Arthur Hayes thinks the next Bitcoin tailwind could come from a place most crypto traders are not watching closely enough: an AI spending bust that forces more liquidity into the system.
- Hayes says AI safety talk may also reflect weaker demand for compute
- Debt, private credit, and insurers are the pressure points
- Any rescue that boosts dollar liquidity could favor Bitcoin
- The setup is plausible, but still a speculative macro thesis
In a Sept. 22 essay titled Safety First, Hayes argued that calls from major AI companies to slow frontier model development may not be driven only by safety concerns. They may also reflect a simple financial problem: the AI boom is expensive, and the financing behind it could wobble if demand cools, as discussed in Arthur Hayes Says AI Drained Bitcoin's Liquidity, Is Now.
Hayes put it bluntly:
“Safety First is by definition compute demand destruction.”
That means exactly what it sounds like. If AI labs slow down model training and release schedules, they need less computing power. Less demand for compute can ripple into the broader stack built around it, chips, data centers, power contracts, and the debt used to pay for all of it.
And that stack is huge.
According to Apollo, AI-related issuance accounted for nearly 40% of longer-duration investment-grade corporate bond supply in August. Apollo also estimated that the AI ecosystem could support more than $2 trillion of additional investment-grade debt, while public markets may absorb less than $1 trillion through 2030.
Those are not small numbers. They are the kind of numbers that make everyone look disciplined right up until refinancing gets ugly.
Apollo has separately estimated that roughly $5 trillion could be spent on AI infrastructure through 2030. In a Sept. 21 note, it said consensus forecasts assume operating cash flow at five major hyperscalers, Alphabet, Amazon, Meta, Microsoft and Oracle, will rise from roughly $600 billion to $2 trillion by 2030. That is an enormous jump in cash generation, and it needs to happen if the buildout is going to justify itself.
If it doesn’t, the financing math gets awkward fast.
That is where Hayes starts looking at the darker side of the AI trade. He points to private credit and insurance, where risk is often harder to see than it is in public bond markets. The National Association of Insurance Commissioners said in its July update that private credit has less liquidity, weaker price transparency and less frequent valuation than publicly traded debt. The NAIC also counted 139 private-equity-owned U.S. insurers by June 2025.
Nick Nemeth has estimated that affiliated reinsurance credits across the U.S. life and annuity industry total $1.54 trillion. In plain English: there is a lot of insurance-linked exposure sitting inside structures that can look solid until assets need to be marked lower, meaning recorded at a reduced value because their market price or credit quality has weakened.
That’s the kind of plumbing problem that doesn’t make headlines until it does.
Hayes’s thesis is that if AI demand weakens enough, stress could spread through these financing channels. At some point, Washington may be forced to respond. He floated the idea that the U.S. could become a
“compute buyer of last resort”or step in to support insurers if losses threaten policyholder claims.
That is not a policy decision on the books. It is a stress scenario. Still, it is not a crazy one. When a big enough credit cycle turns, governments and central banks often wind up providing support, directly or indirectly, to stop the damage from spreading.
That is the part Bitcoin holders should care about.
Bitcoin has historically done well when financial conditions loosen, dollar liquidity rises, and risk appetite returns. It does not need a perfect rescue package to benefit. If a shock in AI financing eventually pushes policymakers toward easier conditions, balance-sheet support, or some other form of intervention, BTC could catch a bid from the fallout, a view Hayes has also tied to a Bitcoin could gain from AI bust, Arthur Hayes says.
That said, this is not a clean or guaranteed chain of events. Hayes is sketching a macro path, not calling the next collapse with certainty. AI demand may keep growing. The financing may keep working. Or the industry may simply keep kicking the can with fresh capital, new structures, and a lot of optimism dressed up as spreadsheets. His earlier warning on how a Bitcoin Rally Possible as US-Iran Tensions and Fed Policy could unfold was another reminder that macro shocks rarely arrive neatly packaged for traders.
The current policy backdrop also does not support the idea that the Fed is ready to flood markets. The Federal Reserve raised its target range by 25 basis points on Sept. 16 to 3.75%, 4.00%, its first increase since July 2023. The vote was unanimous, and officials said inflation remained elevated.
The New York Fed also scheduled no reserve-management purchases for both the Aug. 14, Sept. 14 and Sept. 15, Oct. 14 operating periods. Those operations are not the same thing as quantitative easing, but they matter in the broader liquidity picture because they show the Fed is not currently trying to add extra reserves to the system.
Fed H.8 data also showed seasonally adjusted bank credit rising from $19.74 trillion in July to $19.87 trillion by the week ending Sept. 9. So the plumbing is not broken, and the taps are not wide open either.
Meanwhile, the AI spending machine is still running.
OpenAI said in August that it temporarily slowed parts of frontier model development after cybersecurity concerns and because stronger internal safeguards were needed. Anthropic CEO Dario Amodei later called for the industry to pace model development so safety controls could catch up with capabilities. Reuters reported on Sept. 19 that Anthropic is still considering another model release despite Amodei’s public request for slower development.
In other words: safety talk is real, but the race is still on.
That’s also why the financing side of the AI boom has drawn so much attention. In August, Nvidia said Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR were working on independent AI-compute financing platforms intended to mobilize more than $500 billion in third-party capital over time. When that many heavyweight firms are building dedicated funding rails, it usually means the capex bill is enormous, and that traditional balance sheets are already feeling the strain.
SoftBank is doing its part too. According to the Financial Times, it has begun marketing more than $11 billion of high-yield bonds this week to help finance its OpenAI investment. High-yield debt is exactly what it sounds like: riskier borrowing that pays more because lenders want compensation for taking on more risk. That’s finance with the safety rails removed.
None of this proves that an AI bust is around the corner. It does suggest that the sector is leaning hard on future cash flows, debt markets, and a lot of faith that demand for AI services will eventually justify the buildout. If those assumptions start to crack, the stress could travel beyond tech stocks and into credit, insurance, and private lending.
Hayes is basically saying: if that happens, the response could be more liquidity, and Bitcoin tends to like liquidity. His latest macro view follows a similar line to his earlier call that Arthur Hayes Predicts Bitcoin Surge to $100K with $572B in Treasury liquidity, while his warning about Arthur Hayes: Japan’s Yen Crisis Could Spark Bitcoin Rally showed he is clearly hunting for the next big liquidity mismatch, not chasing meme-chart hopium like a degenerate with a candle fetish.
That is not a moon-boy guarantee. BTC does not print green candles on command just because macro got messy. But it is one of the cleaner reasons crypto bulls keep watching debt markets, central banks, and credit stress instead of obsessing over every shiny AI headline.
The ugly truth is that the AI gold rush is not just about models and product demos. It is about leverage, refinancing, opaque credit structures, and what happens when the future revenue gets priced a little too confidently. If that machine stumbles, some of the fallout could be inflationary for liquidity even if it is ugly for the companies involved.
That is the kind of mess Bitcoin has often benefited from.
Key questions and straight answers
-
Why would an AI slowdown matter for Bitcoin?
If weaker AI demand creates stress in debt, private credit, or insurance markets, policymakers may respond with support or easier liquidity conditions. Bitcoin tends to do better when dollar liquidity improves. -
Is Hayes saying an AI crash is guaranteed?
No. He is laying out a possible macro chain: weaker AI demand, financing stress, policy support, and then a liquidity boost that could help BTC. -
What makes the AI financing side look fragile?
Apollo’s estimates point to huge debt needs, and the NAIC has warned that private credit is less liquid and less transparent than public debt. Insurance and reinsurance structures can also hide risk until assets have to be marked lower. -
Are regulators worried about private credit and insurers?
Yes. The NAIC highlighted liquidity, transparency and valuation concerns, and it counted 139 private-equity-owned U.S. insurers by June 2025. -
Is the Fed already easing because of AI risk?
No. The Fed raised rates on Sept. 16 to 3.75%, 4.00%, and the current setup does not look like an emergency liquidity response. -
Are AI companies actually slowing down?
Some are talking about pacing development for safety reasons, but the capital spending cycle is still alive. Reuters reported Anthropic is still considering another release, and major firms are still building AI financing structures.
Hayes may be early, wrong, or simply ahead of a problem that does not show up for a while. But the thesis is worth taking seriously because it connects two things crypto traders should never ignore: leverage and liquidity. If the AI boom turns into a credit headache, Bitcoin is one of the few assets likely to benefit when the money printer, or something close to it, comes back into play.
