Anthropic Revenue Just Passed OpenAI. The Growth Rate Is the Real Story.

Anthropic hit $30 billion in annualized revenue on April 7, 2026, officially passing OpenAI's $25 billion. If you read that sentence and thought "okay, one AI company passed another," you missed the part that actually matters.

The gap between $30 billion and $25 billion is interesting. The fact that Anthropic's revenue went from $1 billion to $30 billion in fifteen months is extraordinary. For context: OpenAI went from roughly $20 billion to $25 billion in the same window. Anthropic grew 30 times over. OpenAI grew about 25 percent.

That asymmetry is not a blip. It reflects two fundamentally different bets on how enterprise AI is going to work, and right now, one bet is winning by a large margin. It is also prompting OpenAI to do something unusual: draft internal memos questioning the other company's accounting.

What Happened: $30B, 15 Months, and 1,000 Clients Spending $1M Each

The Anthropic revenue trajectory tells a progression that looks like a typo.

In January 2025, Anthropic's annualized revenue run rate was $1 billion. By June 2025, it was $4 billion. By December 2025, $9 billion. By April 7, 2026, $30 billion, up from $9 billion in roughly four months. According to reporting on the milestone, the company simultaneously announced that more than 1,000 enterprise customers now spend over $1 million per year with Anthropic. In February, when Anthropic closed its Series G at a $380 billion valuation, that number was 500. It doubled in under two months.

OpenAI's trajectory over the same 15 months went from $20 billion in annualized revenue at the end of 2025 to $25 billion by February 2026. That is roughly 25 percent growth. Both companies are growing. The rate differential is about 40 to 1.

Claude Code is the accelerant. Launched publicly in mid-2025, Anthropic's coding agent reached $1 billion in annualized revenue within six months of launch. Since the start of 2026, business subscriptions to Claude Code have quadrupled, and enterprise usage now accounts for more than half of all Claude Code revenue. Every enterprise development team that deploys Claude Code becomes a line item in Anthropic's ARR, with very high retention, because switching a coding workflow is not easy.

There is, however, a dispute worth noting. On April 13, OpenAI's chief revenue officer Denise Dresser sent a four-page internal memo to staff arguing that Anthropic's $30 billion figure is overstated by roughly $8 billion. The accounting question is whether Anthropic should report revenues from AWS and Google Cloud at gross value (the full amount billed through the partner) or net (after the partner's cut). Anthropic says it is the principal in these transactions, meaning it sets the price and the cloud providers are distribution channels, so gross revenue recognition is correct. OpenAI argues the comparable figure should be around $22 billion, which would put Anthropic behind OpenAI's $25 billion.

Both positions have defensible accounting logic. The dispute will matter more when both companies file S-1s.

Why Anthropic's Revenue Matters More Than the Number

The accounting question is real, but it distracts from a more durable fact: Anthropic's business is structured differently, and that structure compounds.

Anthropic never built a ChatGPT. It built the AI behind the tools enterprises actually deploy. Roughly 80 percent of Anthropic's revenue comes from enterprise and developer workloads. At OpenAI, enterprise accounts for approximately 40 percent, with consumer subscriptions (ChatGPT Plus, Pro, and Team) making up the other 60. Both revenue streams are large. But they behave differently.

Enterprise revenue has higher retention rates, more predictable expansion economics, and lower churn. Once a company builds a product on Claude's API, switching is expensive: you need to retest every workflow, retrain internal users, and renegotiate vendor contracts. Consumer subscriptions cancel in an afternoon. The composition of Anthropic's book of business means its $30 billion is, in quality terms, likely worth more per dollar than OpenAI's $25 billion, accounting dispute aside.

Enterprise LLM market share data makes the point more bluntly. According to third-party estimates, Anthropic now accounts for roughly 40 percent of enterprise LLM spending, compared to OpenAI at 27 percent and Google at 21 percent. A year ago, those numbers were roughly reversed.

The mechanism behind this shift is what is sometimes called bottom-up enterprise adoption. Individual developers discovered Claude Code, often paying out of pocket or on a personal subscription, and gradually pushed it into team and company-level usage. When enough developers in a company use Claude Code, an engineering manager notices the productivity numbers, and then procurement gets a call. The $1 million-per-year contract is often the end state of a process that started with one developer paying $20 a month. This is how Slack beat email. It is how GitHub Copilot got into enterprises before Microsoft made it a corporate offering. Anthropic appears to be running the same playbook, except the product is stickier because the workflow integration is deeper.

As SaaStr noted, Anthropic is also doing this while spending approximately 4 times less per dollar of revenue on model training than OpenAI is projected to spend. OpenAI's training budget is estimated to reach $125 billion per year by 2030. Anthropic's comparable figure for the same period is projected at around $30 billion. The revenue gap is widening while the cost structure diverges in the same direction.

The historical analogy that keeps coming up is Salesforce displacing Siebel in the early 2000s. Siebel had the CRM market, the brand, and the enterprise relationships. Salesforce had a simpler product and a business model that did not require a six-month procurement cycle. Fifteen years later, Siebel is a footnote. The parallel is imperfect, as OpenAI is not standing still and the AI market is moving far faster than enterprise software in 2003. But the structural logic is similar: a focused, enterprise-first challenger can compound faster than a market leader managing both a consumer product and an enterprise push simultaneously.

The Accounting Fight: Why OpenAI's Math Does Not Change the Trend

OpenAI's memo deserves a serious reading. The gross-versus-net revenue question is not trivial. Under Anthropic's accounting, a dollar that flows through AWS and ends up split between Anthropic and Amazon is counted in full on Anthropic's top line. Under OpenAI's preferred framing, only Anthropic's actual take would count. The $8 billion difference between $30 billion and $22 billion is not rounding error.

But here is what the accounting dispute does not explain.

OpenAI's memo may have closed the accounting gap by $8 billion. It did nothing about the growth rate gap. Even using OpenAI's preferred figure of $22 billion for Anthropic, which would technically place Anthropic behind OpenAI's $25 billion, Anthropic got to $22 billion from $1 billion in fifteen months. OpenAI went from $20 billion to $25 billion in the same window. The rate differential remains approximately 15 to 1.

The memo itself is telling. When a company that is losing ground in a metric decides the right response is to redefine the metric rather than close the gap, that is usually a signal about which direction things are moving. OpenAI is not wrong that gross and net revenue recognition produce different numbers. What the OpenAI accounting dispute cannot address is why 1,000 enterprises are now spending over a million dollars a year with Anthropic when that number was 500 two months ago.

CEO Sam Altman's response was to call Anthropic's safety positioning "fear-based marketing" in a podcast appearance. The characterization is worth noting: it is an argument about brand narrative, not about product performance. The developers who migrated from GitHub Copilot to Claude Code did not do so because Anthropic's safety messaging resonated with them. They did it because the code completions were better.

There is also an IPO dynamic at play. Both companies are in pre-IPO mode. Anthropic has reportedly held conversations with Goldman Sachs, JPMorgan, and Morgan Stanley about a potential October 2026 listing. Secondary market interest has reportedly reached valuations above $800 billion, which Anthropic has declined. OpenAI is valued at roughly $852 billion in private markets. When both companies eventually file S-1 documents, investors will apply a single, consistent accounting methodology, and the real comparables will emerge. Until then, both sets of numbers are promotional documents as much as financial ones.

What will matter for the IPO narrative is not the April 2026 snapshot but the trajectory. Analysts projecting Anthropic's anthropic revenue forward based on its recent growth rate are landing at figures between $80 billion and $100 billion by the end of 2026. That range assumes some deceleration from current pace, which is reasonable. But even half of current growth rates would put Anthropic in a different category than any enterprise software company that has ever gone public.

How OpenAI and Anthropic See the Market Differently

The revenue comparison is a symptom of a deeper strategic divergence that started at the company formation level and has never really converged.

OpenAI built ChatGPT first and figured out enterprise later. The consumer side of the business, ChatGPT Plus, Pro, and Team subscriptions, generates revenue, brand recognition, and a user base that feeds training data. But consumer AI support is expensive, retention is volatile, and the margins are narrower than API contracts. OpenAI is actively building toward enterprise parity: the company says enterprise revenue has grown from roughly 20 percent of revenue in 2024 to 40 percent today, and it is targeting parity with consumer by end of 2026.

Anthropic never had a consumer phase to speak of. The Claude chatbot exists, but it was never positioned as a ChatGPT competitor. It was positioned as a developer tool and a safe enterprise model. That means Anthropic carries almost no consumer support infrastructure. It does not have to maintain 15 million plus consumer subscriptions while simultaneously trying to close enterprise deals. The operating leverage from this structure is real.

The API-first approach also created a developer community that functions as a voluntary sales force. Developers who like Claude recommend it to their managers. Managers approve procurement. The flywheel runs without a dedicated enterprise sales team for every deal.

Smaller players, Mistral and Cohere, occupy the edges of the enterprise LLM market at under 10 percent share combined. Neither has the model quality or the infrastructure relationships to change the head-to-head between Anthropic and OpenAI in the near term.

According to Anthropic's Series G announcement, the company has expanded its compute partnership with Google and Broadcom to bring roughly 3.5 gigawatts of Google TPU capacity online in 2027. That infrastructure commitment is the other side of Google's $40 billion investment. It is not charity; it is Google Cloud securing the highest-growth enterprise AI customer it has.

What Comes Next for Both Companies

The short-term picture is framed by IPOs.

Anthropic is reportedly targeting an October 2026 listing, potentially raising more than $60 billion at a valuation that secondary markets have already priced above $800 billion. The $30 billion ARR figure, or the $22 billion figure if OpenAI's accounting framing sticks, will be the centerpiece of that S-1. Either way, the growth rate will be the headline.

OpenAI faces a more complicated narrative. It is the company that created the modern AI industry, holds the dominant consumer brand, and carries a private valuation of roughly $852 billion. But it is now second in annualized revenue to a company it spun out of, and the competitor has a higher growth rate, a higher enterprise concentration, and two of the three largest cloud providers as both investors and infrastructure partners. OpenAI's IPO will need to tell a story about why a slower-growing, consumer-heavy business deserves a premium to a faster-growing, enterprise-heavy one.

For the broader AI market, the more consequential question is what happens when Anthropic's compute infrastructure expansion comes online. With 3.5 gigawatts of TPU capacity coming in 2027, Anthropic will be able to run more experiments, train larger models faster, and serve more enterprise workloads simultaneously. If the Claude Code flywheel continues, with developers driving enterprise adoption, enterprise adoption funding infrastructure, and infrastructure enabling better models, the anthropic revenue trajectory that looks extraordinary today could look conservative in twelve months.

The accounting dispute will settle itself in the S-1 process. The growth rate will not. That is the number to watch.

FAQ: Common Questions About Anthropic Revenue and the OpenAI Comparison

Did Anthropic really pass OpenAI in revenue?

Anthropic reported $30 billion in annualized revenue (ARR) on April 7, 2026, compared to OpenAI's $25 billion disclosed in February 2026. OpenAI disputes the comparison, arguing Anthropic's gross revenue accounting overstates its figure by approximately $8 billion. Using OpenAI's preferred net calculation, Anthropic's comparable figure would be around $22 billion, which would place it below OpenAI. Both companies use different accounting treatments for cloud partner revenue, and the dispute will likely be resolved when both file IPO prospectuses.

Why is Anthropic growing so much faster than OpenAI?

The core difference is business model composition. Approximately 80 percent of Anthropic's revenue comes from enterprise API usage and developer contracts, while OpenAI is roughly 60 percent consumer subscriptions (ChatGPT Plus, Pro, Team) and 40 percent enterprise. Enterprise contracts have higher retention, lower churn, and expand organically through usage growth. Claude Code, Anthropic's coding agent, has been particularly effective at driving bottom-up enterprise adoption: developers adopt it individually, prove productivity gains, and then enterprises sign company-wide agreements.

What is the accounting controversy about?

OpenAI's CRO Denise Dresser alleged in an April 13 internal memo that Anthropic reports gross revenue from AWS and Google Cloud partnerships, counting the full transaction value even though Amazon and Google take a distribution cut. OpenAI reports its Microsoft partnership revenues on a net basis. Anthropic says it is the principal in these transactions, which under standard accounting rules justifies gross revenue recognition. The $8 billion difference represents the gap between the two methodologies.

What does this mean for their respective IPOs?

Both companies are preparing for IPOs in late 2026. Anthropic is reportedly targeting an October 2026 listing with a potential raise of $60 billion or more. OpenAI is valued at approximately $852 billion in private markets, while Anthropic's current valuation is around $380 billion, with secondary market offers reportedly reaching $800 billion. The revenue comparison and accounting dispute will receive heavy scrutiny from underwriters and public market investors when S-1 filings become available.

As AI coding tools like Claude Code absorb more of the software development workflow, engineering teams are generating unprecedented volumes of technical context: decisions, prompts, outputs, debugging sessions that rarely get captured or reused. If you are thinking about how to build a knowledge base for engineering teams that keeps pace with AI-assisted development, the infrastructure question is changing faster than most knowledge management systems were built to handle. The companies that figure out how to retain and reuse accumulated AI context will compound their productivity the same way Anthropic is compounding its anthropic revenue: by building a flywheel, not a pipeline.