The AI lab’s last private funding round valued the company at $852 billion, but the company could be valued at up to $1 trillion by the time it goes public. A $1 trillion IPO, which would closely follow SpaceX’s record-breaking listing, would be one of the largest wealth events in Silicon Valley history. It would also test whether public markets are willing to bankroll the staggering cost of the AI race and back the heavily loss-making OpenAI.
The company’s S-1 may not arrive for weeks or even months. But whenever the filing does drop, it will certainly be a treasure trove. Here’s what investors will be looking for.
How bad is the burn?
Who stands to gain from the IPO?
OpenAI has an infamously complex structure. Founded as a non-profit, the company finalized its transition into a for-profit public benefit corporation in October 2025.
A $1 trillion listing would turn those stakes into enormous wealth. Microsoft’s stake alone would be worth roughly $270 billion before dilution, while the nonprofit foundation’s stake would be worth roughly $260 billion. OpenAI president Greg Brockman has already testified that his holdings are worth nearly $30 billion, a figure that could rise to roughly $35 billion at a trillion-dollar valuation.
The filing should also give a much clearer picture of the company’s ownership structure, including how much equity is held by insiders, employees, and investors.
What is Sam Altman’s stake and compensation?
There have been persistent questions hanging over Altman’s stake in OpenAI. OpenAI’s CEO previously said he did not directly own equity in the company, though he recently clarified in court that he has a passive stake through Y Combinator.
If Altman has received equity or a new compensation package under OpenAI’s public benefit corporation structure, the S-1 should spell that out as well. It should also show how much he is paid and any performance awards.
What does it cost to serve OpenAI’s models?
Revenue growth is only half the story for OpenAI; there are also questions around what the company spends to generate that revenue and what its so-called “unit economics” are.
In other words: for every token the company serves to a user, how much is it costing it to produce that token? If the S-1 gives any meaningful window into model-serving costs, it could help shed light on whether these products can become truly profitable at scale.
That question is especially important because generative AI does not have the same economics as traditional software. While SaaS products often have relatively low marginal costs per additional user, AI systems incur compute costs every time they generate a response. As usage of tools like ChatGPT, Codex, and APIs grows, providers must scale GPU infrastructure, although efficiency gains and model optimization can reduce the cost per request over time.
How much is spent on compute vs. talent?
OpenAI’s S-1 should also show how its spending breaks down. How much is going toward compute, cloud contracts, chips, and data centers? How much is going toward salaries for some of the most expensive technical talent in the world?
OpenAI has used equity to recruit and retain some of the most sought-after technical talent in the world. The filing should also show how much stock-based compensation it is issuing, how much dilution existing shareholders face, and how much employee wealth is already baked into the company.
What is the revenue mix?
How much is ChatGPT subscriptions? How much is enterprise? How much is API usage? How much is coding products like Codex?
Different revenue streams will give a clearer picture of OpenAI’s customer base. Lots of enterprise revenue may make OpenAI look more durable, while heavy reliance on consumer subscriptions or usage-based API spending could raise questions about churn and margins.
What risks does OpenAI identify?
The risk section could be unusually revealing. OpenAI will likely have to address standard IPO questions such as competition, customer concentration, dependence on partners, and the need for huge amounts of capital.
But OpenAI isn’t just facing standard problems. The company’s own executives have repeatedly acknowledged that their technology might just wipe out all of humanity, help people construct bioweapons, and orchestrate massive coordinated cyberattacks.
The company is also facing a wave of lawsuits over the psychological harms of its technology. OpenAI may have to address everything from national security concerns to psychological risks to the possibility that governments impose much stricter rules on frontier AI systems.
