OpenAI Revenue in 2026: What's Actually Reported, Number by Number

OpenAI's revenue for fiscal 2025 is reported at roughly $13 billion, with annualized run-rate reportedly near $20 billion by the end of the year and heading toward $25 billion by mid-2026, per Sacra's estimates. Every one of those numbers needs a footnote, because OpenAI is privately held and doesn't publish audited financials. What gets called “OpenAI revenue” in most headlines is somebody's estimate, not a filed number — and mixing up which estimate you're looking at is where almost all the confusion starts.

This piece sticks to what's actually been reported, names the source and the date for each figure, and explains why the same company can be described as making “$13 billion” in one article and “$25 billion” in another without either one being wrong.

How Much Revenue Does OpenAI Make?

OpenAI's actual fiscal 2025 revenue is reported at approximately $13 billion, based on aggregated estimates from Sacra and Statista. That's the closest thing to a real top-line number publicly available — money reportedly earned over the year, not a forward projection.

Annualized recurring revenue (ARR) is a different, faster-moving figure. Sacra's estimates put OpenAI's ARR at roughly $20 billion at the end of 2025, climbing toward $25 billion by mid-2026.

The difference between those two numbers is not a rounding error — it's a methodology gap, and it's worth understanding once instead of getting tripped up by it every time a new headline runs:

• Revenue is money earned over a completed period, like a fiscal year. It looks backward.
• ARR takes current monthly recurring revenue and multiplies by 12. It looks forward, assuming today's run rate holds for a full year.

For a company adding subscribers and API customers every month, ARR will almost always sit above trailing revenue. That's normal. It's also why most “OpenAI made $X billion” headlines are quietly citing ARR, not revenue — a distinction the headline itself rarely makes.

How Does OpenAI Make Money?

Two products carry nearly all of it. Per Sacra's breakdown, roughly 70% of OpenAI's reported revenue comes from ChatGPT subscriptions, and about 25% from API access, with the remainder split across smaller lines.

• ChatGPT subscriptions (~70% reported) — Plus, Pro, Team, and Enterprise plans sold directly to consumers and businesses. This is the visible, consumer-facing product.
• API access (~25% reported) — usage-based revenue from developers and companies building their own products on top of OpenAI's models. This is the invisible layer, embedded inside other people's apps.

That mix is the more interesting fact than either percentage on its own. A company built around a research lab and a developer platform is, by reported revenue, mostly a consumer subscription business. ChatGPT's dominant position in the consumer chatbot market (see our ChatGPT market share breakdown) is a big part of why subscriptions bring in far more than API revenue — and it means OpenAI's near-term financial fate rides more on consumer retention than on developer adoption.

Is OpenAI Profitable?

No — not according to current reporting. OpenAI's 2025 spending is reported at around $34 billion, per Ed Zitron's reporting at Where's Your Ed At, against roughly $13 billion in reported revenue. That gap is the central fact of OpenAI's current financial story.

Compute is the reason. Training frontier models and serving inference to a massive, still-growing user base is expensive at a scale most companies never approach, and compute costs reportedly dominate the spending side of the ledger. Growth doesn't close that gap by itself — more users and more prompts mean more inference cost, which is part of why spending has reportedly scaled roughly in step with (or ahead of) revenue rather than behind it.

Investors are reportedly tolerating the gap anyway, for reasons that are more strategic than financial:

• Early, default-model status in a category this new may be worth more than near-term margin.
• Inference costs are widely expected to fall as hardware and model efficiency improve.
• Enterprise contracts typically carry better margins than consumer subscriptions and are still a small share of the reported mix.

That said, reported spending running well ahead of reported revenue isn't a rounding error. It's worth taking seriously as a real gap rather than assuming scale quietly closes it on its own — OpenAI isn't alone in running at a loss during this build-out phase, either; see how Anthropic's revenue and spending reportedly compare.

What Are OpenAI's Key Reported Numbers?

Every figure below is sourced and dated. Treat anything you read elsewhere without both attached the same way you'd treat an unsourced stat — skeptically.

Metric	Reported figure	As of	Source
Fiscal 2025 revenue	≈$13B	FY2025	Sacra / Statista aggregations
Annualized recurring revenue (ARR)	≈$20B	End of 2025	Sacra estimates
ARR (projected)	≈$25B	Mid-2026	Sacra estimates
Revenue mix — ChatGPT subscriptions	≈70%	2025	Sacra
Revenue mix — API	≈25%	2025	Sacra
Reported spending	≈$34B	2025	Ed Zitron, Where's Your Ed At
ChatGPT weekly active users	800M	Oct 2025	OpenAI, via TechCrunch
Daily prompt volume	2.5B+	Jul 2025	via TechCrunch

Why Do OpenAI's Revenue Numbers Vary by Outlet?

Because OpenAI is private, every published figure is a third-party estimate built on its own methodology — and ARR, trailing revenue, and run-rate measure genuinely different things, even though headlines rarely say which one they mean.

• ARR — a snapshot multiplied by 12, forward-looking, always the highest of the three during a growth phase.
• Trailing (GAAP-style) revenue — money recognized over a completed period. The closest thing to “real,” and the smallest number of the three.
• Run-rate — similar to ARR in spirit, but sometimes calculated over a shorter window (a single quarter times four), which means run-rate and ARR can disagree even inside the same report.

There's no SEC filing to check any of this against, no 10-K, no audited statement — just investor materials, leaks, and modeling from research firms like Sacra. Treat any headline that states “OpenAI made $X billion” without naming which metric it means as incomplete at best. Naming the metric, the source, and the date — which is the convention this article follows throughout — is a low bar, and most coverage still doesn't clear it.

What's the Scale Behind OpenAI's Revenue?

ChatGPT reportedly reached 800 million weekly active users as of October 2025, according to Sam Altman via TechCrunch. The service was also processing more than 2.5 billion prompts a day as of July 2025 reporting.

That's the usage base the subscription revenue is actually drawn from, and it's why the ~70% ChatGPT figure carries so much weight in the overall mix. For the fuller set of adoption numbers behind it, see our dedicated ChatGPT statistics roundup.

• 800M weekly users (Oct 2025) — a meaningful share of the internet-using population checking in with ChatGPT every week, and the pool the ~70% subscription revenue is converted from.
• 2.5B+ prompts/day (Jul 2025) — each one a moment where a model might mention, recommend, or ignore a given brand or product.

That second bullet point matters more than it looks like it should.

Why Does OpenAI's Revenue Matter Beyond Finance?

Here's the part most “OpenAI revenue” coverage skips entirely: the same system handling those 2.5 billion daily prompts is also, increasingly, the system deciding which brands, tools, and products get mentioned when someone asks for a recommendation. That's not a neutral utility sitting quietly in the background — it's a business optimizing for engagement and retention, and one that, per its own reported economics, still needs to close a substantial revenue gap.

OpenAI isn't the only model competing for that role, either. The market is going multipolar — Anthropic, Google, Perplexity, and others are all fighting for the same query volume, which means visibility inside any single assistant is no longer enough on its own. That broader shift in how people search is covered in our AI search statistics piece.

The actual stake for anyone running a brand isn't OpenAI's income statement — it's whether your product shows up at all when someone asks an AI assistant to compare options. A company can be running at a massive reported loss and still be the place buyers ask “what should I use for this?” every single day. Tools like AEOeye exist to answer a narrower, more useful question: whether ChatGPT and its peers are recommending you right now, and exactly where you disappear from the answer.