Industry News

Can OpenAI Outgrow Its Losses? AI Revenue Tops $10B in a Single Quarter

OpenAI's Q3 revenue reportedly crossed $10 billion for the first time, up nearly 50% from the prior quarter, even as its adjusted operating loss widened to $10.9 billion.

OpenAI just posted the strongest revenue quarter in its history — and somehow the most worrying loss figures at the same time. According to reports from Bloomberg and The Wall Street Journal, the company told investors this week that its third-quarter revenue crossed $10 billion for the first time, a jump of nearly 50% from the roughly $6.7 billion it booked in the second quarter. Its adjusted operating margin improved too, but the absolute operating loss grew larger, not smaller.

If the numbers hold up, they mark a genuine inflection in the business of AI. Not long ago, the story around OpenAI was stagnation: growth had cooled to about 18% quarter over quarter between Q1 and Q2, enterprise customers were drifting toward Anthropic, and the company's much-discussed profitability path looked more like a rumor than a plan. Now the growth curve has bent upward again, enterprise revenue has doubled in a single quarter, and OpenAI is openly using price cuts as a competitive weapon. But it is doing all of this while burning cash faster than ever — which raises the uncomfortable question in this article's title.

A few caveats matter before going further. These figures come from investor disclosures reported through unnamed sources; OpenAI has not published audited results and remains private, so the numbers are well-sourced estimates rather than confirmed financial statements.

The Numbers Behind the Quarter

Here is what was reportedly shared with investors:

  • Revenue of about $10 billion in Q3, up nearly 50% from $6.7 billion in Q2, and roughly 2.6 times the level of a year earlier.
  • An annualized revenue run rate near $50 billion, meaning OpenAI's recent quarterly pace, multiplied out across a year.
  • Adjusted operating margin improving from -150% in Q2 to -109% in Q3.
  • An adjusted operating loss that nonetheless grew, from $10.1 billion to $10.9 billion.
  • 1.2 billion weekly active users and more than 100 million paying subscribers.
  • Enterprise revenue doubling quarter over quarter, with business and consumer now each contributing roughly half of total revenue.

Two of these items deserve a closer look, because they seem to contradict each other and don't.

Better Margins, Bigger Absolute Losses

The margin improvement is real, but it is a ratio, not a dollar amount. When revenue grows 50% while losses grow roughly 8%, the loss as a percentage of revenue shrinks — from -150% to -109% in this case. That is genuinely good news: it means the company is getting more efficient at converting spending into revenue, and the trajectory is finally pointing in the right direction.

But the absolute number moved the other way. OpenAI lost about $800 million more in Q3 than in Q2 on the adjusted measure. And the adjusted measure itself excludes a large expense: stock-based compensation reportedly reached $2.6 billion in the quarter. In plain terms, the adjusted loss of $10.9 billion understates the true economic cost of running the business, because a meaningful chunk of employee pay arrives as equity rather than cash and is simply left out of the calculation.

This is the central tension of the AI buildout in one income statement. Frontier model development demands enormous upfront spending on compute and talent; the revenue eventually follows if the models are good and the pricing is right. OpenAI is now demonstrating the "revenue follows" half of that equation more convincingly than ever, but the "spending upfront" half has not slowed down to match.

There is also a framing trap in the run-rate figures that float around coverage of both OpenAI and its rivals. A $50 billion annualized run rate is not $50 billion of collected annual revenue — it is the current quarterly pace extrapolated forward. It answers "how fast are we going right now," not "how much did we earn," and readers should always check which framing is being quoted.

Where the Growth Actually Came From

The most consequential shift this quarter is structural, not just quantitative. OpenAI president Greg Brockman has said the enterprise business doubled in the quarter, and the company now describes enterprise and consumer as roughly equal halves of revenue. That is a very different company from the one that grew up on $20-per-month ChatGPT subscriptions.

The drivers, according to the company, were new frontier models, deliberate price reductions, and a wave of enterprise-facing products — coding tools, cybersecurity offerings, and agent platforms designed to complete real work rather than answer chat prompts. The pattern matters more than any single product: enterprise buyers purchase AI the way they purchase other infrastructure, by evaluating whether it saves money or produces output in production environments. Once a tool passes that test, budgets follow, and those budgets are typically larger and stickier than consumer subscriptions.

The consumer side did not stall, either. Subscriptions crossed 100 million paying users, weekly active users reached 1.2 billion, and OpenAI's advertising product reportedly hit a $1 billion annualized run rate as far back as August. Advertising is still small next to $10 billion in quarterly revenue, but it opens a second monetization channel for the enormous free-user base below the paying tier — a lever Anthropic, which has no consumer product, simply does not have.

The Anthropic Gap — and Why You Can't Compare the Two Directly

Any discussion of OpenAI's numbers now has to run through Anthropic, which reported its own Q3 revenue of about $17 billion and has become the enterprise market's default alternative. Anthropic also announced that its annualized run rate climbed roughly $10 billion in a single month, reaching about $81 billion as of the end of October.

Before anyone concludes that Anthropic is nearly twice OpenAI's size, one detail has to be on the table: the two companies count revenue differently. Anthropic includes sales made by its cloud distribution partners in its reported figures. OpenAI reports only the share of partner-channel sales that belongs to it. The comparison is directionally meaningful — Anthropic's enterprise momentum is real — but the raw numbers are not an apples-to-apples score.

What is unambiguous is the competitive dynamic in the enterprise segment. Anthropic moved first with products enterprises actually adopted, and OpenAI spent much of the year playing catch-up. Its response this quarter was twofold: sharper pricing on new models, and a public relations push aimed directly at the low-cost narrative. CFO Sarah Friar has called China's apparent lead in low-cost AI models one of the biggest myths of 2026, pointing to GPT-6 Luna, the company's model for high-volume work, priced at $0.10 per million input tokens and $0.50 per million output tokens.

That pricing tells you what OpenAI thinks the market has become. When your flagship low-cost model undercuts the kind of prices that seemed aggressive a year ago, you are no longer selling exclusivity — you are selling throughput. Friar has also been candid about the strategic lesson behind the price war: Anthropic's success distributing through third-party cloud providers convinced her that OpenAI needs to break out of its historic dependence on a single partner relationship, and she has framed broader cloud distribution as a major growth lever for 2027.

Price Cuts as a Weapon, and What They Signal

There is a deeper signal in the pricing moves. If OpenAI — the company with the largest consumer distribution in AI — is cutting prices on frontier-adjacent models while revenue accelerates, it is betting that volume and enterprise lock-in will more than compensate for thinner per-token margins. That bet only makes sense in a market where open-weight and Chinese models have collapsed the price floor for "good enough" intelligence. The reasoning is straightforward: when good-enough is nearly free, being excellent must either justify a premium or be delivered so cheaply that switching costs do the work.

The catch, of course, is that price cuts and surging losses are awkward companions. OpenAI is improving margins while reducing what it charges per unit of output — so the margin gains have to come from efficiency: cheaper inference, better compute utilization, models that deliver more value per dollar of training cost. The -109% adjusted operating margin suggests those gains are happening but are nowhere near sufficient yet. The quarter's arithmetic only closes if growth keeps compounding faster than costs, and nobody can promise that.

The IPO Context Makes This Week's Numbers a Preview

The timing of these disclosures is not accidental. OpenAI has reportedly raised funding at a valuation around $1.25 trillion and has postponed its own IPO to next year. Anthropic, meanwhile, launched its IPO roadshow this week with a price range of $70 to $78 per share, seeking a valuation of up to roughly $1.4 trillion, with trading expected within days.

Seen in that light, the same-week revenue revelations from both companies look less like routine transparency and more like the private-market equivalent of a pre-IPO earnings season. Investors weighing Anthropic's offering want to know whether OpenAI is collapsing behind it (it is not — growth re-accelerated). Investors tracking OpenAI's future listing want to know whether the loss curve is bending (the margin says yes; the absolute loss says not yet).

One more comparison deserves honesty: neither company is profitable in any ordinary sense, and both owe their compute bills to the same small set of chipmakers and cloud providers. The revenue race is real, but it is a race run on borrowed infrastructure, financed by increasingly large circular investment flows. When the two largest AI revenue stories go public within a year of each other, public-market investors will get their first real look at the unit economics behind the boom — and the gap between the run-rate narratives and audited financials will finally be measurable.

What Builders and Buyers Should Take From This

For anyone choosing AI tools rather than investing in them, the most useful lesson from this quarter is not about stock tickers at all. It is about how competition is actually being fought: not on leaderboard rankings, but on cost per task. OpenAI is cutting prices because enterprises compare what it costs to complete a unit of work, not what a model scores on a benchmark. Anthropic's enterprise lead exists because customers ran those calculations and liked the answers.

That is exactly the right way to evaluate tools for your own work, too. When you compare an AI writing assistant, a coding agent, or a research tool, the question that matters is what it costs to get your actual task done reliably — subscription price, token pricing, and how much retrying and editing the output requires. Comparing models by a single benchmark number is like comparing airlines by their top speed; it ignores everything that determines what you'll actually pay and experience.

This is the lens we use when organizing tools at AI Bookmarks: hundreds of AI tools, sorted by task category and pricing model, so you can line up options on the same terms — per-task cost, free tier availability, and what the pricing actually covers — before committing to any of them. In a market where the providers themselves are racing to the bottom on price, the winners are the users who compare like buyers rather than like fans.

The Road Ahead

Three things are worth watching as this story develops. First, whether enterprise revenue can keep doubling — that is the load-bearing wall of the entire valuation story, and enterprise budgets are subject to procurement cycles that can stall momentum quickly. Second, whether continued price cuts flatten the margin improvement; the -109% figure has room to improve, but only if inference efficiency keeps rising. Third, Anthropic's trading debut, which will give the entire AI sector its first public-market price anchor and, by extension, a market verdict on what these revenue trajectories are worth.

OpenAI's $10 billion quarter answers one question and sharpens another. The answered question: demand for frontier AI is growing fast enough to fund a genuine business, not just a research lab with a donations jar. The sharpened one: can the company's costs bend down toward its revenue before the capital that bridges the gap gets impatient? The next four quarters — with both leading labs either public or about to be — will finally put that question to a public test.

---

This article synthesizes reporting from TechCrunch, Bloomberg, The Wall Street Journal, and The Decoder, based on figures OpenAI shared with investors. OpenAI has not formally confirmed the reported figures; Anthropic's figures come from its own public disclosures, and the two companies' revenue metrics are calculated differently, as noted above.

— The AI Bookmarks News Desk

Published by AI Bookmarks · More AI news →

← Back to News