OpenAI Revenue 2026: $25B ARR, a $20.9B Leaked Loss, and Why Anthropic Just Passed It

$25 billion annualized run rate for OpenAI as of early 2026 — but leaked audited financials show it actually booked just $13.07B in 2025 revenue against a $20.9B operating loss, and Anthropic's ARR just blew past it to $69B. That's the short answer. The longer answer is more interesting.

OpenAI has spent 2026 talking in run-rate language — $2 billion a month, a $25 billion annualized pace, a record $122 billion funding round. All of that is true. What's also true, thanks to a June 2026 leak of its audited financial statements to the Financial Times, is that the company's actual booked 2025 revenue was $13.07 billion against a $20.92 billion operating loss, and that its closest rival just lapped it on growth. Both stories are real, and an investor who only reads the press releases is missing the second one entirely.

What Is OpenAI's Revenue in 2026?

OpenAI's annualized revenue run rate reached $25 billion by February 2026, up 17% from the $21.4 billion run rate it closed 2025 with, and has held roughly flat in that range through the spring and into summer. That run-rate figure is a projection — it takes the most recent month of revenue, currently about $2 billion, and multiplies by 12. It is not the same thing as revenue the company has actually booked and recognized over a full year, a distinction that matters enormously once you look at the audited numbers underneath it.

Full-year 2025 revenue, per financial statements obtained by blogger Ed Zitron and independently verified by the Financial Times, came in at $13.07 billion — more than tripling from $3.7 billion in 2024, which itself was up from roughly $1.6 billion in 2023. Growth has been extraordinary by any normal company's standard. It has also flattened: OpenAI's run rate held near $25 billion from February through the spring of 2026 without meaningfully accelerating, a pause worth watching given how fast the market has been pricing in continued exponential growth.

The Leaked Financials: A $20.9B Operating Loss

The headline number from the June 2026 leak is a $20.92 billion operating loss on $13.07 billion of 2025 revenue — a loss larger than the revenue that produced it. Total costs and expenses came to roughly $34 billion for the year, split between $19.18 billion in research and development spending and $5.73 billion in sales and marketing. Of that $34 billion, $17.2 billion flowed directly to Microsoft for Azure cloud infrastructure and compute — meaning compute costs to a single vendor alone exceeded OpenAI's entire annual revenue.

The net loss figure is even larger: $38.5 billion, once you layer in one-time, non-cash accounting charges tied to OpenAI's 2025 conversion from a nonprofit-controlled structure to a for-profit public benefit corporation, which produced a roughly $41.55 billion charge from changes in the fair value of convertible interests and warrant liabilities. Strip that structural noise out and the underlying operating loss of $20.9 billion is still the number that matters for anyone modeling the business, not the smaller headline losses OpenAI itself has emphasized in public commentary.

Looking forward, OpenAI's own internal forecasts reportedly project close to $14 billion in losses for 2026 on the higher revenue base, with some outside analysts projecting cash burn as high as $27 billion for the year. Longer-run projections vary wildly depending on the source: one internal scenario has OpenAI losing a cumulative $44 billion from 2023 through 2028 before turning a $14 billion profit in 2029, while a more conservative estimate puts cumulative cash losses closer to $115 billion through 2029, and OpenAI's own internal planning reportedly models cumulative cash burn reaching $218 billion by 2030 in a slower scenario.

OpenAI vs Anthropic Revenue: Who's Actually Winning in 2026?

This is the part of the OpenAI revenue story that most 2026 coverage has missed. At the end of 2025, OpenAI's $21.4 billion run rate dwarfed Anthropic's $9 billion. Seven months later, that gap has not just closed — it has inverted. Anthropic's run-rate revenue climbed from $9 billion in December 2025 to $14 billion in February, $19 billion in March, $30 billion in April, $47 billion in May, and an estimated $69 billion by early July 2026, an increase of well over 80x from its January 2024 starting point of $87 million ARR. OpenAI's $25 billion run rate, by comparison, has been essentially flat since February.

Figures blended from OpenAI's leaked audited financials (Financial Times, June 2026), Anthropic's fundraising disclosures reported by SaaStr and VentureBeat, and Bloomberg's coverage of both companies' 2026 funding rounds. Anthropic's growth is run-rate ARR, an annualized projection, not booked revenue; OpenAI's $13.07B figure is actual booked 2025 revenue.

The comparison isn't perfectly apples-to-apples — Anthropic's numbers are run-rate ARR announced alongside fundraising, the same kind of projection OpenAI itself leans on for its $25B figure, and Anthropic hasn't had its own financials leaked the way OpenAI's have. But directionally, the signal is hard to miss: the company built around enterprise API consumption and coding-assistant demand is compounding faster than the company built around a mass-market consumer chatbot, at exactly the moment enterprise budgets are the thing everyone in AI is fighting over. For more on how the frontier labs stack up on pricing and performance, see our breakdown of Anthropic vs OpenAI on safety, performance, and pricing.

Where OpenAI's Revenue Actually Comes From

Roughly 70% of revenue still comes from ChatGPT consumer and Plus/Pro/Business subscriptions, spread across an estimated 20 million paid seats. Another 25% comes from API usage — developers and enterprises paying per token for GPT-5-class models embedded in their own products. The remaining 5% is a grab bag of Sora video generation, licensing deals, and a new advertising pilot inside ChatGPT that reportedly crossed $100 million in ARR within just six weeks of launch, an early signal that OpenAI is serious about diversifying beyond subscriptions.

The more important shift is enterprise mix: enterprise revenue now exceeds 40% of the total and is on pace to reach parity with consumer revenue by the end of 2026. With roughly 700 million weekly active users across ChatGPT, OpenAI's blended revenue per weekly active user works out to about $36 a year — a figure that looks thin next to Anthropic's far smaller, far more concentrated, and far higher-paying enterprise and developer base. If OpenAI is going to close the growth gap with Anthropic, enterprise and API revenue, not consumer subscriptions, is where it has to happen. Track how the broader AI valuation picture is shifting on our AI valuations dashboard.

The $122B Raise and What It's Actually Funding

OpenAI closed a $122 billion funding round on March 31, 2026 at an $852 billion post-money valuation — up from the $110 billion the company had previously guided to. Amazon committed $50 billion, with $35 billion of that contingent on OpenAI going public or reaching an internally defined AGI milestone. Nvidia and SoftBank each put in $30 billion, and OpenAI opened the round to individual investors through bank channels for the first time, raising $3 billion from that tranche alone.

Nearly all of that capital is earmarked for one thing: compute. With $17.2 billion already flowing to Microsoft Azure in 2025 alone and 2026 Azure-related compute spend estimated near $13 billion on top of other infrastructure commitments, OpenAI's cash needs are effectively unbounded relative to its current revenue base. That is the core tension in the OpenAI 2026 revenue story: a company with a genuinely enormous and still-growing top line that is nonetheless burning cash faster than almost any private company in history, betting that scale now converts into pricing power and margin later.

What OpenAI's Revenue Numbers Mean for Investors

For LPs and crossover investors underwriting the next OpenAI round, or evaluating secondary exposure, three numbers matter more than the $25 billion headline: the $20.9 billion 2025 operating loss, the flattening of ARR growth since February, and the widening gap with Anthropic. A company can lose money while scaling and still be a great investment — that's the entire venture model — but a plateauing top line combined with accelerating losses is a different and more concerning pattern than hypergrowth combined with losses.

For operators, the practical read is that OpenAI's talent costs alone run near $4 billion annually across roughly 4,500 employees, or about $900,000 fully loaded per employee — a cost structure only a company burning $20 billion a year could sustain, and a data point worth keeping in mind before assuming AI-native headcount economics translate to smaller, better-capitalized startups. It's also a reminder that headline ARR and run-rate figures, whether from OpenAI or any other AI company, deserve the same scrutiny a public-market analyst would apply to a 10-K, not the press-release read.