Anthropic has reportedly begun early-stage development of its own custom artificial intelligence chip and has held preliminary discussions with Samsung Electronics to handle manufacturing. According to The Information, citing three sources familiar with the matter, the San Francisco-based startup is currently in the foundational stages of defining the processor's specifications, power requirements, and server cluster configurations. The company is exploring Samsung's advanced 2-nanometer manufacturing and packaging technology as it seeks greater control over its computing infrastructure and skyrocketing costs. While the San Francisco-based startup has engaged with several chip design firms, it has not yet progressed to formal blueprinting, testing, or production, as reported by The Information.
Anthropic PBC has hired Amir Salek, a founder of the custom chip programme at Alphabet Inc.'s Google, as the AI lab lays the groundwork for a push into making its own semiconductors. According to reports from Bloomberg, Salek is joining the compute team of Anthropic, maker of the Claude platform, with the executive running Google's tensor processing unit business until 2022 and delivering the first seven generations of those chips. In his new role, Salek will report directly to Anthropic compute lead James Bradbury, underscoring the strategic importance of this executive hire as the company builds its internal silicon capabilities. Salek's résumé represents one of the most direct paths in the industry from "built the chip" to "leads the next chip," having founded Google's TPU program in 2013 and built TPUv1-v4 and Edge TPU systems before his departure in 2022.
Anthropic, which buys chips from a variety of sources, including Nvidia Corp., Google and Amazon.com Inc., has indicated recently that it wants to build an in-house silicon business. As reported by Bloomberg, the San Francisco-based company has begun hiring and listing jobs for the effort as it scrambles to get enough data center infrastructure to support its ambitions. Establishing an internal silicon division is expected to help the company tailor hardware design to its operational requirements while mitigating ongoing supply chain bottlenecks. Custom chips would potentially help the company cope with supply shortages and better meet its specific AI computing needs. The hire signals that Anthropic is compressing the "how do we even start" phase, with Salek having already made every mistake a from-scratch custom-silicon program makes once at Google, on infrastructure with similar scale demands to what Anthropic now runs for Claude.
Like other AI giants, Anthropic is taking similar steps to Rival OpenAI, the company behind ChatGPT, which has unveiled a chip called Jalapeno, co-developed with Broadcom Inc., and plans to begin using it later this year. According to explainx.ai, Anthropic is pursuing a three-pronged compute strategy: rent capacity at scale (AMD, Google), buy efficiency software (Decart, if it closes), and now, build the chip itself. The company recently outlined an initial $250 million commitment to UK chip firm Fractile, while simultaneously securing additional data center capacity through deals with Riot Platforms Inc. and Volta Infra Holdings Ltd. Each chapter targets a different point on the build-vs-buy spectrum, with AMD and Google deals locking in someone else's silicon at scale, Decart (if it closes) buying software that makes existing chips run more efficiently, and Salek's hire representing the build-your-own-from-scratch option.
To strengthen its chip-design capabilities, Anthropic has begun recruiting specialized engineering talent, including the recent hire of Clive Chan, a key engineer from OpenAI's dedicated silicon division. As reported by The Information, this move aligns with a broader industry trend where top-tier AI developers are moving to command their entire backend supply chains, ranging from silicon and cloud partnerships to energy sources and physical data centers. Several hyperscalers are planning to develop their own chips to boost computing power and lower reliance on Nvidia's Blackwell chips or Google's Tensor chips, but surging memory costs due to bloated demand have created cost concerns. Anthropic's exploration of custom hardware aligns with a broader industry trend where top-tier AI developers are moving to command their entire backend supply chains, with the company noting that Amazon Web Services' Trainium chips, Google's Tensor Processing Units (TPUs), and Nvidia's graphics processors will continue to serve as core elements of its long-term computing strategy.
