Everyone writing about the OpenAI IPO is asking the same question: can a company losing $1.22 for every dollar it earns justify an $852 billion price tag? That’s the wrong question, or at least the boring one. The interesting question is why OpenAI is racing toward a September listing at all, when the textbook move for a company this hot would be to stay private another year and let the valuation keep compounding. I think the answer has less to do with OpenAI’s own numbers and more to do with what’s happening three floors down from the AI desk, in the Treasury market.
Here’s the thing nobody’s connecting. At the same time OpenAI has been stacking up roughly $600 billion in compute commitments — Azure, Oracle, AWS, CoreWeave, Google Cloud, the Oracle piece alone running $60 billion a year — the U.S. government has been out there this summer paying the richest yields on 30-year debt since 2001. Corporate AI issuance and Treasury supply are now fishing in the same pond, competing for the same pool of long-duration capital, and strategists on the Street have flagged the AI borrowing wave specifically as one of the forces pushing term premiums higher. That’s not a coincidence OpenAI’s bankers can ignore. It’s a closing window.
Private mega-rounds like the $122 billion one OpenAI closed in March get priced in a world where capital is cheap and patient — SoftBank, Amazon, Nvidia writing nine- and ten-figure checks on the assumption that money stays abundant long enough for the bet to compound. That assumption gets shakier every week long rates sit above 5%. Private investors who backed the $852 billion round at roughly 34 times revenue did so believing the next round, or the IPO, would come at an even fatter multiple. If the cost of capital across the entire market keeps climbing, that math doesn’t hold indefinitely, and the smartest move for a company sitting on paper gains this large is to lock in liquidity while public investors are still willing to underwrite growth stories at any multiple — not wait for a macro backdrop that’s visibly turning less generous.
Look at who OpenAI brought in to write the round: Amazon’s $50 billion included $35 billion contingent specifically on hitting an IPO or AGI milestone, whichever comes first. Read that clause again. Amazon structured its own check around the assumption that a public listing was the more likely trigger. That’s not blind faith in the mission — that’s a bank shot on timing, and it tells you the smart money already priced in urgency months before Goldman and Morgan Stanley showed up.
None of this means the fundamentals don’t matter. They do. Revenue run-rate near $25 billion annualized, up from $13 billion at the end of last year, is real growth by any standard. But set against $600 billion in compute obligations and a guided $14 billion loss for the year — some analysts tracking the burn put it closer to $27 billion once the full buildout lands — the gap between OpenAI’s income statement and its balance sheet commitments isn’t the kind of thing that closes on its own. It closes with either dramatically faster revenue growth, or new capital, or both. And new capital, in a world where the 30-year just touched its highest yield in a quarter century, is getting more expensive for everyone, including a company that isn’t the U.S. government.
There’s a second pressure nobody in the OpenAI coverage seems to want to say plainly: Anthropic isn’t behind. It edged OpenAI in global LLM revenue share in the first quarter of this year, 31.4% to 29%, on a materially cheaper revenue multiple. If OpenAI waits and Anthropic keeps closing that gap, the “we’re simply the market” story that justifies the premium multiple gets harder to tell a public shareholder base with a fiduciary duty to ask why they’re paying up for the label rather than the lead.
So the honest read on the September timeline isn’t confidence. It’s arithmetic. Rates are rising, compute obligations are locked in for years, and the competitive gap that used to be a moat is narrowing. Going public now isn’t OpenAI declaring victory. It’s OpenAI locking in a price before the market that’s currently willing to pay it starts asking harder questions — the same market that’s already asking those questions of the U.S. Treasury.
