🔥 Arthur Hayes: The AI Bubble Will Dwarf Subprime — And Send Bi…

📊 The Core Thesis: AI's Bubble Setup

Arthur Hayes has a provocative view on the markets right now: the AI infrastructure buildout is creating a credit bubble that could dwarf the 2008 subprime crisis. And when it pops, the response will be so large that it sends capital flooding into crypto — potentially pushing Bitcoin toward seven figures.

The setup is straightforward but ominous. According to Hayes, hyperscalers and AI labs are financing GPU purchases with five- to six-year depreciation schedules, even though compute power demonstrably improves every two years. That mismatch between economic reality and accounting fiction is the foundation of what he sees as a multi-trillion-dollar misallocation of capital.

"If I amortize an asset at six years but it has a useful life of two, then all of a sudden all those H100s and Blackwells that you underwrote are essentially worthless."

Hayes points to historical precedent: the 19th-century railroad buildouts led to two major financial crises in 1873 and 1893, both rooted in overoptimistic credit extended to infrastructure projects. He believes AI is following the same arc — except this time, the scale is unprecedented.

🛢️ Energy, Iran, and Oil: The Hidden Accelerant

Beyond balance sheets, Hayes is watching the physical commodity markets closely, particularly oil. Despite the recent peace deal between Iran and Israel, he views current WTI prices as a "fake out."

His reasoning:

Restocking demand is coming. Governments and commercial players drew down inventories during the Iran conflict. Once the Strait of Hormuz reopens, those reserves need to be rebuilt — and that takes months of sustained buying.
Political risk hasn't disappeared. Hayes notes that Trump and Netanyahu have a history of striking Iran after signing deals. Countries that were caught unprepared once won't make the same mistake twice.
Oil could hit uncomfortable levels. He expects WTI to climb toward levels that matter — potentially breaching targets within six to twelve months.

Why does this matter for AI? Because AI is fundamentally about converting energy into intelligence. If energy input costs rise meaningfully, the unit economics of hyperscaler capex start to deteriorate — especially when Chinese models like DeepSeek deliver "good enough" performance at a fraction of the cost.

💸 Circular Accounting and the Nvidia Flywheel

Hayes critiques the circular revenue dynamics propping up AI valuations. Google invests in a lab. The lab buys Nvidia chips. Nvidia reports record revenue. Rinse, repeat.

"It's all just circular accounting. The depreciation is pushed off into the future. You're taking a chip that depreciates every two years, but you're giving it a six-year life."

He sees echoes of Michael Burry's subprime thesis: investors are buying into the brand of American tech dominance without scrutinizing whether returns will actually materialize. And when Chinese competitors offer models at 1% of the cost with comparable performance, that brand premium starts to erode fast.

Hayes points to consumer behavior as the ultimate equalizer:

"Do you care what brand of car you drive if you're not in a Lambo or Ferrari? Get me from A to B. Tesla's fine. BYD is fine. I don't give a [expletive] what the car brand is. I want safe, cheap, reliable."

The same logic applies to AI tokens. If a Chinese open-source model delivers the same rental car search result as GPT-5 for free, brand loyalty vanishes.

🗳️ The Political Reckoning: 2028 as a Referendum on AI

Hayes believes the real danger point comes in late 2027 to 2028, when two forces collide:

Credit reality catches up. Loans underwritten in the 2025–2026 capex boom start coming due. Depreciation schedules force mark-to-market on assets that have already been technologically leapfrogged.
AI becomes a political wedge issue. By the 2028 presidential election, Hayes expects anti-AI sentiment to have crystallized among voters who see data centers polluting their towns, raising electricity costs, and threatening jobs — all while they own none of the winning stocks.

"There's going to be a whole class of politicians that are going to tap into this anger. Where's mine? Why is this data center coming into my town making stuff more expensive, polluting my water? I don't own any of this [expletive]. I'm struggling."

He draws parallels to the early 20th-century backlash against the robber barons. The industrial revolutionaries stayed wealthy, but secondary market shareholders got crushed as capital and labor renegotiated the rules.

🪙 Why Crypto Wins When AI Fails

Here's where the macro thesis turns bullish for Bitcoin and Ethereum. Hayes argues that when the AI credit event materializes, the Fed's first instinct will be to print money aggressively — just like in 2008.

But this time, there's a critical difference:

"The Fed can't print Moore's law. I don't care how much money you throw at this thing, you can't change the fact that a chip gets better every two years if you pump 10 trillion into the economy."

Investors will recognize that no amount of liquidity can fix a depreciation schedule or alter the pace of technological progress. At that point, newly printed capital won't flow back into AI — it will rotate into assets that benefit from monetary expansion without needing to meet a cost of capital tied to obsolescence risk.

That's where crypto enters the picture.

"At that point, investors are going to say, 'I do not want to put any new capital, whether it's free or not, into AI because it does not meet its cost of capital.' So therefore, this capital goes straight to crypto."

Hayes sees this as the trade of the decade — comparable to John Paulson's subprime short or Michael Burry's housing bet. If timed correctly, it's the kind of asymmetric opportunity that changes generational wealth trajectories.

📉 Current Positioning: Risk-Off and Watching

So how is Hayes positioned today?

He's sold out of shorter-term altcoin positions in tokens like HYPE, NEAR, and Zcash — coins he was publicly bullish on just months ago. His reasoning: the asymmetry that existed when he entered those trades (buying HYPE in the high 20s to low 30s, exiting in the 70s) has compressed. The next leg higher is "a harder slog," and with summer volatility and macro uncertainty ahead, he's prioritizing capital preservation over appreciation.

Right now, he's sitting in Treasury bills, earning yield and waiting for the next asymmetric setup. He remains perennially long Bitcoin, viewing it as the ultimate safe haven in a world of infinite fiat debasement.

🔀 Market Dispersion: Majors Lagging, Alts Ripping

One unusual feature of this cycle: Bitcoin and Ethereum are underperforming while select altcoins break all-time highs.

Bitcoin is sitting near its 200-week moving average. Ethereum is trading roughly 30% below that same metric. Meanwhile, tokens like Hyperliquid have pushed through prior peaks, and smaller-cap projects in the infrastructure and DeFi categories are seeing sustained momentum.

Hayes attributes this to AI sucking all the oxygen out of the room. Incremental fiat capital isn't flowing into crypto — it's being funneled into AI equity, debt, and capex. For crypto to see explosive upside again, the AI narrative needs to stumble.

"Without the AI implosion, I don't see how Bitcoin gets to $500,000. If Bitcoin went to $500,000, that's a 10x return. But you could find some second- or third-derivative play in the AI supply chain and 20x in six months. Why would you buy Bitcoin?"

That opportunity cost dynamic is why Hayes sees crypto in a holding pattern — at least until the AI trade cracks.

📈 The Ethereum Setup: A Coiled Spring?

Despite his cautious stance on the broader market, Hayes sees Ethereum as one of the cleanest risk/reward setups among large-cap assets.

ETH is trading well below its prior all-time high, while most other major tokens have at least tested or broken through their 2021 peaks. That underperformance, combined with Ethereum's established network effects and Lindy credibility, makes it attractive on a relative value basis.

"If you gave me a dollar of fiat capital and said choose one, I'd choose Ether over Bitcoin purely just on a chart perspective."

He's evaluating whether to deploy capital into ETH, viewing it as a large-cap crypto asset with asymmetric upside potential — especially if market sentiment shifts and ETH begins to catch a bid relative to Bitcoin.

⚙️ Perpetual Swaps: The Retail Magnet

On the infrastructure side, Hayes remains bullish on the perpetual swap as the dominant retail trading instrument. He credits the product's success to its simplicity, 24/7 availability, and high leverage — all powered by the socialized loss mechanism pioneered by BitMEX and refined across the industry.

Traditional exchanges like CME are launching perp products, but Hayes is skeptical they'll capture meaningful retail flow. Why? Because they're constrained by guaranteed settlement clearing models that don't allow the same leverage or risk flexibility.

"The only way this is going to work for you as an exchange is socialized loss. If you continue to cling to this guaranteed settlement model that's been around for 50 to 100 years, you will not be able to offer a compelling product to the average retail trader."

He believes Hyperliquid will eventually flip Binance as the leading offshore perp platform, citing superior UX, decentralized security, and lower operational overhead. Meanwhile, onshore U.S. perp markets will be highly competitive and commoditized, with multiple licensed players fighting for distribution through marketing spend.

🧭 The Big Picture: Timing the Implosion

Hayes doesn't claim to know when the AI bubble will pop. But he's mapping out the conditions that would trigger it:

Depreciation reality hits in 2027–2028 as GPU financing comes due and technological obsolescence becomes undeniable.
Revenue models shift from subsidized subscriptions to usage-based pricing, exposing the true unit economics and causing demand to crater.
Political pressure mounts ahead of the 2028 election, turning AI into a referendum issue and spooking institutional capital.

If all three align, Hayes sees a credit event larger than subprime, followed by aggressive monetary response — and a historic rotation into crypto.

"This is the John Paulson subprime trade. This is the Michael Burry trade of this particular epoch. If you get this right and time it well, you'll never work again."

For now, he's staying liquid, staying patient, and staying long Bitcoin.

— End of Recap —