Anthropic explores building its own AI chips amid supply crunch

Early-stage discussions signal a potential shift in strategy as the company evaluates long-term hardware independence, balancing rising AI workloads with the high costs and complexities of building advanced semiconductor capabilities.

Amid a global shortage of advanced semiconductors, Anthropic is reportedly evaluating the possibility of developing its own artificial intelligence chips. The discussions are still at an early stage, and the company has not yet committed to building proprietary hardware, according to reports citing people familiar with the matter.

As per reports, sources indicate that Anthropic is currently assessing whether designing chips in-house would provide long-term advantages or if continuing to rely on external suppliers remains a more viable approach. No formal team has been assembled for the project, and the company has yet to finalise any technical roadmap.

Growing demand driving strategy

The exploration comes at a time when demand for Anthropic’s AI offerings is rising rapidly. Its flagship model, Claude, has seen significant adoption in 2026, contributing to a sharp increase in the company’s annualised revenue. This growth has intensified the need for high-performance computing resources capable of supporting large-scale AI training and deployment.

Currently, Anthropic depends on a mix of hardware solutions, including tensor processing units developed by Google and custom chips from Amazon. These partnerships enable the company to build and operate its AI systems, but also tie it to external supply chains that are facing increasing pressure.

In a related development, Anthropic recently strengthened its ties with Google and Broadcom through a long-term agreement focused on chip design and infrastructure expansion. The collaboration is part of a broader commitment to invest heavily in enhancing computing capacity within the United States.

Industry-wide shift toward custom chips

Anthropic’s internal discussions reflect a wider trend across the technology sector, where leading companies are considering custom silicon to optimise performance and reduce dependency on third-party suppliers. Firms such as Meta and OpenAI are also exploring similar strategies to meet the growing demands of AI workloads.

However, building advanced AI chips is a complex and costly undertaking. Industry estimates suggest that developing a single high-performance chip design can require investments of hundreds of millions of dollars, along with specialised engineering expertise and rigorous manufacturing processes.

While Anthropic has not yet made a final decision, its exploration signals the increasing importance of hardware innovation in the evolving AI landscape, where access to computing power is becoming as critical as advancements in software.