AI hard to profit from? Anthropic lowers gross margin guidance: Though revenue increased 12-fold, it's hard to overcome a 23% surge in inference costs.

AI hard to profit from? Anthropic lowers gross margin guidance: Though revenue increased 12-fold, it's hard to overcome a 23% surge in inference costs.

Claude chatbot developer Anthropic is facing profitability challenges as its revenue surges. According to a report by The Information on Wednesday, the company lowered its 2025 gross margin forecast to 40% last month, 10 percentage points below previous optimistic projections. This was due to AI model inference costs on Google and Amazon servers being 23% higher than expected. This figure highlights that while AI companies pursue large-scale growth, they still struggle to reduce reliance on cloud service providers and contend with the resulting cost pressures. Despite gross margins falling short of expectations, Anthropic’s financial performance still shows strong growth momentum. Sources say the company expects revenue to reach $4.5 billion in 2025, nearly 12 times the $381 million expected in 2024. This rapid growth is mainly attributed to the success of its Claude Code coding tool and Cowork assistant in the enterprise market. Currently, at least nine customers spend over $100 million annually, with Microsoft’s spending expected to reach $500 million. In comparison, rival OpenAI expects its 2025 gross margin to be about 46%, including inference costs for both paid and free ChatGPT users. OpenAI’s 2025 revenue will exceed $13 billion. Though its absolute scale still leads, Anthropic is closing the gap—their current monthly revenues are $1.7 billion and $750 million, respectively. This comparison shows that in the race to commercialize AI, operational efficiency and cost control have become the key factors determining success. These figures are significant for investors. Although Anthropic expects an EBITDA loss of roughly $5.2 billion in 2025, and OpenAI's pre-tax loss may reach $21.2 billion, equity investors remain optimistic. Anthropic is in talks to raise over $10 billion at a pre-money valuation of $350 billion, led by Singapore’s GIC and Coatue Management; OpenAI is seeking to raise up to $100 billion at a valuation of about $750 billion. **Inference Costs Squeeze Profit Margins** Anthropic’s gross margin dilemma stems from its dependence on cloud service infrastructure. The company calculates gross margin by deducting inference and other sales costs, yet inference costs—charges for running paid AI models on Google and Amazon servers—are 23% higher than expected. If inference costs for free Claude chatbot users were also included, Anthropic’s gross margin would drop further to about 38%. Although this is far below traditional software companies, it is a marked improvement from the negative 94% gross margin in 2024. Anthropic previously projected it would surpass a 70% gross margin by 2027; OpenAI also expects to reach at least 70% gross margins by 2029, approaching levels seen in listed software and cloud computing companies. However, besides inference costs, both companies face enormous model training costs—Anthropic expects training expenses of about $4.1 billion in 2025, up around 5% from summer estimates; OpenAI’s training compute expenditure last year was estimated at $9.4 billion. These training costs are not included in gross margin calculations but further increase the difficulty of achieving net profitability. To control costs, both companies are moving to take charge of hardware. Anthropic recently agreed to purchase $21 billion worth of Google’s Tensor Processing Units to decrease compute costs. OpenAI is developing server chips mainly for running ChatGPT, as an alternative to expensive Nvidia chips. OpenAI’s CFO Sarah Friar said Wednesday that the inference chip has been “taped out,” meaning the company has delivered the final design to the chip manufacturer. **Enterprise Market Competition Intensifies** Anthropic’s revenue growth is mainly driven by enterprise clients. It’s estimated that about 86% of its 2025 revenue will come from selling AI models to enterprises via API, with the rest from Claude chatbot subscriptions. The recent success of Claude Code and Cowork in coding and office scenarios has been compared by insiders to the hype OpenAI triggered in early 2023. Microsoft is projected to pay Anthropic $500 million for GitHub CoPilot, while other hot coding tools like Cursor and Cognition are also major clients. OpenAI’s enterprise business likely remains larger. Sources say about 40% of OpenAI’s revenue comes from enterprise clients, including API and business-oriented chatbot sales, amounting to about $5.2 billion in enterprise revenue, compared to Anthropic’s $3.9 billion. OpenAI’s ChatGPT has about 900 million weekly active users, 95% of whom are free users; the company just announced plans to launch an ad business to subsidize free user costs. Despite the unprecedented surge in sales at both firms, some lenders remain cautious about providing loans for data center projects involving these companies, believing that free cash flow may not be generated until the end of the century. However, this has not dampened equity investor enthusiasm. Nvidia and Microsoft have previously pledged up to $10 billion and $5 billion in investments to Anthropic, respectively. In last September’s $13 billion funding round, investors valued Anthropic at $170 billion pre-money. **Risk Disclosure and Disclaimer** Markets involve risk; investment should be undertaken cautiously. This article does not constitute personal investment advice, nor does it consider individual users' specific investment objectives, financial situation, or needs. Users should evaluate whether any opinions, views, or conclusions in this article suit their particular circumstances. Investments based on this article are at the user’s own risk.