Anthropic Building Custom AI Chip Design Team

Anthropic has confirmed it is assembling a team to design custom AI chips, a strategic move aimed at running its Claude models faster and more efficiently as demand accelerates.

The company told TechCrunch it plans to co-design hardware and models together, an approach that could optimize performance beyond what off-the-shelf accelerators can deliver. Business Insider first reported the development.

According to a job listing, Anthropic is seeking engineers with chip design experience for its "custom silicon team." The Information reported last month that the company has been exploring Samsung as a potential manufacturing partner for these chips.

Why it matters

Custom chip development represents a fundamental shift in how leading AI companies approach infrastructure. While Anthropic currently relies on partnerships with AWS, Google, Nvidia, and AMD for computing hardware, building proprietary silicon signals that third-party access alone cannot meet the scale Claude requires. This vertical integration also gives AI companies more control over performance optimization, cost structure, and supply chain resilience—critical advantages as competition for GPU capacity intensifies across the industry.

Following industry precedent

Anthropic's move follows a pattern established by its competitors. OpenAI unveiled its Broadcom-manufactured Jalapeño chip in June, designed specifically for inference workloads—the process of running trained models to generate responses. Google DeepMind has long used Alphabet's custom TPU (Tensor Processing Unit) chips to power its AI systems. Meta has been developing its MTIA accelerators for similar purposes.

The trend reflects a broader reality: as AI models grow more complex and user bases expand, companies face a choice between competing for limited third-party hardware or investing in custom solutions tailored to their specific architectures.

Infrastructure constraints driving change

The push toward custom silicon comes as AI companies race to secure computing infrastructure amid persistent supply constraints. High-performance AI accelerators remain scarce, with long lead times and premium pricing. Designing purpose-built chips allows companies to optimize for their particular model architectures and workloads, potentially achieving better performance per watt and lower long-term costs than general-purpose alternatives.

For Anthropic, which has positioned Claude as a competitor to OpenAI's ChatGPT and Google's Gemini, controlling more of the hardware stack could provide both technical and economic advantages as the company scales.

Business Insider first reported the news, which Anthropic subsequently confirmed to TechCrunch.