Anthropic designs custom hardware to power Claude
▼ Summary
– Anthropic is hiring engineers to design custom silicon chips for running its AI models, confirming earlier rumors.
– The company will use a multi-chip approach, combining its own hardware with chips from other vendors.
– Anthropic’s spokesperson confirmed the plans to Business Insider and TechCrunch.
– The move follows a previous report that Anthropic was considering Samsung as a manufacturing partner.
– Competitors like OpenAI, Google, Meta, and Mistral are also developing or deploying custom chips.
Anthropic is moving chip development in-house, announcing plans to assemble a dedicated team tasked with designing custom silicon to power its AI models.
The company posted a job opening for a senior engineer with a proven track record in semiconductor design, a listing first spotted by Business Insider. Additional roles, including a silicon engineer and a technical program manager for silicon, are now live on Anthropic’s job board. A company spokesperson subsequently confirmed the initiative to both Business Insider and TechCrunch.
While the shift signals a major step toward vertical integration, Anthropic will not go all-in on proprietary hardware just yet. The spokesperson emphasized a multi-chip strategy, meaning the company will continue to rely on processors from external vendors alongside its own custom designs as it scales its infrastructure.
This development puts an end to speculation that has been building for weeks. Earlier reports from The Information suggested Anthropic was exploring a manufacturing partnership with Samsung to bring its hardware ambitions to life.
Anthropic is hardly the first mover in this direction. OpenAI recently unveiled Jalapeño, a custom chip built specifically for large language model inference in data centers, developed in collaboration with Broadcom. Google has long relied on its own internal hardware to run models, while Meta has already designed and deployed its own silicon. Mistral, another major player in the AI space, is reportedly exploring similar options.
By building its own chips, Anthropic is positioning itself to optimize performance and reduce dependence on third-party suppliers, a critical consideration as AI workloads become increasingly compute-intensive. The move also reflects a broader industry trend where leading AI labs are seeking greater control over the hardware that underpins their models.
(Source: Ars Technica)
