On Friday, October 16, U.S. tech stocks came under further pressure, with CoreWeave, an AI-powered cloud computing company backed by NVIDIA, plunging more than 10% during intraday trading. The stock had already faced heavy selling pressure over the previous two trading days, crashing 15.5% on October 14.
Selling quickly spread to the chips, cloud computing, and data center sectors. The Nasdaq index faced one of its worst weekly declines in months, as the global AI investment boom encounters a crisis of confidence.
But this decline is different from previous ones.
In the past, investors worried that AI was overvalued; now, they are beginning to worry that AI has too much debt.
Among these, a long-dormant but critical indicator—the credit default swap (CDS)—is now at the heart of this storm.
Deadman's insurance is increasing in price
A CDS is essentially a form of credit default insurance. Bondholders pay premiums regularly to the seller, and if the bond issuer defaults, the CDS seller is responsible for compensating the losses.
These tools are rarely in demand when markets are calm. But when CDS prices begin to rise, it often signals that someone is paying for the “worst-case scenario.”
According to Bloomberg, since autumn 2025, CDS prices for tech giants such as Microsoft, NVIDIA, Meta, and Amazon have surged dramatically, with some companies' prices doubling. The increases for second-tier giants like Oracle and Broadcom have been even more dramatic.
According to the latest statistics from Reuters on October 15, CoreWeave’s five-year CDS surpassed 1,100 basis points, setting a new record. Oracle’s CDS neared 300 basis points, while credit risk indicators for tech giants such as NVIDIA, Meta, Amazon, and Alphabet also rose significantly.
What does 1100 basis points mean? For a simple example, if you were to buy credit protection on a $1 million debt, the annual premium would be as high as $110,000.
The Reuters model shows that CoreWeave’s CDS implies a five-year default probability exceeding 60%, compared to around 30% a year ago. Of course, this is merely model pricing and does not mean the company has a 60% actual chance of going bankrupt.
But a dangerous shift has occurred: the risks of AI investing are spreading from stock valuations to debt repayment capacity.
CoreWeave is precisely the most typical example. Its business model involves borrowing money to purchase NVIDIA GPUs, building data centers, and leasing computing power to customers such as Microsoft and OpenAI. Demand is indeed strong. CoreWeave’s second-quarter revenue reached $2.575 billion, a 112% year-over-year increase, with a contract revenue backlog of approximately $104 billion.
But behind the impressive growth, the financial report hides another set of numbers: a net loss of $626 million for the quarter, with net interest expenses alone reaching $640 million.
The company's revenue has grown rapidly, yet it continues to rely heavily on financing to build data centers. According to LSEG data, its debt-related obligations have approached $50 billion, with yields on some bonds一度 exceeding 16%.
The key concern here is whether the money earned in the future can cover today’s financing costs.
Expected contract revenue is not the same as cash on hand. If data center construction is delayed, hash rate prices fall, or GPU upgrades occur faster than anticipated, a business model that once seemed guaranteed to profit could face significant pressure.
Inflation suddenly heated up, bursting the funding fantasy around AI.
To make matters worse, the macroeconomic environment is moving in the direction AI companies least want to see.
According to data released by the U.S. Bureau of Labor Statistics on October 14, the CPI increased by 3.7% year-over-year in September, up from 3.4% in August, with gasoline prices rising 9.1% month-over-month.
The following day, the PPI rose 1.3% month-over-month, the largest increase since March 2022.
The Middle East conflict has pushed up energy prices, reigniting inflationary pressures and leading markets to bet that the Federal Reserve may further raise interest rates.
What does this mean for the AI industry?
Over the past two years, technology companies have been able to continuously build data centers, largely relying on capital markets willing to fund future growth.
However, when interest rates remain high, debt interest costs consume an increasing share of profits. At the same time, high interest rates also reduce the present value of future cash flows, placing greater selling pressure on highly valued tech stocks.
Thus, the two previously separate risks have combined: inflation has raised financing costs, and those higher costs have caused investors to once again question the returns on AI investments.
Trillion-dollar AI investment, Wall Street finally starts asking for returns
CoreWeave is just one part of the entire AI capital expenditure frenzy.
Microsoft's capital expenditures for the fourth quarter of fiscal year 2026 reached $41 billion, while Meta expects its full-year 2026 capital expenditures to reach $130 billion to $145 billion.
Behind these numbers are substantial investments in GPUs, data centers, power, and network infrastructure.
However, during the same period, Microsoft generated $19.6 billion in free cash flow, fundamentally differing from CoreWeave, which heavily relies on external financing.
This means the AI industry is not collectively facing financial difficulties, but rather the capital market is beginning to distinguish: who is investing profits into AI, and who is betting on AI with borrowed future funds.
Deutsche Bank strategist Jim Reid noted that as AI capital expenditures become increasingly reliant on debt rather than free cash flow, markets must place greater emphasis on changes in credit pricing.
The International Monetary Fund's (IMF) latest Global Financial Stability Report also warns that if AI investment returns fall short of expectations, high leverage and concentrated positions could amplify market volatility.
However, whether the AI bubble has burst remains controversial. NVIDIA CEO Jensen Huang continues to support CoreWeave; while prominent tech analyst Dan Ives expects that 5% to 10% of AI projects may ultimately fail to materialize, as funding requirements have become more stringent.
The AI revolution may be far from over, but Wall Street's patience for endless spending is running out.
Author: Bootly
Twitter: https://twitter.com/BitpushNewsCN
BitPush Telegram community: https://t.me/BitPushCommunity
BitPush TG subscription: https://t.me/bitpush