Sources Say Anthropic Considers Building Custom Hardware
According to three independent sources, artificial intelligence company Anthropic is evaluating the possibility of designing its own chips to power its AI systems. The decision comes as competition intensifies in the AI sector, with companies seeking greater control over hardware that drives advanced models like Claude.
Early-Stage Discussions and Uncertainty
The initiative remains in its preliminary phases. While Anthropic has not finalized plans or assembled a dedicated team, insiders suggest the company may still opt to purchase chips rather than develop them internally. A spokesperson for the San Francisco-based firm declined to comment on the report.
Revenue Growth Accelerates Amid AI Demand
Anthropic’s revenue from its AI model Claude has surged in 2026, reaching over $30 billion in annual run-rate income—a significant jump from $9 billion at the end of 2025. The company attributes this growth to rising demand for its services and expanded applications of its technology.
Current Chip Strategy Relies on Partnerships
To date, Anthropic uses a mix of chips, including tensor processing units (TPUs) from Google and Amazon. The firm recently signed a long-term agreement with Google and Broadcom to secure TPUs, aligning with its $50 billion investment in U.S. computing infrastructure.
Industry Trends Mirror Larger Tech Players
Anthropic’s exploration of custom chips reflects broader industry trends. Companies like Meta and OpenAI are also pursuing independent chip design to reduce reliance on third-party suppliers. Industry experts estimate that developing a high-performance AI chip can cost up to $500 million, requiring significant engineering and manufacturing resources.
Challenges in Chip Development
The process involves not only hiring skilled engineers but also ensuring flawless manufacturing processes. While Anthropic has not announced a timeline or specific design goals, the move underscores its commitment to maintaining competitive advantage in an increasingly hardware-driven AI landscape.
