Anthropic to Design Custom AI Chips for Claude Models

Anthropic has announced the creation of an in-house hardware team tasked with designing custom AI chips to overcome the global hardware supply crisis and boost the processing power of its Claude AI models. Through this new structure, which involves the co-development of software and hardware architecture, the company aims to meet its high-scale processing needs more efficiently.

Multi-Chip Strategy and Hardware Partnerships

While establishing its own chip design team, Anthropic emphasized that it will not terminate its existing hardware partnerships, noting that the move is part of a "multi-chip strategy." The company will continue to utilize the hardware and cloud infrastructures of industry giants such as Amazon Web Services (AWS), Google, Nvidia, and AMD. By recruiting expert engineers in both software and hardware, Anthropic aims to pursue a hybrid infrastructure model.

Production Timeline and Industry Costs

Anthropic has yet to make an official statement regarding the market release schedule or the specifics of the manufacturing process for the new chips. Industry experts point out that the cost of designing an advanced AI processor from scratch can reach up to $500 million. Investments in skilled hardware engineering and production risks during the foundry phase constitute the most critical cost items in this process.

Industry Implications

It is becoming an increasingly widespread trend in the industry for companies that train and operate AI models to design their own hardware to reduce chip dependency. Anthropic's move is regarded as a critical threshold in optimizing the hardware costs of large language models and overcoming bottlenecks in the supply chain.

*This news is based on data published by Webtekno — Artificial Intelligence.