House Progressives Throw Down a Marker: No AI Bailouts

Since the dawn of the artificial intelligence disruption, its leading privately held firms have not been required to release financial details in any rigorous way. There was a growing sense that the revenues of the leading AI firms in no way matched the prodigious sums they were spending on infrastructure build-out. But this could be massaged or explained away through a lack of specificity.

But on Monday, Anthropic released a prospectus for its proposed initial public offering, scheduled for sometime next month. And the truth in the numbers could no longer be hidden, even if they could be spun.

Yes, revenues for the maker of Claude jumped 1,300 percent to $16.3 billion in the first six months of this year. But despite that, the company lost $80.7 billion over the same period. Anthropic chalked up much of that loss to an accounting change. But the far bigger numbers to understand are between $518 billion and $807 billion.

Depending on your calculation, that’s the amount of planned obligations Anthropic is on the hook for over the next decade to power and manage its AI models and operations. As much as 80 percent of that is unable to be canceled and is “regardless of actual usage,” and most of it is owed to other Big Tech firms for cloud and compute services and chips: companies like Google ($111 billion), Amazon ($110 billion), Microsoft ($31 billion), Nvidia ($86 billion), and Broadcom ($161 billion). To put this in perspective: For every dollar Anthropic made in 2025, it owes between $113 and $200 to its creditors, most of whom are the biggest tech companies on the planet.

Anthropic doesn’t own its own data centers; it leases the power from other companies. In this sense, Anthropic is “effectively a tenant in a digital landscape owned by the very entities that provide its capital and infrastructure,” as one analysis put it.

Both Anthropic and its partners are now reliant on each other. If Anthropic cannot repay its debts, or Google’s AI ambitions outpace Anthropic’s and it cuts its rival off, or any number of other possibilities in the 94 pages of risk factors listed in the prospectus come to pass, all these counterparties will feel the pain. These kinds of tight connections are precisely what crumbled the housing finance system when the bubble collapsed in 2008.

It’s these kinds of concerns that have led the Congressional Progressive Caucus to write a new House resolution, obtained exclusively by the Prospect, demanding that Congress “refuse to provide any additional funds or exceptional regulatory support to companies or financial institutions facing insolvency or financial collapse due to exposures to acute stress within the artificial intelligence market.”

The text of the seven-page resolution recounts the history of the financial crisis, the lack of prison time for those whose illegal actions caused it, and the congressional bailout delivered to big banks instead of workers or homeowners. The resolution attempts to commit Congress to not make the same mistakes again.

“Many of us, myself included, came into political life watching as many people who caused the crisis in 2008 or cheered it on weren’t held accountable,” said Rep. Greg Casar (D-TX), chair of the Progressive Caucus, who co-authored the resolution. “We know that the American people were appalled at these Wall Street executives nearly two decades ago, and they’re even more pissed off at these tech oligarchs today. We shouldn’t be trying to make them whole.”

His co-author on the measure is Rep. Maxine Waters (D-CA), who is in line to chair the House Financial Services Committee (barring a leadership fight with Illinois Democrat Sean Casten) if Democrats take back the House. “Big tech and Wall Street are taking enormous risks with billions in borrowed money to fuel the AI boom, despite growing warning signs of a dangerous bubble,” Waters said in an email to the Prospect. “We cannot let history repeat itself. The American people should never again be forced to foot the bill for billionaires’ reckless bets.”

The resolution is a nonbinding “sense of the House” measure. But it sets a marker that no bailout funds should be made available to AI companies. Casar wants the measure to be voted on early in the next Congress if Democrats take power, if not sooner, to get everyone on the record about an AI bailout.

The resolution highlights the trillions of dollars in opaque debt instruments that AI companies are liable for, without projected revenues that approach that level of spending. Much of this debt has been kept off corporate balance sheets, even as the tech firms have committed to being responsible for it through guarantees.

Already we’re seeing softening in what has become known as the “AI trade.” Nvidia-backed infrastructure provider CoreWeave is collapsing, with the stock falling close to 50 percent in a month. Its three co-founders have cashed out over $3 billion in stock since going public, fleeing the sinking ship with a payout to themselves.

The stakes surrounding AI’s financial future are incredibly high, critics point out. “Since October 2022, artificial intelligence-related stocks were responsible for as much as 75 percent of the returns of the Standard and Poor’s 500,” according to the resolution, “and the price-to-earnings ratio of the stock market is currently almost as high as it was at the peak of the 1990s tech bubble.” It notes that Nvidia’s market capitalization is nearly $6 trillion, close to 16 percent of total U.S. gross domestic product by itself. If AI is driving the economy—one incredible chart shows data center construction up 64 percent while every other type of construction is down heavily—then a downturn would have serious effects on the entire nation.

That’s why the Progressive Caucus believes that a plan for what to do in a financial crisis is sorely needed. What they’ve laid out in the resolution goes beyond opposing all bailouts, whether through Congress or through special emergency assistance by the Treasury Department or Federal Reserve.

The resolution advises that any bankruptcy of an AI company or affected bank should go through the current bankruptcy processes reserved for those businesses, in ways that support workers, consumers, and pension and retirement fund holders. It states that senior executives liable for excessive risk-taking or any other unlawful action “should be held fully accountable under civil and criminal law,” with disgorgement of ill-gotten gains. It says that the government should take action, but bail out workers instead of executives through fiscal stimulus, job creation, and retirement fund backstops. It suggests automatic stabilizers to prepare for this possibility, so that workers and consumers are protected with immediate benefits if an economic collapse commences. And it proposes a “reform agenda to prevent future speculative bubbles.”

Technology stocks rose cautiously on Tuesday, the day after Anthropic’s prospectus release, in part because all those binding contracts boosted its Big Tech partners. But Casar questioned why the company, which included in its prospectus the destruction of humanity among the risk factors that could tank the stock, was being allowed to cash in at all.

“Before any of these companies should be going public, we need financial transparency across the entire ecosystem,” Casar said. But he characterized the numbers in Anthropic’s release as alarming. The prospectus “should be a wake-up call to all of us that we’re sitting on top of a bubble,” he added.

He saw demands for transparency among the first items in a future oversight agenda if Democrats retake Congress in next week’s election. “The CEOs of the top AI firms should be the first corporate actors hauled in front of our committees,” he said. “They should explain what they thought they were buying when they gave money to Trump’s pet projects. And they should be clear with the American people about making enormous financial bets on AI with the expectation of a bailout. They shouldn’t be counting on taxpayers to save them.”

Read the entire resolution below: