OpenAI’s $122 billion funding round, closed at an $852 billion valuation in the first half of 2026 (Forbes), is the largest private technology raise on record, and it lands alongside reporting that the company has pushed its initial public offering from late 2026 into 2027. The pairing is the real story. Investors are not buying a near-term liquidity event. They are underwriting OpenAI’s ability to keep monetization ahead of compute burn for another year behind closed doors, with public-market scrutiny deferred, not cancelled.
What investors actually bought at the headline valuation
The March 2026 round put $122 billion of committed capital into OpenAI at an $852 billion post-money valuation, the highest price ever paid for a private company, per Forbes’ company profile. It surpasses every prior private tech deal on record.
That valuation trajectory matters more than the headline. The price climbed from a $500 billion valuation in an October 2025 share sale to $852 billion five months later, per Wikipedia, while the company carries an estimated $9 billion net loss for 2025. The round was anchored by Amazon, Nvidia, and SoftBank, per Forbes. Investors paid up anyway, which is the bet: that the revenue curve bends faster than the cost curve, and that a 2027 listing clears a higher number than a 2026 one would have.
Why the IPO slipped to 2027
OpenAI has signaled a listing could come “within the next year,” per media reports. June reporting indicates the actual listing has slipped from a late-2026 window into 2027, with Sam Altman refusing to lower a $1 trillion valuation target, according to a June 2026 analysis.
Holding out for that price while pushing the listing out has a cost that lands on someone else’s balance sheet. SoftBank, whose OpenAI exposure is approaching roughly $65 billion by October 2026, saw its stock drop 13 percent on the delay news, per Chinese-language coverage of the SoftBank hit.
Can revenue outrun the compute burn?
OpenAI’s numbers disagree across sources in a way that matters for anyone pricing the equity. Wikipedia lists $13.1 billion in 2025 revenue against an estimated $9 billion net loss. Forbes reports a roughly $2 billion monthly run rate, about $24 billion annualized, with 40 percent coming from enterprise contracts and 2 million weekly Codex users. A June 2026 analysis puts the 2025 net loss far higher, near $39 billion, on compute-infrastructure spending.
These figures are not necessarily contradictory, but reconciling them requires reading each as a different measurement. That revenue line is most likely GAAP revenue recognized over the calendar year. The monthly run rate is a later point-in-time snapshot, and the June analysis reports 2025 revenue roughly doubled year-over-year. What does not reconcile away is the gap between a low-teens top line and a net loss that even the conservative estimate puts at roughly $9 billion (Wikipedia): every additional revenue dollar currently rides on heavy compute spend underneath it.
The forward spend is where the bet gets tested. The core driver of those losses is compute infrastructure, training successor models, building proprietary data centers, and purchasing hundreds of thousands of GPUs, per the QQ analysis. A 2027 IPO means public investors arrive while that burn is still running ahead of revenue, which is precisely the window where the narrative has to hold without audited proof.
Can enterprise contracts carry the margin story to public markets?
Enterprise contracts are the part of the business that has to carry the margin argument to public markets, and OpenAI is accumulating the case studies now. The company’s MUFG deployment, a vendor-published case study, reports that the Japanese bank rolled out ChatGPT Enterprise to approximately 35,000 employees.
Those numbers are company-reported and should be read as marketing evidence, not audited outcomes. But the structural point holds: with roughly 40 percent of revenue from enterprise and 2 million weekly Codex users, per Forbes, the defensible, sticky revenue is concentrated in large multi-year seats. That is also where pricing pressure will land on buyers. Defending an $852 billion valuation (Forbes), let alone a $1 trillion one (QQ analysis), requires extracting more margin from those contracts rather than simply adding customers.
The Microsoft stake after the restructuring
The Microsoft partnership, once the arrangement keeping OpenAI solvent, now sits inside a restructured corporate form. Microsoft holds a 27 percent stake, per Wikipedia, having previously invested over $13 billion and continuing to provide Azure cloud computing resources. A 2025 restructuring converted OpenAI’s for-profit subsidiary into a public benefit corporation, with the OpenAI Foundation nonprofit retaining 26 percent (Wikipedia).
Anthropic’s IPO could set the benchmark OpenAI has to beat
Anthropic filed confidentially for an IPO in early June 2026 and could list in fall 2026, potentially months ahead of OpenAI, according to the same June reporting. If that timeline holds, Anthropic becomes the first pure-play foundation-model lab to face public-market scrutiny of its unit economics, and its reception sets the comp OpenAI has to clear.
This is the underappreciated risk in pushing to 2027. Private markets are pricing OpenAI at $852 billion (Forbes), but public markets have already shown appetite for repricing AI exposure fast. If Anthropic lists first and the market punishes its loss trajectory or rewards its margins, that single data point reframes what a defensible AI-lab valuation looks like before OpenAI gets to set its own range. The first lab to go public becomes the bellwether whether it wants the role or not.
What will public markets demand from a 2027 OpenAI listing?
By the time OpenAI actually lists, public investors will demand proof of unit economics that private markets have so far waived. The pressure to show margin improvement will land across three surfaces: enterprise pricing, where higher per-seat costs and longer contract terms are the obvious lever; corporate structure, already partially clarified through the 2025 PBC restructuring; and product cadence, with Codex and the ChatGPT Atlas browser expected to show that revenue can compound without a proportional jump in compute spend.
The $852 billion already prices in a great deal of optimism about that curve (Forbes). Holding out for $1 trillion (QQ analysis) prices in another year of having to prove it under private cover, with Anthropic potentially exposing the category to public repricing first.
The bet embedded in this round is that the wait is worth it. Whether SoftBank, Nvidia, and Amazon still agree by 2027 is the part nobody can verify yet.
Frequently Asked Questions
How does OpenAI’s Azure spending differ from ordinary cloud costs?
OpenAI has committed to $250 billion in Azure services and pays Microsoft 20 percent of revenue until AGI. That turns compute from a variable expense into a long-term contractual obligation and a profit-share drag.
What enterprise proof points beyond seat count does OpenAI need?
The MUFG case study reported 1,800 custom GPTs built in four months and 20 to 30 percent workload reductions among 35,000 employees. Public investors will look for that kind of per-customer expansion and measurable efficiency gain, not just more logos.
Why does Anthropic’s IPO timing threaten OpenAI’s valuation?
Anthropic would be the first pure-play foundation-model lab to price its unit economics in public. If the market reprices it sharply, the SpaceX precedent suggests public investors can move faster than private marks, resetting what counts as reasonable for OpenAI before OpenAI sets its own range.
What is the biggest risk in how OpenAI reports its losses?
The $39 billion 2025 net-loss figure may include capitalized data-center and GPU spending rather than pure operating expense. If OpenAI capitalizes infrastructure, current losses look smaller but future depreciation weighs on margins just as public investors arrive.
How does the 2025 restructuring affect what public investors will actually own?
The public-benefit-corporation conversion left the OpenAI Foundation with 26 percent and Microsoft with 27 percent, so public investors would buy a minority economic interest inside a dual-mission structure. The nonprofit’s AGI-benefit mission could constrain returns or capital allocation in ways a standard C-corp would not.
