non-AI business investment growing under 4.5%. But not one of these figures traces to a filing, a national accounts table or a company disclosure in our coverage, so the shape is credible and the decimals are not.
Framing outruns the receipts
"Hinge point in history" and a 500% jump carry a load the sourcing cannot bear, and the implied arithmetic makes it worse: a $16.7B base a year ago growing by $83.3B is an extraordinary claim to leave unattributed. The piece's own best paragraph is the deflationary one — productivity cannot be measured in real time and cannot be pinned on one cause — which quietly concedes that the urgency in the headline will not be settled for years.
A rate-cut read for a rate-sensitive audience
Crypto Briefing's readers are long assets that move on the Fed, and the piece closes by pointing them at the minutes for rate signals. The benign branch — faster growth without inflation, rates lower for longer — is rendered in a clean single sentence; the bubble-and-displacement branch is longer, vaguer and hedged as something that "occupies space in Warsh's thinking". Nothing here is undisclosed conflict, but the framing runs with the audience rather than against it.
Thin and uncorroborated
We can say with confidence what Crypto Briefing published; almost nothing beyond that. The two most quotable numbers are the least sourced, the scenarios are assertions rather than findings, and basic checkable facts — who chaired the Fed on August 28, what the keynote actually said — have no second account anywhere in our coverage to confirm them.