Anthropic, the developer of the Claude family of neural networks, has faced a notable decline in operating profitability. The key factors putting pressure on margins were an aggressive reduction in tariffs for the use of Claude models and, at the same time, a sharp increase in spending on server infrastructure.
Lowering prices for access to the neural networks is a strategic step aimed at expanding the user base and strengthening competitive positions in the generative AI market. However, such a pricing policy comes into direct conflict with the economics of computing: scaling inference load requires significant computing power, and server costs are growing at a faster pace.
The economics of computing under pressure
The company has found itself in a trap classic for the generative AI sector: the more users turn to the models, the higher the load on infrastructure and the more substantial the costs of renting and maintaining server capacity. Lowering tariffs while costs rise creates negative operating leverage, in which each additional unit of demand does not improve but worsens the financial result.
The situation reflects the fundamental challenge facing the entire AI services industry: the balance between product accessibility for the mass market and the need to maintain sustainable unit economics. Anthropic, competing with the largest players in the sector, is forced to invest in price advantage, sacrificing short-term profitability for long-term market capture.
Cryptalist analysis: The Anthropic case illustrates a systemic contradiction in the generative AI market — price competition inevitably outpaces the decline in computing costs, which increases pressure on margins even for technologically strong players. For investors, this is a signal that the sustainability of AI companies' business models will be determined not only by the quality of models but also by the ability to optimize infrastructure costs and build differentiated pricing.