iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Arthur Hayes argues that the AI infrastructure boom could leave the industry with more data-center and computing capacity than customers will pay to use. If AI labs slow frontier training or make models more efficient, he says, compute demand could disappoint the assumptions behind AI-related borrowing. His next-step forecast is that government support for the sector or exposed financial institutions could add liquidity and benefit Bitcoin—but that is a conditional market prediction, not an announced rescue or a guaranteed price move.
What Hayes means by saying trillions are being “wasted”
At the Gamma Prime Investing Conference in Singapore, Hayes told CNBC, in remarks reproduced in a syndicated report, that humanity is “wasting multi-trillion dollars” building AI data centers. He believes a supply glut could eventually make computing capacity “extremely cheap and extremely plentiful.” Those phrases describe his opinion about the buildout, not an independently measured tally of losses. The available report does not establish that trillions of dollars have already been lost or wasted. The syndicated account of Hayes’s remarks is the source for those quotations.
Why compute demand matters to the debt behind data centers
In his September 21, 2026 essay, “Safety First,” Hayes focuses on whether expected purchases of computing capacity will be sufficient to support the borrowing used to build data centers and acquire chips. He estimates that more than $1 trillion in investment-grade debt, along with hundreds of billions of dollars in lower-credit-quality debt and loans, is backed by leading AI labs’ expected compute demand. This is Hayes’s estimate; the essay does not independently document or calculate the total.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →The concern is not simply that companies might buy fewer chips next quarter. Hayes’s argument is that financing depends on expectations about future demand. If AI labs reduce spending on frontier training or get more output from less computing capacity, their purchases may not meet those expectations. That could weaken the economics of facilities and equipment financed on the assumption of sustained growth.
#1 Best Overall
How a slowdown could lead to cheaper compute—and financial stress
Hayes’s scenario has two linked effects. First, if data centers and chips are built faster than customers’ demand grows, excess capacity could push down the price of compute. Second, lower-than-expected demand could leave the borrowers and lenders tied to that capacity facing weaker assumptions about future revenue and repayment. Cheap compute would benefit users who need capacity, but it would not automatically make the financing of the infrastructure sustainable.
The pivotal question is whether demand for AI computing remains strong enough to support the capacity and debt being built around it. Hayes argues for the demand-shortfall case. The figures and reports cited here do not establish how much AI infrastructure is currently being used, what revenue it generates, or whether future demand will meet financing expectations.
Rank #2
What Hayes thinks governments or markets might do next
Hayes sketches two possible responses if weaker compute demand creates pressure on debt holders or insurers: the government could buy compute as a buyer of last resort, or it could print money to support financial institutions exposed to losses. These are scenarios he proposes, not confirmed government plans or evidence that a rescue is underway.
His Bitcoin thesis follows from that possible intervention. If public support adds liquidity to financial markets, Hayes expects that liquidity could benefit Bitcoin and other crypto assets. The chain of reasoning depends on several uncertain steps: compute demand must disappoint, financial exposure must prompt intervention, and that intervention must translate into buying pressure for crypto. Nothing in the cited material establishes that these events will occur or that Bitcoin will rise if they do.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How to read the forecast
Hayes’s essay is an argument about AI infrastructure financing as well as a forecast for crypto markets. He frames a potential financial response as favorable to Bitcoin holders, so his evident perspective as a crypto investor matters when weighing the prediction. His estimate of debt exposure, his conditional account of how a demand slowdown could affect financing, and his prediction about Bitcoin are distinct claims—not a verified debt tally, an established crisis, or a guaranteed market outcome.
For readers assessing the thesis, the key question is whether durable demand for AI computing will support the infrastructure being financed. Hayes warns that efficiency gains or slower frontier-model training could undermine that demand; the sources cited here do not settle how likely that outcome is.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
