Anthropic Just Passed OpenAI in Revenue. While Spending 4x Less to Train Their Models

Both OpenAI and Anthropic dropped major revenue disclosures within days of each other this week. OpenAI confirmed $2 billion in monthly revenue alongside a $122 billion raise at an $852 billion valuation. Anthropic confirmed $30 billion in annualized run-rate, up from $9 billion at year-end 2025. Then the WSJ published confidential financials from both companies ahead of their IPOs — and the training cost numbers make the revenue story even more interesting.

Put it all side by side and a few things become very clear.

1. Anthropic Just Passed OpenAI on Run-Rate Revenue

This was supposed to happen even under the most optimistic analyst assessments in August 2026. Epoch AI modeled it. The analysts debated the timing. It happened in April.

Anthropic is now at $30 billion annualized run-rate. OpenAI is at $24 billion ($2B per month, confirmed by the company). A year ago, Anthropic was at roughly $1 billion ARR and OpenAI was at $6 billion. The gap looked insurmountable.

It wasn’t.

Anthropic got there by doing almost the opposite of OpenAI. No viral consumer app. No 900 million weekly users. Instead: enterprise API contracts, developer adoption, and Claude Code. The company that most people outside of B2B circles couldn’t name two years ago is now generating more run-rate revenue than the company that invented the consumer AI category.

The lesson for B2B founders is straightforward: consumer scale and revenue scale are not the same thing. Anthropic has roughly 5% of ChatGPT’s consumer user base. It just passed them on top-line run-rate.

2. The Growth Rates Are Genuinely Without Precedent

OpenAI CFO Sarah Friar said it plainly: “never-before-seen growth at such scale.”

She’s right. OpenAI went from $2 billion ARR in 2023 to $6 billion in 2024 to $20 billion by end of 2025. Now $24 billion run-rate in April 2026. That is 3x per year, sustained, at a scale where 3x means adding billions of dollars every quarter.

Anthropic’s trajectory is even steeper. $87 million run-rate in January 2024. $1 billion by December 2024. $9 billion by end of 2025. $14 billion in February 2026. $19 billion in March. $30 billion in April.

That last sequence — $14B to $30B in roughly 8 weeks — is hard to make sense of in traditional software terms. Meritech’s Alex Clayton has said he reviewed the IPO trajectories of over 200 public software companies and never saw a growth rate like this. He said that in 2025. It has only accelerated since.

For context: Salesforce took about 20 years to reach $30 billion in annual revenue. Anthropic did it in under 3 years from a standing start.

3. Enterprise Is the Real Engine — for Both of Them

OpenAI announced that enterprise now makes up more than 40% of revenue, up from around 30% last year, and is on track to reach parity with consumer by end of 2026. APIs process more than 15 billion tokens per minute. Nine million paying business users as of February.

Anthropic never really had a consumer phase. Enterprise API contracts and cloud provider deals — primarily Google Cloud and AWS — built the base. Eight of the Fortune 10 are now Claude customers. Over 500 companies spend more than $1 million annually.

The convergence is notable: the company that started consumer-first is rapidly becoming enterprise-first. The company that was enterprise-first from day one is pulling ahead on run-rate as a result.

This matters for how you think about building in AI. Consumer virality gets you to a big user number fast. Enterprise contracts get you to durable, high-ACV revenue that compounds. The trajectory of both companies in 2026 is validation that the B2B motion — slower to start, harder to crack — is where the money actually is.

4. Coding Tools Are the Category That Changed Everything

Claude Code launched publicly in May 2025. By February 2026, it was at $2.5 billion annualized run-rate. That figure had more than doubled since January. Business subscriptions quadrupled in the same period. According to recent data, Claude Code now authors 4% of all public GitHub commits, with projections of 20%+ by year-end.

A product that did not exist 11 months ago is now generating more revenue than most public SaaS companies ever will.

OpenAI is seeing the same dynamic with Codex. Two million weekly users as of this week, up 5x in three months, growing 70% month over month. That is OpenAI’s answer to the coding agent category — and it is growing faster than most companies’ entire product lines.

The implication for anyone building developer tools or adjacent infrastructure is significant. The AI coding category went from zero to a multi-billion dollar market in under a year. That is not a feature. That is a new layer of the software stack, and the companies that own it are going to own an enormous amount of enterprise spend.

5. Neither Company Is Profitable. And That Is A Big Part of the Bet

OpenAI is burning approximately $17 billion in cash this year. Internal documents project a $14 billion loss for 2026. The company has committed over $1 trillion to infrastructure over the next several years and does not project positive free cash flow until 2029.

Anthropic’s burn is proportionally large as well. It has raised over $18 billion in funding. The $30 billion run-rate is real, but so is the cost structure.

The investors funding both companies — SoftBank, Amazon, Nvidia, Google, a16z, Lightspeed, ICONIQ — are not confused about this. They are making a specific bet: that compute costs continue to fall per unit of intelligence, that revenue keeps compounding faster than burn, and that whoever owns the AI infrastructure layer in 2029 will generate returns that make the interim losses irrelevant.

That bet might be right. It might not be. But the numbers are large enough now that it is no longer a venture bet — it is a macroeconomic one. Amazon alone committed $50 billion to this round. That is not a financial bet on a startup. That is a strategic bet on where enterprise compute is going.

For everyone building in B2B software right now: the companies you are potentially competing with, building on top of, or selling into are operating at a scale and a loss tolerance that has no historical parallel. That is either the best time in history to be a founder building AI-native software — because the infrastructure is being subsidized at massive scale — or it is a warning sign that the economics of this market are going to be very difficult for anyone who is not one of the two or three companies at the top.

Probably both.

6. The Training Cost Gap Is the Most Underreported Number in AI Right Now

The WSJ just published confidential financial documents from both companies ahead of their IPOs. The core finding should reframe everything above.

OpenAI projects spending $121 billion on compute in 2028 alone. That year, even after strong revenue growth, it projects losses of $85 billion. It does not expect to break even until after 2030. Anthropic’s training costs peak at around $30 billion in that same period — roughly 4x less — and the company projects reaching profitability in 2028 or 2029.

The company that just passed OpenAI on run-rate revenue is doing it while spending a fraction of what OpenAI spends on model training.

The standard assumption has been that whoever spends most on training wins. Anthropic is now running a meaningful test of whether that’s actually true.

There is an important caveat: the two companies count revenue differently. Anthropic books revenue through cloud partners (Google Cloud, AWS); OpenAI does not count that the same way. So the direct run-rate comparison has some noise in it. But even accounting for that, the capital efficiency gap is real.

Strip out training costs entirely and both companies are near operating profitability right now. Add training costs back in and the paths diverge sharply: OpenAI is making a 2030s bet on model supremacy justifying the spend; Anthropic is making a sooner-is-better bet on capital efficiency.