Arthur Hayes’ Bitcoin call is easy to boil down to one line: $1 million by 2030. His actual argument is less comfortable.
Hayes argues the next major Bitcoin rally could follow a breakdown in debt-financed AI infrastructure. If data centers, GPUs and computing projects fail to generate enough revenue to support their borrowing, governments may respond with new liquidity. In that scenario, Bitcoin benefits not because AI succeeds, but because the financial system eventually needs another rescue.
It’s a bold thesis. Recent market data shows why it can’t be dismissed outright — though the evidence still falls well short of proving an AI credit crisis is imminent.
AI Infrastructure Is Becoming a Credit Story
The AI buildout is no longer being funded only from the cash flows of large technology companies. Debt, private credit, project finance and outside investors are increasingly involved.
Morgan Stanley data reported by Reuters shows that companies raised about $466 billion in AI-linked debt during the first nine months of 2026, compared with roughly $101 billion during the same period a year earlier. Yet issuance slowed sharply in September, when companies raised approximately $23 billion, less than half the August total.
The slowdown does not necessarily mean that AI demand has collapsed. It may reflect the amount of debt already issued, higher borrowing costs and growing questions about whether massive capital expenditure will generate acceptable returns. But it does show that the financing side of the AI boom is becoming as important as the technology itself.
The Numbers Must Eventually Work
A data center can be strategically important and still be a poor investment. The business must generate enough revenue to pay for electricity, land, construction, servers, maintenance, financing and hardware replacement.
That calculation is becoming harder as the industry expands. Research discussed by Axios estimates that large technology companies may need approximately $2.9 trillion to expand computing capacity through 2028, with more than half coming from outside investors. The same analysis estimates that the infrastructure would need to produce about $3.7 trillion in annual revenue within six years to deliver a 10% return.
Those figures do not prove that AI infrastructure is overbuilt. They do explain why investors are paying closer attention to utilization rates, customer contracts and the timing of cash flows. A project can look attractive when demand is rising quickly, but its economics become fragile if customers delay spending or newer chips make existing equipment less competitive.
GPU Values Are Part of the Collateral Risk
Hayes’ thesis depends partly on the idea that lenders may be financing assets whose value falls faster than the loans used to buy them.
GPUs are productive assets, but their resale values are linked to technology cycles. A new generation of hardware can reduce demand for older equipment, even if the older chips continue to function. If a borrower defaults, lenders may not recover the value they assumed when the loan was issued.
The latest resale data offers a reason for caution. American Compute’s index for used data-center GPUs fell to 75 in August 2026, its lowest reading in the 13-month series and 12% below the previous month. The report said older cards led the decline, while H100 systems held up better.
That does not amount to a collapse in GPU values. It does show why GPU-backed lending requires more careful assumptions than ordinary equipment financing. Lenders must estimate how much a chip will be worth several years from now, not simply how much demand exists today.
Private Credit Is Spreading the Exposure
The financial risk is also harder to measure because the debt is not concentrated in one obvious place. AI infrastructure financing can appear through corporate bonds, private credit funds, asset-backed structures, commercial real estate vehicles and project finance.
S&P Global has warned that data-center exposure may look diversified across different products while still depending on the same small group of hyperscalers and AI customers. If the same companies are ultimately responsible for demand across multiple financing structures, a problem at the customer level could affect more lenders than headline diversification suggests.
This is where the 2008 comparison enters the discussion. The point is not that AI infrastructure is identical to subprime mortgages. The similarity is that leverage, optimistic assumptions and complex distribution channels can make risks difficult to see before asset prices begin to fall.
Why This Could Become a Bitcoin Catalyst
If AI investment simply slows, the result could be weaker technology stocks, lower GPU demand and tighter credit. That would probably be negative for Bitcoin in the short term because investors tend to reduce exposure to volatile assets during periods of stress.
Hayes is describing a later stage of the cycle. His view is that losses could become large enough to trigger policy intervention. Governments might increase spending, support financial institutions or expand the supply of fiat currency to contain the damage. If that response produced a broad liquidity wave, Bitcoin could benefit as a scarce digital asset outside the traditional banking system.
The sequence matters. A credit shock would not automatically send Bitcoin higher. The market could first experience forced selling, falling liquidity and a sharp BTC decline. The bullish case depends on what happens after the initial stress: whether policymakers respond with measures that restore liquidity, and whether investors choose Bitcoin as a hedge against currency debasement and financial instability.
Current Bitcoin Flows Are Not Yet a Clear Confirmation
Bitcoin’s institutional demand remains mixed. On October 12, U.S. spot Bitcoin ETFs recorded a net outflow of approximately $65 million. However, the same data showed net inflows of around $1.9 billion over the previous month, suggesting that one negative session does not define the broader trend.
This is important for the $1 million thesis. Before any hypothetical AI-led rescue, Bitcoin would still need durable demand from ETFs, institutions, corporations and long-term holders. Liquidity alone may not be enough if investors are unwilling to allocate capital to BTC.
The market is therefore watching two separate questions. The first is whether AI infrastructure can generate enough earnings to justify its financing. The second is whether Bitcoin can attract sustained capital when macro conditions become more supportive. One does not guarantee the other.
The Bottom Line
Arthur Hayes’ Bitcoin $1 million forecast is not a conventional price prediction. It is a scenario built around debt, technology cycles and monetary policy.
The AI boom is clearly attracting enormous amounts of capital, and the financing structure is becoming more complex. There are already signs of slower debt issuance, higher borrowing costs and greater scrutiny of data-center economics. But there is no confirmed evidence yet of a system-wide AI credit crisis.
For now, the better conclusion is conditional. If AI infrastructure produces strong cash flows, the credit-crisis thesis weakens. If demand disappoints, GPU values fall and private-credit losses spread, the risk becomes more serious. Only a large policy response could turn that stress into the kind of liquidity event Hayes believes might push Bitcoin toward $1 million.
Readers can follow Bitcoin markets, macro developments and digital-asset research through Tapbit. Users who already have an account can log in, while those exploring the platform for the first time can register here.
Frequently Asked Questions
Why does Arthur Hayes think Bitcoin could reach $1 million by 2030?
Hayes argues that a debt-fueled AI infrastructure downturn could eventually lead governments to inject liquidity into the financial system. He believes Bitcoin could benefit from that liquidity and from stronger demand for scarce assets.
Is an AI credit crisis already happening?
Not at a confirmed system-wide level. AI-related debt has increased sharply, while investors are becoming more selective about data centers and financing structures. The current evidence points to rising risk, not a proven crisis.
Why do GPU prices matter for AI financing?
GPUs may serve as collateral for loans used to build AI infrastructure. If newer chips reduce the value of older hardware, lenders could recover less than expected if a borrower defaults.
