Anthropic tells IPO investors its 15-person chip team designs custom silicon with Claude
What's happening? Anthropic has told investors in pre-IPO meetings that its new in-house chip design team, about 15 people, is already using Claude for engineering and design work on the company's custom chips, The Information reports, citing four people familiar with the discussions. According to those accounts, Anthropic's engineers say Claude cuts design time compared with their estimates of traditional workflows at Nvidia and AMD.
What is confirmed? Anthropic announced the chip team in August and has previously held talks with Samsung Electronics as a possible manufacturing partner. On a chip-design benchmark released last month by ChipAgents and Nvidia, Claude Fable 5.1 posted the top score, but that score was only 16.6%, so fully automated chip design is still far off.
What is reported but unconfirmed? A startup executive told The Information that Anthropic has approached some electronic design automation (EDA) companies about buying or using their chip design data. Anthropic has not disclosed any plan to sell AI chip-design tools, and the claimed time savings have not been quantified.
The race: Rival OpenAI says it built its Jalapeño chip with Broadcom in nine months using its own frontier models, has a multiyear deal with Synopsys for a model tuned to Synopsys tools, and its CFO Sarah Friar has named chip design as a target sector.
Why it matters: Frontier labs are pointing their models at the hardware that runs them. For Anthropic, whose backers Amazon, Google, Nvidia and AMD have committed about $90B, a custom-silicon push could sit awkwardly with those compute partners, and a move into design data would put it closer to Cadence and Synopsys. It also lands weeks before an IPO that could value the company around $1.5T.
Read more: The Information
