Arthur Hayes Is Betting on the Bailout, Not the AI Crash
The BitMEX co-founder says he won't wager on whether the AI data center buildout holds or breaks. He's betting on what happens after it breaks, and he wants Bitcoin and Ethereum to catch the money. Here's what his call gets right, and what it glosses over.
I noticed Arthur Hayes's latest note while scrolling past yet another "AI is a bubble" thread, and it stopped me. Mostly because he isn't making that bet.
Hayes, the BitMEX co-founder who's turned contrarian macro calls into a second career, told traders to go long Bitcoin and Ethereum instead of shorting the AI data center trade. He paired the call with a chart of CoreWeave's falling stock. That's a deliberate choice of visual aid. CoreWeave priced its IPO at $40 in March 2025, ran past $180 by that summer, and has handed back a big slice of it since. Charts like that are how crash narratives get started.
But Hayes isn't calling the crash. He's calling the rescue.
The Mechanics He's Betting On
Here's the part most coverage skips. Hayes explicitly says he won't wager on AI either way. He doesn't care whether the buildout holds or breaks. He cares about what happens to the money afterward.
The thesis runs like this. Data center construction is debt-funded. CoreWeave alone carries billions in borrowings against GPUs that depreciate on a schedule nobody outside the company fully understands. Global data center capex is tracking somewhere near half a trillion dollars for 2025, and a large share of it sits on credit rather than cash flow. If the revenue doesn't arrive fast enough, that debt has to be rolled, restructured, or written down. And because compute capacity now sits next to national security priorities, Washington has a reason to step in.
So Hayes expects a multi-trillion-dollar bailout. Then he wants to own the assets that absorb the printed money.
The question worth asking: does an AI infrastructure rescue actually look like 2008? The template from March 2023, when the Fed backstopped bank deposits within days, suggests yes. Granted, that was depositors, not data center debt. But the precedent is fresh.
Who's Actually Holding The Risk
This matters well beyond crypto. A lot of that AI debt has been sliced up and sold into private credit funds, insurance portfolios, and pension allocations. When people ask where the next credit event comes from, this is one of the honest answers.
If Hayes is right, the knock-on effects touch anyone with a 401(k), not just someone holding ETH on a hardware wallet.
And there's a reflexive wrinkle he doesn't fully address. If enough traders front-run the bailout by buying Bitcoin now, the trade gets crowded before the catalyst ever arrives.
My Take
Color me skeptical, but the direction here feels more sound than the timing.
Hayes has a decent track record on the big macro turns and a much messier one on when they arrive. "The crash is coming, own hard assets" is a thesis that can be right for two years and still lose you money if you size it wrong. Admittedly, that's true of most macro calls.
What I'd watch isn't the equity. It's the credit. CoreWeave's bond spreads, the private credit vehicles with AI infrastructure exposure, and any sign that lenders are getting twitchy about GPU collateral. That's where a crack shows up first, and it'd be visible months before any bailout conversation reaches Congress.
Long BTC and ETH as insurance against a money printer is a defensible position. Just don't confuse it with a trade you can time.
