Anthropic confirms in-house silicon team to design custom hardware for Claude
Anthropic has confirmed it will build an in-house silicon team and design its own hardware for its Claude AI model, aiming to reduce reliance on Nvidia and improve performance through co-design. The effort is still in early hiring stages.
Anthropic has confirmed plans to create an internal silicon team to develop hardware specifically for its Claude AI models. The company says it wants to bring more chip expertise in-house and reduce its dependence on external suppliers.
Why the shift?
A major factor is the industry's heavy reliance on Nvidia, whose hardware powers many AI models, including Anthropic's. This dependency is seen as a strategic vulnerability, particularly in a market where compute capacity is in high demand and supply is limited. Anthropic also believes that designing chips and models together can lead to performance gains. The company says its teams will co-design new hardware and models side by side.
Competitive considerations
This is part of a broader pattern among AI companies. If OpenAI can benefit from vertical integration, other frontier model providers will likely seek the same advantage. Anthropic has previously co-designed certain hardware components with partners, but now aims to internalize more of that knowledge. This could provide an edge as developers and users begin exploring smaller, cheaper, or open-weight models on their own hardware.
What to expect next
Anthropic is still in the process of hiring key team members, so significant time is likely needed before the initiative yields results. The company acknowledges that neither it nor its users will see benefits immediately.
