Claude chatbot developer Anthropic is confronting profitability challenges even as its revenue soars. The company last month revised down its 2025 gross margin forecast to 40%, a reduction of 10 percentage points from its earlier optimistic projection, due to inference costs for running AI models on Alphabet and Amazon servers being 23% higher than anticipated. This data underscores the persistent difficulty AI companies face in reducing their dependence on cloud service providers and the associated cost pressures, even while pursuing rapid growth. Despite the disappointing gross margin, Anthropic's financial performance still demonstrates robust growth momentum. The company anticipates 2025 revenue will reach $4.5 billion, a nearly twelvefold increase from the $381 million recorded in 2024. This explosive growth is primarily fueled by the success of its Claude Code programming tool and Cowork office assistant in the enterprise market, with at least nine clients now spending over $100 million annually, including Microsoft, whose expenditure is projected to hit $500 million. In contrast, competitor OpenAI expects a 2025 gross margin of approximately 46%, which includes the inference costs for both paid and free ChatGPT users. While OpenAI's projected 2025 revenue exceeding $13 billion maintains its absolute scale advantage, Anthropic is steadily narrowing the gap—their current monthly revenues stand at $1.7 billion and $750 million, respectively. This comparison highlights that operational efficiency and cost control have become decisive factors in the race for AI commercialization. These figures hold significant implications for investors. Despite Anthropic forecasting a 2025 EBITDA loss of around $5.2 billion and OpenAI reporting a pre-tax loss of $21.2 billion, equity investors remain optimistic. Anthropic is currently negotiating a funding round exceeding $10 billion, led by Singapore's GIC and Coatue Management, at a pre-money valuation of $350 billion; meanwhile, OpenAI is seeking to raise up to $100 billion at a valuation of approximately $750 billion. Anthropic's gross margin predicament stems from its reliance on cloud provider infrastructure. The company calculates gross margin by deducting inference costs and other sales costs, and the inference costs—fees for running AI models for paying customers on Alphabet and Amazon servers—are 23% higher than expected. If the inference costs for free Claude chatbot users were also included, Anthropic's gross margin would drop further to about 38%. While this performance remains far below that of traditional software companies, it represents a marked improvement from the negative 94% gross margin in 2024. Anthropic had previously projected gross margins would surpass 70% by 2027, and OpenAI also anticipates achieving at least 70% gross margins by 2029, nearing the levels of publicly traded software and cloud computing firms. However, beyond inference costs, both companies bear enormous model training expenses—Anthropic expects 2025 training costs to be around $4.1 billion, up approximately 5% from summer forecasts; OpenAI's spending on model training compute last year is estimated at $9.4 billion. These training costs are not factored into gross margin calculations but significantly increase the difficulty of achieving net profitability. To control expenses, both companies are taking steps to gain more control over hardware. Anthropic recently agreed to purchase $21 billion worth of Alphabet's Tensor Processing Units to reduce computing costs. OpenAI is developing its own server chip for inference, primarily for running ChatGPT, as an alternative to expensive Nvidia chips. OpenAI's CFO, Sarah Friar, stated that this inference chip has "taped out," meaning the final design has been sent to the chip manufacturer. Anthropic's revenue growth is primarily driven by enterprise clients. An estimated 86% of its 2025 revenue is expected to come from selling AI models to businesses via API, with the remainder from Claude chatbot subscriptions. The recent success of Claude Code and Cowork in coding and office scenarios has been likened by industry insiders to the frenzy OpenAI ignited in early 2023. Microsoft's payments to Anthropic for GitHub CoPilot are projected to reach $500 million, and popular coding tools like Cursor and Cognition are also major customers. OpenAI's enterprise business scale likely remains larger. Approximately 40% of OpenAI's revenue comes from enterprise clients, including API sales and chatbot sales to businesses, implying enterprise revenue of around $5.2 billion, compared to Anthropic's $3.9 billion. OpenAI's ChatGPT boasts about 900 million weekly active users, with roughly 95% being free users; the company recently announced it will introduce advertising to subsidize the costs of serving these free users. Although both companies are growing at a pace almost unprecedented in sales history, some lenders remain cautious about providing loans for data center projects involving them, as they are not expected to generate free cash flow until the latter part of this decade. However, this caution has not dampened the enthusiasm of equity investors. Nvidia and Microsoft have previously committed to invest up to $10 billion and $5 billion, respectively, in Anthropic, and investors valued the company at a pre-money valuation of $170 billion during a $13 billion funding round last September.
Comments