OpenAI’s valuation is the number that most often makes headlines, and the one most easily misread. Within a single year, its post-money valuation jumped from around $300 billion to over $800 billion, cementing its place as one of the most highly valued AI companies in the world; in June 2026, the company officially confirmed it had quietly taken the first step toward going public.
Where do these astronomical numbers actually come from, and how should you read them? This piece walks through them with a research framework, where the focus isn’t “whether it’s worth buying” but “what these numbers mean.” If you want to get to know the company as a whole first, you can start with What kind of company is OpenAI.
One caveat up front: OpenAI hasn’t gone public and hasn’t disclosed complete financial statements, so a number of the figures below are the company’s public framing or third-party estimates. I try to make clear which are official and which are estimates.
The funding rounds OpenAI has been through
OpenAI grew from a donation-funded nonprofit lab into one of the most highly valued AI companies in the world. Lay out a few of its key funding rounds and you can see just how fierce that growth has been (amounts and valuations are approximate figures from media and official announcements; the 2026 row is an official-announcement figure):
| Time | Round / Nature | Amount | Post-money valuation | Lead / Main backers |
|---|---|---|---|---|
| 2015 | Nonprofit founding | ~$1B committed | (none) | Musk, Altman, et al. |
| 2019 | Microsoft’s first strategic investment | ~$1B | (undisclosed) | Microsoft |
| 2021 | Employee liquidity transaction | (mostly secondary) | ~$14B | Tiger, a16z, Sequoia |
| 2023 | Microsoft multi-year top-up | ~$10B | ~$29B | Microsoft |
| 2024 | Series E | ~$6.6B | ~$157B | Thrive Capital |
| 2025 | Series F | ~$40B | ~$300B | SoftBank |
| 2026 | Latest round | ~$122B committed capital | ~$852B | Amazon, NVIDIA, SoftBank |
The most eye-catching part is those last two steps: in a single year, the post-money valuation jumped from around $300 billion to over $800 billion, nearly tripling. It’s worth emphasizing that these are all private-market valuations, that is, the price investors are willing to assign during private fundraising, not a price traded on the public stock market. Private valuations often reflect a handful of large investors’ bets on the future, and they’re volatile: on May 28, 2026, peer Anthropic officially announced it had closed a Series H round at a post-money valuation of $965 billion, formally overtaking OpenAI’s $852 billion round from March, currently the most highly valued AI startup. Rankings can shift, and this is the best proof of it; for the full breakdown on Anthropic’s side, see Anthropic’s valuation approaches a trillion dollars.
ARR is not the same as actual revenue
When discussing OpenAI’s revenue, you absolutely have to distinguish between two measures first, or it’s easy to be misled by the numbers.
- ARR (annualized revenue): an “annualized” figure derived by annualizing recent revenue run rate, landing on the order of $20 billion for 2025. It reflects the current pace of revenue, not that the full year actually books that much.
- Actual revenue: third-party estimates cited by media put what OpenAI actually booked in 2025 at around $13.1 billion (not confirmed by the company), clearly below the annualized estimate.
This doesn’t mean OpenAI is padding the numbers, fast-growing startups generally use ARR to convey momentum. But for readers, when you see “annualized revenue of $20 billion,” keep in mind that it’s a different thing from “actually earning $20 billion over the full year.”
A reminder that applies to any AI-startup figure: private companies have no audited financials, and revenue is mostly an estimate or the company’s own framing, it’s more grounded to grasp “order of magnitude and trend” than to chase precise values.
High valuation, but still burning cash heavily
Even as the valuation surges, OpenAI isn’t making money. It’s still in a phase of scaling losses, burning far more cash each year than it earns, with the biggest cost being that enormous compute bill: every time a user asks a question, computing resources get burned behind the scenes, the bigger the scale, the heavier the bill.
So there’s a tension here: the valuation reflects investors’ bet on the “future,” while the financials right now are continuous losses. Understanding OpenAI’s valuation is, at its core, about understanding how much the market is willing to pay for “what it might one day become.”
IPO: officially confirmed a confidential filing, but the timeline is all over the place
On June 8, 2026, OpenAI officially confirmed, in its own words, that the company had “recently submitted a confidential draft registration (S-1) to the U.S. Securities and Exchange Commission.” This is the first time the “OpenAI is going public” story that had been circulating for over a year got an official stamp of confirmation, rather than staying at “according to reports.”
What is a confidential S-1? It means privately submitting the filing to regulators for review first, without immediately disclosing financial details, and later converting it into a public filing depending on how things go — there’s still a stretch of process between this and an actual listing. Peer Anthropic, by contrast, publicly confirmed its own confidential filing all the way back on June 1; for the full breakdown, see The Anthropic IPO Explained.
That’s as far as the official line goes. What about the timeline? The company itself has stayed guarded, saying only “it may be a while,” with no specific date, no fundraising size, and no pricing range given.
What’s genuinely interesting is what happened over the following month: the same company got reported with three IPO timelines that don’t line up with each other.
- May 20 (before the official confirmation): Reuters reported that OpenAI was preparing a confidential filing and could list as soon as September 2026, naming Goldman Sachs and Morgan Stanley as the underwriting banks (the investment banks that handle selling a company’s new shares to investors and arranging listing pricing) involved. The report also noted this move came after the Musk lawsuit against OpenAI turned out in OpenAI’s favor, which the market read as removing a major piece of legal uncertainty.
- June 8: The company officially confirmed submitting a confidential S-1, while stressing the timeline “may be a while.”
- June 10: Financial outlet The Information reported, and Reuters relayed, that Altman told employees he expects the company to go public “within a year,” and that it plans to launch an employee tender offer “soon” (a tender offer: the company, or investors arranged by the company, buy shares from employees and early shareholders at an agreed price, letting them cash out early), with reports pointing to a price of around $687.69 per share [media / leaked insider information].
- June 25: The New York Times reported that OpenAI internally actually leans toward waiting until 2027 to go public, in order to protect its $1 trillion valuation target. The report said advisors gave the company only two options: wait until 2027 and push for $1 trillion, or go public sooner at a discounted valuation. CFO Sarah Friar reportedly also privately told some people that her real target is 2027.
Within a single month, “as soon as September” became “within a year,” which then became “possibly not until 2027.” This is the lesson most worth remembering about OpenAI’s IPO: a confidential filing buys optionality, not a timetable. Until there’s a public S-1, these dates are all just reporting and insider chatter, not official commitments.
Moving into July, the signs of preparation got more concrete. Reuters reported on July 8 that Bank of America had extended a first $520 million loan to OpenAI, while also competing for a role as an IPO advisor bank for both OpenAI and Anthropic — seen as one sign that Wall Street is starting to place its bets on these two IPOs. There’s a new wrinkle, too: on July 10, Apple sued OpenAI and two former employees in Northern California federal court over alleged theft of hardware-related trade secrets, and some reports treat it as a potential risk to the IPO timeline [media]. Earlier we said the Musk lawsuit wrapping up was a tailwind — the same logic cuts the other way, too. On July 21, OpenAI officially announced that Nubank founder and CEO David Vélez and BNY (Bank of New York Mellon) CEO Robin Vince were both joining the boards of the OpenAI Foundation and OpenAI Group PBC; the Wall Street Journal reported that Vince is also set to join the audit committee (the board subcommittee responsible for overseeing financial reporting and internal controls). This kind of governance strengthening is consistent with the direction of IPO prep, but on its own it isn’t evidence of listing timing.
CFO Sarah Friar’s posture can also be read as part of this “internal pace out of sync” thread. The Wall Street Journal reported in early May that she’d told investors privately that total planned spending through 2030 comes to around $600 billion, somewhat more conservative than the figure the company had floated publicly before. But the Wall Street Journal reported again on July 22 that OpenAI had revised its pre-2030 compute-spending plan upward to roughly $750 billion: the CFO reining things in on one hand while the spending plan keeps ballooning on the other is a fitting picture of that internal tension [media]. In a public interview in early June, she also described the IPO as “a milestone, not a finish line,” which lines up with the company’s official tone of “hold onto optionality first, figure out the timing later.”
When you see this kind of “officially confirmed, but the timeline keeps clashing” news, remember one thing: the filing is real, but whoever first reports “it’ll list in such-and-such month” doesn’t count — only a date the company itself puts in writing counts.
Pulling the threads together: the filing is real, and the company has confirmed it; but the listing date hasn’t been locked down in any company document, and as of July 24, 2026 there is still no public S-1 and no officially announced listing date. Whenever you see a headline claiming “OpenAI will list in [month],” it’s worth pausing to ask: is this the company saying it, or just another round of media relay?
An often-overlooked player: the nonprofit foundation
When discussing OpenAI’s equity, you can’t leave out its two-tier structure. The entity that actually operates is the public benefit corporation (PBC), but above it sits a nonprofit, the OpenAI Foundation, which holds roughly a quarter of the public benefit corporation’s equity and wields the power to appoint directors.
This means a sizable chunk of the valuation is “attributable to the nonprofit parent,” differing from the equity structure of an ordinary for-profit company. For the detailed power arrangements and the controversy over this structure, see the governance section in What kind of company is OpenAI.
Penchan’s take
OpenAI hasn’t gone public, so ordinary retail investors simply can’t buy its stock. The various “how to get in on OpenAI” routes floating around online are mostly indirect holdings, secondhand stakes, or derivative products, and you have to investigate the true exposure and risks yourself. Lumping it together with those public stocks that get hyped just for “touching OpenAI” is even more dangerous.
Valuation multiples, ARR, IPO timeline, these are more like an exercise in understanding a company’s financial constitution than market pricing, and certainly not entry or exit signals. Reading them clearly is about looking at this company more coolly, not about chasing the highs.