Hayes said that if the AI bubble bursts, Bitcoin may benefit

BitMEX co-founder Arthur Hayes published a new article "Situational Relations" on August 4, 2026, arguing that the boom in artificial intelligence infrastructure may end in a credit crisis rather than an Internet bubble-style equity collapse.

Summary

Hayes believes that artificial intelligence infrastructure is similar to leveraged real estate, so future recessions will be a credit crisis. Ultra-large cloud service providers in the United States continue to increase capital expenditures against the backdrop of rapid expansion of cloud computing demand and artificial intelligence workloads. Google parent company Alphabet has raised its 2026 capital expenditure guidance to $195 billion to $205 billion. The Federal Reserve has kept interest rates between 3.5% and 3.75%, and has not announced any artificial intelligence rescue plan. Bitcoin is trading at close to $64337, but Hayes 'article does not provide traders with a proven real-time market catalyst.

Hayes describes data centers as leveraged real estate that contains computing equipment that loses economic value as more efficient chips are introduced. He said an eventual slowdown in data center construction could expose weak borrowers and financiers, prompting governments to intervene and implement broader monetary easing. He believes the resulting liquidity may support a new Bitcoin bull market. However, this is still his personal prediction, not a confirmed crisis or official policy outlook.

Bitcoin traded at approximately $64150 in early trading on August 5. The evidence reviewed in this report does not link this immediate price movement to Hayes 'article. Hayes also admitted that he could not identify the borrowers who could trigger the crisis or the precise bottom of Bitcoin.

Arthur Hayes: AI spending is a real estate credit transaction

Hayes 'core argument is that investors view AI capital spending as supporting high-margin technology businesses with every money they make. He holds a different view on this. Data center land, buildings, power connections and cooling systems are similar to real estate development, and processors lose their value when new equipment provides higher computing power at a lower cost.

This distinction led him to compare it to the global financial crisis. Hayes described the artificial intelligence boom as "a credit story like 2008, not a profit story like 2000." In his vision, banks, insurance companies, private credit funds and infrastructure investors would continue to finance construction projects even as earnings demand began to slow. Losses occur when weaker projects cannot generate enough cash to pay off debt, lease or interest obligations. Even if leading technology companies remain profitable, financial pressures could spread to lenders and investors with exposure to AI infrastructure.

Hayes predicts that announced growth in artificial intelligence capital expenditures will begin to slow in the second half of 2027 and become more pronounced in 2028. He also expects the market to eventually reward companies that reduce construction plans. These dates are forecasts. Company documents reviewed for this report do not confirm that an industry-wide contraction has begun.

His bullish view on Bitcoin stems from expected policy responses. Hayes believes that U.S. authorities will protect strategically important artificial intelligence companies and their lenders because computing power has become part of the U.S. economic competition with China . He discussed a possible trading range for Bitcoin between $60,000 and $70,000, with downside risks close to $50,000, before eventually rising to $1 million. These levels are not guaranteed goals and depend on whether monetary policy, credit creation and investor demand develop as Hayes expects.

This article continues the earlier argument. As previously reported, Hayes has warned that major technology listings, including possible IPOs by OpenAI, Anthropic and SpaceX, could absorb liquidity that might otherwise have entered the crypto market.

Official documents show artificial intelligence spending is still accelerating

The latest company results do not show a collapse in AI capital expenditures. Alphabet reported capital expenditures of $44.9 billion in the second quarter, with approximately 60% of technology infrastructure investment in servers and 40% in data centers and network equipment. Alphabet raised its 2026 capital expenditure guidance to US$195 billion to US$205 billion from the previous US$180 billion to US$190 billion, which the company attributed to faster capacity delivery needed to meet demand. Google cloud revenue increased 82% year-on-year to US$24.8 billion, cloud business operating revenue reached US$8.8 billion, and the backlog of orders increased to US$514 billion. Alphabet said it expects capital expenditures to increase again in 2027.

Microsoft also reported continued expansion. Its quarterly capital expenditure reached $41 billion, with about two-thirds of it being spent on CPUs and GPUs. Microsoft cloud revenue increased 27% to US$59.3 billion, with commercial residual performance obligations reaching US$678 billion. The company expects capital expenditures to grow in fiscal 2027, with expenditures exceeding $50 billion in the next quarter, but some of the amounts reflect changes in certain data center leasing classifications.

Amazon reported a similar combination of investment growth and enhanced cloud revenue. AWS's second-quarter revenue increased 37% to $42.2 billion, the fastest growth in 18 quarters, with operating income reaching $16.6 billion. However, Amazon's free cash flow over the past twelve months turned into an outflow of $7.6 billion, largely attributed to the company's $66.1 billion increase in property and equipment purchases related to artificial intelligence investments.

These results have mixed benefits for Hayes 'argument. Strong cloud growth and large customer backlogs weaken the argument that demand has fallen. At the same time, falling free cash flow, increased depreciation and increased contractual obligations suggest that even as revenue expands, construction can put pressure on finances. High spending alone will not trigger a credit crisis, which requires reduced cash generation, refinancing problems, default or impairment of infrastructure assets by multiple companies and lenders.

U.S. financing exposure is growing, but the 2008 scenario has not yet confirmed

Regulatory documents support Hayes 'narrower view that artificial intelligence infrastructure increasingly involves leasing, guarantees, joint ventures and external capital. Alphabet disclosed $85.2 billion in future lease payments (mainly related to data centers) that had not yet begun as of June 30. These leases are planned to begin between 2026 and 2031, with contract periods of up to 26 years. Alphabet also reported $811 billion in procurement commitments and other contractual obligations, mostly related to technology infrastructure, inventory, energy agreements and other long-term contracts. The company holds $98.2 billion in long-term debt and issued more than $51 billion in fixed-rate bonds in the first half of 2026.

Microsoft disclosed $62.9 billion in financial lease liabilities as of March 31, and another $196.6 billion in leases that had not yet begun (mainly for data centers). Meta reported approximately $182.88 billion in unstarted lease obligations and $237.67 billion in non-cancellable contract commitments as of March 31, and signed another $24 billion in infrastructure contracts in April. Private financing is also becoming increasingly visible in U.S. data center projects. Meta and BlackRock announced the construction of a 1gigawatt campus in El Paso, Texas, which Meta said represents more than $10 billion in investment. Meta's previous joint venture with Blue Owl Capital covered an estimated $27 billion data center park in Louisiana, with Blue Owl Fund receiving 80% of the interest and Meta retaining 20%. Some of the external funds are raised through debt privately sold to PIMCO and other bond investors.

Meta agreed to lease the Louisiana facility with a capped residual value guarantee under certain conditions. These arrangements show how data center exposure is distributed among technology companies, infrastructure funds, landlords and debt investors. They do not prove that a 2008-style bankruptcy chain has begun. Alphabet, Microsoft, Amazon and Meta remain profitable companies with significant operating cash flow and growing customer commitments. Documents reviewed did not report widespread defaults on AI infrastructure debt or official government bailout plans. As a result, the 2008 comparison remains a stressful scenario rather than a current diagnosis. Mortgage losses have become systemic risks because weak lending, securitization, leverage and opaque counterparty exposure have spread among major financial institutions. A downturn in artificial intelligence infrastructure may follow different paths, involving unused capacity, declining rental values, obsolescence of equipment, concentration of tenants and long-term power commitments. Whether these risks become systemic risks will depend on utilization rates, refinancing conditions and where losses end up.

Bitcoin's trend depends on policy, liquidity and timing

On July 29, the Federal Reserve kept the target range of the federal funds rate unchanged at 3.5% to 3.75%, with a vote of 9 - 3. The central bank has not announced an artificial intelligence rescue mechanism, emergency loan program or new asset purchase program. The Federal Reserve conducts treasury bond reserve management purchases to maintain sufficient reserves in the banking system. Its July monetary policy report stated that since early January, total treasury bond purchases have been close to US$250 billion, of which approximately US$160 billion are reserve management purchases. The operations were not officially described as quantitative easing or artificial intelligence bailouts, and the Fed said they were aimed at maintaining adequate reserve levels and supporting control of short-term interest rates.

Hayes interprets balance sheet growth and stable policy rates as support for bank credit and future market liquidity. This interpretation remains controversial because reserve management can expand the Fed's assets rather than represent the broad crisis response assumed in its forecasts. Bitcoin could benefit if future downturns lead to interest rate cuts, emergency loans or larger asset purchases. However, the first phase of a credit shock could hurt Bitcoin as investors sell liquid assets, meet margin calls and reduce leverage. As previously reported in a study of changes in the Bitcoin market cycle, Fed policy and global liquidity are now the main drivers of crypto prices, along with the halving cycle.

The next piece of evidence will come from corporate guidance and credit markets, not Hayes 'article. Investors can focus on 2027 spending plans, cloud backlog conversions, data center occupancy, lease commitments, private credit spreads and any defaults related to artificial intelligence infrastructure. The next scheduled meeting of the Federal Reserve will be held on September 15 and 16. Unless corporate demand weakens or financing pressures begin to emerge, Hayes 'argument remains a forward-looking bullish view of Bitcoin built around the credit crisis that has not yet occurred.

FAQs

Has the artificial intelligence bubble burst? The latest document does not show industry-wide contraction. Alphabet raised spending guidance, Microsoft expects capital expenditures to continue to grow, and AWS reports accelerating revenue. Financial stress is visible in free cash flow and contractual obligations, but these conditions do not constitute a credit collapse.

Why does Hayes compare artificial intelligence to 2008 instead of 2000? [TAG 73]Hayes believes that the main vulnerabilities lie in debt, leasing and infrastructure financing, rather than technology companies with little or no revenue. The comparison depends on the spread of credit losses through financial intermediation, but this has not been confirmed.

Will an AI crash automatically push up the price of Bitcoin? No. Bitcoin may fall during the initial liquidation. Subsequent recovery will depend on the size, speed and form of monetary support, as well as continued demand for Bitcoin. Central bank easing does not guarantee any particular price.

What would weaken Hayes 'argument? Continued cloud revenue, strong data center utilization, profitable artificial intelligence services and stable credit performance will weaken its argument. The argument also loses credibility if companies finance construction projects without causing losses to pressure borrowers or centralized lending institutions.