Anthropic Lowers Gross Margin Projection as Revenue Skyrockets
The Takeaway
- Anthropic late last year lowered 2025 gross profit margin projection to 40% due to higher inference costs
- Anthropic projected $4.5 billion revenue in 2025, a nearly 12x increase from 2024
- Anthropic, OpenAI are inking deals to get computing costs under control
Anthropic last month projected it would generate a 40% gross profit margin from selling AI to businesses and application developers in 2025, according to two people with knowledge of its financials. That margin was 10 percentage points lower than its earlier optimistic expectations, though it’s still a big improvement from the year before.
The lower-than-expected gross profit margin resulted from the costs of running Anthropic models for paying customers, in a process known as inference, on servers from Google and Amazon. Those inference costs were 23% higher than the company had anticipated, the projections showed. Anthropic calculates gross margins by subtracting inference costs and other costs of selling its products.
The expected 2025 result suggests Anthropic, maker of the Claude chatbot and coding agent, which has generated tremendous buzz in recent weeks, may be operating less efficiently than archrival OpenAI as measured by gross margin. Still, Anthropic’s gross margin has improved significantly from negative 94% in 2024.
The data show how both companies’ reliance on renting specialized servers from cloud providers makes it harder to generate a net profit, which is why they are taking steps to create or control server hardware themselves and are in the process of raising tens of billions of dollars to shore up their balance sheets. OpenAI just announced it would launch ads to subsidize nonpaying users of its chatbot.
If Anthropic also counted inference costs for Claude chatbot users that don’t pay for a subscription, its gross margin would be about 38%, or a few percentage points lower than for paid users, based on The Information’s analysis.
In contrast, OpenAI projected a gross margin of around 46% in 2025, including inference costs of both paying and nonpaying ChatGPT users. Nonpaying users make up roughly 95% of the chatbot’s roughly 900 million weekly active users.
Anthropic has previously projected gross margins above 70% by 2027, and OpenAI has projected gross margins of at least 70% by 2029, which would put them closer to the gross margins of publicly traded software and cloud firms. But both AI developers also spend a tremendous amount on renting servers to develop new models—training costs, which don’t factor into gross margins—making it more difficult to turn a net profit than it is for traditional software firms.
The inference costs are in addition to costs from training the models. Anthropic last month expected its costs for training its AI models for 2025 to be roughly $4.1 billion, up roughly 5% from its summer projections. OpenAI, meanwhile, expected to spend $9.4 billion on compute for training its AI models last year.
Both companies have taken steps to get those costs under control. Anthropic recently agreed to purchase $21 billion of Google’s tensor processing units to gain more control over the hardware it uses so it can reduce its computing costs. OpenAI, meanwhile, has been developing a server chip to power inference—mainly for running ChatGPT—that it plans to use in the coming years as an alternative to expensive Nvidia chips.
On Wednesday, OpenAI Chief Financial Officer Sarah Friar said that the upcoming inference chip was “taped out,” implying the company had shared the final design with its chip manufacturer.
Anthropic recently projected that its 2025 revenue would be $4.5 billion, marginally lower than its previous optimistic projection of nearly $4.7 billion. Still, its 2025 revenue is nearly 12 times higher than its 2024 revenue of $381 million, a remarkable growth rate for a company that size. (In comparison, OpenAI generated revenue of more than $13 billion in 2025, up from around $3.7 billion in 2024.)
The revenue gap between the companies has been shrinking, as OpenAI is currently generating a little more than double what Anthropic is generating per month: $1.7 billion compared to $750 million.
Coding Bonanza
Anthropic’s AI for coding and white-collar tasks—Claude Code and Cowork, respectively—has been the biggest AI success story of the past month, drawing parallels with the excitement OpenAI generated among businesses and software engineers in early 2023.
Anthropic privately has disclosed it has at least nine customers spending more than $100 million a year on its products, according to one of the people. Microsoft’s spending on Anthropic, for instance, was on track to hit $500 million, in part for Microsoft’s GitHub CoPilot. Other popular coding tools such as Cursor and Cognition also are large customers of Anthropic.
About 86% of Anthropic’s 2025 revenue was expected to come from its sale of AI models to such businesses through an application programming interface, the company estimated. The rest of its revenue comes from subscription revenue from its Claude chatbot, which competes with ChatGPT.
OpenAI, too, has been gaining business customers. In fact, its enterprise business may still be larger than Anthropic’s when including API and chatbot sales to businesses. OpenAI generates roughly 40% of revenue from business customers, according to a person familiar with its financials. That implies roughly $5.2 billion of OpenAI’s revenue is from business customers, as compared to Anthropic’s $3.9 billion.
While both companies’ sales are growing at a nearly unprecedented rate, some lenders are apprehensive about lending money to data center projects involving such firms because they don’t plan to generate free cash flow until near the end of the decade.
Their cash burn has left equity investors in the firms undaunted, however. Anthropic as of mid-December expected to lose about $5.2 billion, based on its earnings before interest, taxes, depreciation and amortization, last year, according to its most optimistic projections. OpenAI earlier projected a loss of $21.2 billion before interest and taxes, though its projected cash burn was less than half that amount.
Anthropic is in talks to raise more than $10 billion at a $350 billion valuation before the financing in a round led by Singapore’s GIC and Coatue Management. Nvidia and Microsoft previously committed to investing up to $10 billion and up to $5 billion, respectively, in Anthropic. Investors last valued Anthropic at $170 billion before the financing in a $13 billion September round.
OpenAI, meanwhile, is in talks to raise as much as $100 billion in a round that could value it around $750 billion before the investment.