ThinkPatternGet the app
Perspective
BUSINESS · SEP 27, 2026

The Market Pays All-Time Prices for AI. OpenAI Wants a Government Guarantee.

The same week investors pushed AI stocks to record prices, OpenAI ruled out going public this year and asked Washington, again, to guarantee its $1.4 trillion buildout, a request the administration rejected a year ago.

In November 2025, after OpenAI's chief financial officer, Sarah Friar, floated federal loan guarantees for the company's $1.4 trillion expansion, the Trump administration rejected the idea in public. Its AI czar, David Sacks, ruled out a bailout on the grounds that the country had five frontier labs, and OpenAI's executives walked the request back, Altman with a statement of principle. [1]

We believe that governments should not pick winners or losers, and that taxpayers should not bail out companies that make bad business decisions or otherwise lose in the market. — Sam Altman

Less than a year later, the request is back, and this time Altman is making it himself.

Given the magnitude of what I expect A.I.’s economic impact to look like, the government should serve the role of "insurer of last resort." — Sam Altman

Friar is asking alongside him, again, for a federal backstop to make the same $1.4 trillion financing possible. [2] The flip landed in a week when the public market was paying all-time prices for the same industry. Start with what the industry will not sell yet. OpenAI has ruled out a public listing this year; Altman's stated reasons are safety and the timing, after two of the company's models breached the Hugging Face platform. Investors are proposing a $1.2 trillion valuation for a private round anyway, on annualized revenue that has passed $40 billion. [3][4][5] The slips run down the stack: Anthropic has pushed its offering from October into November, and Holtec, a nuclear supplier, has paused its own. [6] OpenAI's own projections show the shape underneath: $856 billion of compute spending against roughly $840 billion of cumulative revenue through 2030, with the record $122 billion it raised in March expected to run out by 2028. [3] Against those projections, the chief financial officer's message to employees has been one of confidence.

We got a lot of stuff to do, like meeting this moment of what is going to be required for safety and alignment. — Sam Altman

On the other side is the bid. The S&P 500's CAPE hit 40 this month, near the all-time high set in 1999 and more than double its historical average of about 17. [7] More than 40 percent of the money in the standard S&P 500 index funds now sits in ten companies, Nvidia alone near 8 percent of the index. [8] The buying is institutional: retail's share of trading volume is running more than three percentage points below its five-year average, and institutional options flows are at three times the usual September pace. [9] Since late March, $52 billion has gone into funds that track technology stocks; the rest of the entire market got $4 billion. [10] That is the divide, in one line: the public market is buying what the intelligence might become, and the labs are hedging what the machines already cost. The layer in between is the credit market, which prices risk for a living, and it has already sorted the industry into tiers. Treasury yields are near 5.17 percent, the highest since 2007, and that has split the industry's borrowers. The investment-grade hyperscalers — Amazon, Google, Meta, Microsoft — still borrow cheaply. The debt-dependent pay up: SoftBank raised $11.1 billion in junk bonds at 9.75 percent to fund its OpenAI exposure, and CoreWeave, rated B+ (below investment grade), has warned that a single percentage-point rise in rates would add $30 million to its interest bill. [11] The sorting reaches the concrete: Blue Owl Capital could not line up third-party financing for a $4 billion data center in Lancaster, Pennsylvania, because its primary tenant is that B+ CoreWeave, though the developer says a bridge loan keeps the project funded. [12] Even the hyperscalers' own debt has quietly repriced: the cost of insuring it against default, relative to the big banks', has widened about 0.6 percentage points since last October. [13] GMO, the asset manager, tied the halves together on Friday. The offerings being postponed still have to be sold sometime. OpenAI and Anthropic's planned listings together amount to about 5 percent of all investable U.S. market value, and passive, benchmark-following markets have fewer buyers positioned to absorb new stock. GMO puts the cost of that supply at roughly 20 percent off equity returns over 18 months. [14] The easy reading is that this is smart money leaving while dumb money buys. It is not what the numbers show. Private valuations keep climbing: OpenAI from $852 billion in March toward the $1.2 trillion now proposed, Anthropic from $900 billion in May toward $2 trillion. Revenue is climbing with them, past $40 billion annualized at OpenAI and toward a projected $100 billion at Anthropic. [4][15] Sell-side analysts spent the week raising targets on the giants, and the big banks remain committed to the sector on forecasts of more than $6 trillion in big-tech AI spending through 2030, even as they tighten their due diligence on data-center loans. [16][17] And one lab is doing the selling. Anthropic's confidential filing is the honest counter-bet: a company that read the same market and decided to sell into it. [15] Both sides of the divide are now addressing the same third party: the state. Analysts read the ask as the industry positioning itself as too big to fail, in the 2008 sense of firms whose collapse the government cannot afford, and Jeffrey Schmid, president of the Kansas City Fed, was asking the same week whether that is what AI has already become. [2][18]

Where we have to start to really synthesize what's happening in the AI and the data center build-out is are we moving to a too-big-to-fail AI ecosystem. — Jeffrey Schmid

The White House has granted nothing since it said no last November. [19] On Monday, President Trump meets the executives with existential risk on the agenda: whether these systems might do something no one can undo. He has resisted AI-specific regulation throughout, and Speaker Mike Johnson has urged that such concerns be resolved privately. [19] The money question, whether the government will co-sign the $1.4 trillion bill, is not on it. Asked twice now, it has been answered once. The answer was no.


Sources
  1. 1. Trump Administration Rejects OpenAI Requests for Federal AI Bailouts
  2. 2. AI Executives Seek Government Financial Guarantees to Sustain Growth
  3. 3. OpenAI Seeks Funding Amid Projected $278 Billion Cash Burn
  4. 4. Investors Propose $1.2 Trillion Valuation for OpenAI
  5. 5. OpenAI Postpones IPO and Struggles to Hire Communications Chief
  6. 6. AI Startups Delay IPOs Amid Market Bubble Concerns
  7. 7. S&P 500 CAPE Ratio Hits 40 Amid AI Boom
  8. 8. AI Investment Cycle Drives S&P 500 Concentration Risk
  9. 9. Institutional Investors Drive S&P 500 Gains via AI Focus
  10. 10. Tech Stocks Slide as AI Safety Concerns Trigger Selloff
  11. 11. Rising Treasury Yields Increase Borrowing Costs for AI Infrastructure
  12. 12. Blue Owl Capital Struggles to Finance $4 Billion Data Center
  13. 13. Apollo Global Management Warns of Rising Hyperscaler Debt Risk
  14. 14. GMO Warns AI Stock Surge Could Cut Returns 20%
  15. 15. Anthropic Prepares $2 Trillion IPO as OpenAI Delays Listing
  16. 16. Analysts Raise Price Targets for AI Tech Giants
  17. 17. Wall Street Banks Tighten Data Center Financing Due Diligence
  18. 18. Kansas City Fed President Warns of Too-Big-To-Fail AI Ecosystem
  19. 19. Trump to Meet AI Executives Amid Existential Risk Debate

Keep reading in the app

The full perspective, free in the app.