The time-honored cliché that owing a bank $100 is your problem, but owing a $1 million dollars is the bank’s problem found new life during the global financial crisis.

Individual mortgage borrowers were on the hook for relatively modest amounts, but those sums totaled many hundreds of billions of dollars.

And we know how that ended.

That is what makes a hint from OpenAI finance chief Sarah Friar all the more worrying. During a discussion at The Wall Street Journal’s Tech Live Event on Wednesday in New York City, Friar detailed the need for an “ecosystem” of financing that could ultimately fund the trillions in artificial intelligence infrastructure the tech industry says it needs to function.

She also mused about the idea of a government “backstop,” or “the guarantee that allows the financing to happen” as OpenAI and other companies “build out the full infrastructure today that allows more compute to come into the world.”

It’s probably not the sort of language investors want to hear from the head bean counter at the beating heart of the AI investment boom; OpenAI has made around $1.5 trillion in spending commitments against current annual revenue of around $13 billion.

It’s also not want investors want to hear from OpenAI, which is the world’s most-valuable start up and could come to market with a $1 trillion price tag next year if it so desired. And certainly not when fixed-income traders are marking up debt protection costs for Oracle, the tech giant whose future revenue growth is tethered to OpenAI’s ambitions. Credit default swaps on Oracle’s five-year bonds are trading at the highest levels since 2023, and have risen nearly 50% over the past month alone, based on fixed-income market data obtained by CNBC.

Oracle raised $18 billion in a multipart bond issue in early October that included paper maturing in 2065. Reports suggested investor demand for the sale—the second-largest corporate bond offering of the year—reached $88 billion.

But a lot has changed since then.

Cracks in the private credit debt market are starting to cause ripple effects in other corners of the fixed-income space.

The use of special purpose vehicles and other exotic forms of financing to keep new debt off the balance sheets of certain tech companies is raising eyebrows. Titanic levels of capital spending from the biggest AI hyperscalers has also continued to increase.

Global tech junk bond issuance, meanwhile, is quietly on pace to surpass the record high reached in 2020, while overall debt sales from companies, governments, and agencies is likely to top $6 trillion before the end of the year.

Friar walked back some of her remarks in a Wednesday evening LinkedIn post, claiming her use of the word “backstop…muddied the point.”

“OpenAI is not seeking a government backstop for our infrastructure commitments,” she wrote. “American strength in technology will come from building real industrial capacity which requires the private sector and government playing their part.”

Backstop or no, the taxpayer is already doing its fair share of the heavy lifting.

Tax breaks from the One Big Beautiful Bill Act make AI investments immensely more attractive by allowing companies to deduct the full cost of new assets. Support for the chip sector among both Democrats and Republican members of the House of Representatives remains central to the government’s economic agenda.

Semiconductors are at the heart of American policy on trade, intellectual property protection, and, in many cases, national security. Top tech executives are regular visitors to the White House and often feted as new industrial heroes when they travel abroad—as was the case when Nvidia’s Jensen Huang visited South Korea last week and Apple’s Tim Cook traveled to China last month.

Meanwhile, Tesla CEO Elon Musk—the world’s richest man— wants a $1 trillion pay deal tied in part to wealth he can create with his AI vision.

Friar’s use of the word “backstop” wasn’t a slip of the tongue, it was a candid admission of what everyone already knows: The stock market, the economy, and the U.S. government are all-in on AI.

If OpenAI can’t find the funding for the hundreds of gigawatts of computing power it’s promising—let alone source the energy required to fuel it—the spillover effects will be significant. Nvidia has a stake in OpenAI. So does Microsoft. OpenAI has a stake in AMD. Amazon added nearly $105 billion in stock market value when it unveiled a $38 billion contrac t between OpenAI and its web services division on Monday.

Oracle, Intel, and Microsoft, as well as smaller companies such as CoreWeave, are all tied deeply into the OpenAI spending web.

The trillions in AI capex, meanwhile, are responsible for the lion’s share of this year’s economic growth—and are powering an outsize portion of the stock market’s record highs.

Put simply, whatever AI owes, it’s likely now everyone’s problem if it can’t be paid back.