AI Companies by Revenue 2026: The Leaderboard, the Sources, and Which Are Profitable

Two companies are in a league of their own. Anthropic reported a run rate above $65 billion at the end of July, and its IPO filing now shows $11.56 billion of revenue in the second quarter of 2026 alone. OpenAI's September run rate was put at roughly $50 billion by the Financial Times and Bloomberg and near $70 billion by Axios. The next company on the board, Databricks, is at about $7 billion, and Cursor, now part of SpaceX, booked $608 million in its first 48 days inside the group. Everyone else is at roughly a billion or below.

This is the November 2026 edition, checked on 4 November. Every row says where its number comes from, when it dates from, and what it measures, because these companies do not report the same thing. A run rate, a year of booked revenue and an analyst estimate are not interchangeable, so read the table as "latest known figure, largest first", not as a like-for-like race.

November edition: Anthropic filed its S-1 with the SEC on 26 October, so its revenue, accounting basis and profit are now stated by the company itself rather than reported from investor documents. SpaceX's third-quarter report gave the first figure for Cursor since the acquisition. No newer figure was found for the other nine. Full list in the change log.

The board, November 2026

Why the first two share a place. Anthropic's figure is above $65 billion for July; OpenAI's two September figures run from about $50 billion to nearly $70 billion. The range crosses Anthropic's number, the months differ, and the two companies count partner sales differently. On these figures nobody can say which of the two earns more on a comparable basis, so this edition does not rank them against each other.

Anthropic reports on a gross basis. It counts the full end-customer spend that comes through cloud resellers (AWS, Google, Microsoft) as revenue and books the partners' share as an expense. Its S-1 now states this in its accounting policies: for sales through cloud marketplaces Anthropic is the principal and "recognizes revenue on a gross basis". OpenAI treats partner sales differently. That is a large part of why OpenAI has two numbers: the FT reports that the ~$70 billion figure came from investors applying Anthropic's method to OpenAI. Why the same company has two revenue numbers explains the gap and which figure to use for what.

From row 5 down the figures are close enough that the order means little. Scale AI's figure is a full year of booked revenue, Copilot's is an estimate, Cognition and Lovable report a run rate, and Midjourney's is two years old. A company growing fast can pass the one above it between two disclosures.

GitHub Copilot. Microsoft has never disclosed Copilot revenue. The $0.9-1.1 billion is an estimate by Axis Intelligence Research, built on the last paid-subscriber count Microsoft published (4.7 million in January 2026) and public pricing tiers; the FT put a floor of at least $550 million using the cheapest tier only. In July Microsoft said Copilot now has 50 million users (all users, not paid seats) and that Copilot revenue grew more than 60% quarter on quarter after the move to usage-based billing. The January-based estimate is therefore probably below the current figure, and we print it as dated rather than guess a new one.

Private-company figures are what the company said or what credible outlets reported. None of them is audited. When a company publishes a newer number, the row changes and the change log says so.

What SpaceX buying Cursor and xAI changed

Cursor (Anysphere) was the fastest-growing company on the August board. Forbes reported it above $4 billion in annualised revenue in early June. SpaceX bought it in an all-stock deal at an implied equity value of $60 billion, closed on 14 August 2026 (SpaceX Form 8-K). It stays on the board, marked as owned by SpaceX, because what this business earns is still the question readers ask.

SpaceX's third-quarter release on 20 October gave the first figure since the deal: from the closing date through 30 September, Cursor contributed $608 million of revenue. That is 48 days. Spread evenly over a year it would come to roughly $4.6 billion, our own arithmetic, not a SpaceX figure, and consistent with the June number. The release also puts the GAAP purchase price at $42.5 billion.

xAI became part of SpaceX in February 2026, before SpaceX listed on Nasdaq in June. Its revenue now appears inside SpaceX's AI segment, renamed Super Intelligence: $7.6 billion of revenue in Q3 2026, up from $2.56 billion in Q2, and income from operations of $431 million, after a $1.26 billion operating loss in Q2 (SpaceX Q2 and Q3 earnings releases). That segment also includes X advertising and subscriptions, cloud services deals and now Cursor, so it is not an "xAI revenue" figure and cannot sit in this table.

The rule is about whether the revenue can still be traced, not about who owns the company. An acquired business stays, marked with its owner, while its own revenue can be identified. When it disappears into a parent's accounts, as xAI's has, it leaves the table and we say where it went.

What the table shows

The table mixes run rates, booked revenue and estimates, so it would be dishonest to add the rows up. What it does show without arithmetic:

• The two frontier labs are an order of magnitude ahead. Databricks, third, is at roughly a tenth of Anthropic's figure.
• Below the top four, the companies on this board do around a billion or less. Seven of the eleven rows are at or near $1 billion or under it.
• The fastest risers are application companies. Cursor, Cognition and Lovable sell tools that sit on top of the labs' models, and they grew into this range within about a year.

Key takeaways

Two leaders, no settled order, and only one reports a profit

Anthropic's S-1 puts its own numbers on the growth that had only been reported second-hand. Revenue was $386 million in 2024 and $4.59 billion in 2025, then $4.73 billion in the first quarter of 2026 and $11.56 billion in the second: $16.29 billion in six months, fourteen times the same half a year earlier. The run rate shared with investors went from roughly $9 billion at the end of 2025 to $47 billion in May and above $65 billion at the end of July.

On the quarterly figures Anthropic is ahead: $11.56 billion against OpenAI's $6.7 billion in Q2 2026. Same quarter and no annualising, but still not fully comparable, because Anthropic counts sales through cloud partners on a gross basis and OpenAI does not. That is why this page does not call either company number one.

The profit question now has a company answer too. In Q2 2026 Anthropic had operating income of $974 million, after an operating loss of $1.67 billion in Q1; for the half year that is an operating loss of $693 million, down from $8.06 billion for all of 2025. The net loss for the half, $80.7 billion, is mostly a $72.4 billion non-cash change in the value of investor instruments that convert into shares at the IPO. In September the FT reported that Anthropic had told investors Q3 would be profitable too; the S-1 covers results only to June. We covered the shift in the first big lab to make money and the filing itself in what Anthropic's S-1 shows.

One caution that still stands: the often-quoted "$559 million profit" was a projection made in May. The reported Q2 result is $974 million of operating income.

In September the New York Times and Axios reported Anthropic "pacing" to more than $100 billion of revenue. The metric behind that phrase was not defined, so it is not in the table.

OpenAI: the biggest consumer business, the biggest losses

OpenAI expects to book about $36 billion of revenue across 2026, according to investor materials reported by the FT, and told investors it expects its run rate to reach at least $70 billion by the end of the year. The same materials project cumulative negative free cash flow of roughly $278 billion between 2026 and 2030. Its last closed round valued it at $852 billion; talks reported in September and October put it between $1.2 and $1.4 trillion, not closed.

Revenue per employee is the most abused number in AI coverage

This page used to carry the line everyone else carried: Cursor doing $2 billion with about fifty people. We corrected it in August, when Anysphere's own figure of over 300 staff put it closer to $6.7 million per head. That headcount is now almost a year old and Cursor is part of SpaceX, so this edition publishes no revenue-per-employee figures at all.

The reason is dates. Of the eleven, only ElevenLabs has stated its headcount recently: over 800 people on 30 September. Its latest company-wide revenue figure is from May. Dividing a May number by a September headcount is exactly the mismatch that produced the famous "$40 million per head". Midjourney's popular per-employee figure has the same problem: its headcount estimates range from about 40 to 163.

The insight survives without the arithmetic: AI-native companies earn far more per person than software companies did a few years ago. The same economics are behind the solo founders and tiny teams building real income with AI.

Bootstrapped can still win

Midjourney has raised no outside money, booked about $300 million in fiscal 2024 per Forbes and PitchBook, and is reported profitable by Forbes and by its CEO. Its financials are private, and no valuation has ever been set by a priced round, so the "$5 billion" and "$10 billion" figures that circulate online have no source we can find. You do not need billions to build a profitable AI company. You do need them to train frontier models and sit in the top three.

Which AI companies are actually profitable?

Revenue rankings quietly dodge this question, so here is the straight answer: only a few of them say anything about profit, and the largest consumer business is losing money on purpose. Training and serving models is paid for up front; revenue arrives month by month.

What is known as of 4 November 2026:

For comparison, SpaceX's AI segment, which contains xAI and since August Cursor, swung from an operating loss of $1.26 billion in Q2 2026 to income from operations of $431 million in Q3.

Two honest conclusions. First, the most profitable AI businesses are probably not on this list at all: the reliable money in this cycle has been made selling chips and cloud capacity. Second, when a headline calls an AI company "profitable", check whether it means operating profit, adjusted profit, cash flow or a projection. In 2026 the answer is rarely the first.

What this means

For entrepreneurs: The companies rising fastest on this board, Cursor, Cognition and Lovable, sell tools built on top of other companies' models. That is the layer a small team can enter. The cheapest on-ramp is still selling a service, as I argue in why AI belongs in your services, not just your back office, and if you run a small business, start with the AI implementation playbook for small business, no CTO required.

For job seekers: Companies doing tens of billions in AI revenue are hiring hard and paying up. The premium goes to people who can use AI, not just describe it. See the highest-paying AI jobs of 2026.

For investors: Anthropic filed its S-1 on 26 October to list on Nasdaq under the ticker ANT; the price range and the date come in an amended filing, which is also where its run rate as of 31 October, left blank in the first filing, should appear. OpenAI says it will not list in 2026. What Anthropic's S-1 shows walks through the filing, and the AI IPO race covers what to watch.

Frequently Asked Questions

Which AI company makes the most money in 2026?

Anthropic and OpenAI, and on the published figures it is not possible to say which of the two earns more. Anthropic reported a run rate above $65 billion at the end of July 2026, and its S-1 shows $11.56 billion of revenue in Q2 2026; OpenAI's September run rate was reported at roughly $50 billion (FT, Bloomberg) and near $70 billion (Axios), and its Q2 revenue at $6.7 billion. Anthropic counts sales through cloud partners on a gross basis and OpenAI does not, so the numbers are not like for like. Until both publish figures on the same basis, first place is not settled.

Which AI companies are profitable?

Anthropic's S-1 shows operating income of $974 million in Q2 2026, after a loss in Q1, and the company has told investors it expects Q3 to be profitable too. Midjourney is reported profitable. Databricks reports positive adjusted free cash flow. OpenAI is lossmaking by design. The others on the board do not disclose profit.

How much revenue do AI companies make?

It falls off a cliff after the top two. Anthropic and OpenAI are in the tens of billions, Databricks is around $7 billion, Cursor was above $4 billion in June, and the next seven on this board are between roughly $0.3 billion and $1 billion.

Which AI company has the highest revenue per employee?

There is no reliable answer this month, so we do not publish one. Only ElevenLabs has stated a recent headcount, and its revenue figure is from a different month. Treat any per-employee figure in AI coverage with suspicion, including the "$40 million per head" for Cursor that this page itself carried until August 2026.

Why is xAI not on the list, and why is Cursor marked SpaceX?

SpaceX bought both. Cursor's own revenue can still be traced, so it stays on the board marked as owned by SpaceX. xAI's revenue now sits inside SpaceX's AI segment together with X advertising and cannot be separated, so it left the table.

Cutting through the noise

The scoreboard is clear about direction and vague about precision. One leader reports a run rate above $65 billion, the other has two credible figures $20 billion apart that straddle it, and most of the board is around a billion. Mind the gross-versus-net caveat, mind that private figures are unaudited, and read the date on every number. These are real revenues from real customers, in a market that reorders itself every few months.

Methodology, sources and limits

What gets included. AI-native businesses whose own revenue comes primarily from AI products they build and sell, and can be identified. Ownership does not decide it: Cursor (SpaceX) and GitHub Copilot (Microsoft) stay because their revenue can still be traced or estimated from disclosures, and each is marked with its owner. That excludes the biggest AI earners of this cycle, Nvidia and the cloud providers, because their AI revenue is not separable from hardware and infrastructure, and it excludes xAI since its revenue was folded into SpaceX's AI segment.

Order. Rows are ordered by the size of each company's latest published figure. Anthropic and OpenAI share first place because their figures overlap and are not on the same basis. Because the figures measure different things, the order below the top four is approximate too.

Where the numbers come from. Company statements and filings where they exist; reporting by established outlets (Bloomberg, Reuters, the FT, the WSJ, Forbes, CNBC, Axios, TechCrunch) where they do not. Every row names its source and its date. Where credible sources disagree, we print both rather than pick the flattering one, as with OpenAI.

Run rate, booked revenue, estimate. A run rate is the latest month multiplied by twelve. Booked revenue is what came in over a past period. An estimate is someone else's calculation. For companies growing at these speeds the three can differ by a factor of two, and mixing them without saying so is the most common error in AI revenue coverage.

If no reliable newer figure exists, the row keeps the last figure we can confirm, with its date, instead of a fresher number we cannot trace.

Headcount. We publish a per-employee figure only where the company stated its headcount and the headcount and the revenue figure date from the same period. This edition has none.

Verification. All eleven rows were re-checked on 4 November 2026. Anthropic and Cursor have new figures; for the other nine no newer published figure was found, so they keep their dates.

Corrections. If a figure is wrong and you can show us, we correct the row and say so below. Corrections are published, not quietly edited.

Change log

• 4 November 2026 (November edition): Anthropic: added Q2 2026 revenue of $11.56 billion and the full revenue history ($386M in 2024, $4.59B in 2025, $4.73B in Q1 2026) from its Form S-1 filed on 26 October; the gross-basis accounting is now quoted from the S-1; profit status changed from "positive adjusted operating income, no amount" to operating income of $974 million in Q2 after a $1.67 billion loss in Q1. Cursor: added the first figure since the acquisition, $608 million of revenue booked by SpaceX between 14 August and 30 September (SpaceX Q3 release). SpaceX AI segment (xAI, Cursor) updated to Q3: $7.6 billion of revenue and $431 million of income from operations. Other nine rows re-checked, no newer figures.
• 9 October 2026 (second review): Anthropic and OpenAI now share first place: OpenAI's two September figures straddle Anthropic's July figure and the two count partner sales differently, so a comparable leader cannot be named; the FAQ says so. The GitHub Copilot estimate now names its source, Axis Intelligence Research, next to Microsoft's subscriber disclosure it is built on. Wording fixes: "a billion or below" made "roughly", since the Copilot range reaches $1.1 billion, and a leftover "top ten" removed.
• 9 October 2026 (same day, after review): Cursor restored to the table, marked as owned by SpaceX, with its last published figure (above $4B annualised, June 2026); the inclusion rule now depends on whether a business's own revenue can be traced, not on ownership, which also explains why GitHub Copilot stays and xAI does not. Anthropic versus OpenAI wording softened: the Q2 figures are the same quarter but not fully comparable, because the two account for partner sales differently. Dataset: open-ended figures such as "above $65B" were stored as exact values (65 to 65); they now have an open upper bound and a revenue_bound column.
• 9 October 2026 (October edition): Every row re-checked; each now shows what the figure measures, its date and its source. Cursor and xAI removed after their acquisition by SpaceX (xAI in February, Cursor closed 14 August 2026); Cognition and Lovable added. Anthropic updated from $47B (May) to $65B+ (end of July) and its profit status from "projected" to "positive adjusted operating income reported for Q2"; the $559M figure is now described as the May projection it was. OpenAI updated from "est. $24-33B" to the September run-rate figures of ~$50B and ~$70B with their sources. Databricks $5.4B at $134B to $7B+ at $190B. Scale AI from an untraced $1B estimate to just under $1B of 2025 revenue stated by the company, valuation from $14B to over $29B. Mistral from an untraced $600M estimate to $400M+ stated by its CEO, valuation from ~$10B to over €21B. ElevenLabs from an untraced $500M estimate to $500M+ ARR stated by the company, valuation from $3.3B to $22B. Midjourney: unsourced $5B valuation removed. GitHub Copilot: estimate kept and dated to January 2026, with Microsoft's later disclosures noted. Removed the computed totals and the median, which added up figures that measure different things. Removed the remaining "$2B with 50 people" and "studio of 40" lines that contradicted our own August correction. Headline no longer carries revenue figures, which go out of date within weeks.
• 7 August 2026 (second correction): Cursor's revenue per employee was wrong. We carried "~50 employees, ~$40M per head". Anysphere reported over 300 staff in November 2025, which put the figure near $6.7M. We also removed Midjourney's per-employee calculation: its headcount estimates range from about 40 to 163, so any ratio built on them is invented precision. Midjourney's profitability is stated as reported by Forbes rather than inferred from its lack of funding.
• 7 August 2026: All ten rows re-verified. Added a profitability section, computed aggregates, this methodology, and a downloadable dataset.
• 4 August 2026: Corrected GitHub Copilot from the widely quoted "$2B+ ARR" to an estimated $0.9-1.1B, based on the disclosed 4.7M paid subscribers and independent analyst work.
• August 2026: Anthropic updated to a ~$47B run rate and $965B valuation; OpenAI's $122B round added.

Use this data

The table is free to reuse with attribution to AI Business (aibusiness.vc) and a link to this page. A machine-readable copy is at ai-revenue-leaderboard.csv: same figures, same date, with the source link, the measure and the confidence level for every row. Open-ended figures stay open: "above $65B" is a lower bound of 65 with no upper bound, and the revenue_bound column says whether a figure is a lower bound, an upper bound, an approximation, a range or two conflicting reports. Where a figure is someone's estimate built on another disclosure, as with GitHub Copilot, revenue_source_url points to the estimate and revenue_basis_source_url to the disclosure it was built on.

If you are citing it and want to check a figure before you publish, write to info [at] aibusiness.vc and we will point you to the source for that row.

Sources for this edition: Anthropic, PBC: Form S-1, SEC EDGAR, 26 Oct 2026; SpaceX Q3 2026 earnings release, SEC EDGAR, 20 Oct 2026; TechCrunch: Anthropic's annualized revenue surges to $65B; Bloomberg: Anthropic Q2 revenue over $11.5B; Reuters via The Star: Anthropic expects a second profitable quarter (FT); Axios: OpenAI's ARR nears $70B; Irish Times: OpenAI tells investors its run rate is $20B below the earlier indication; Reuters via The Star: OpenAI forecasts cash burn near $280B by 2030 (FT); Databricks: $7B revenue run rate; Forbes: inside Scale AI after the Meta deal; Scale AI: next phase of company evolution; Microsoft FY26 Q2 earnings and FY26 Q4 earnings; Reuters via The Star: Cognition raises $2B at $48B; Dealroom: Lovable passes $600M run rate; ElevenLabs: $500M ARR and $22B tender; The Decoder on Mistral's revenue (FT); Mistral: Series D; Forbes: Midjourney; Cursor: joining SpaceX; Forbes: Cursor above $4B annualized revenue; SpaceX Form 8-K on the Cursor acquisition; SpaceX Q2 2026 earnings release.