Anthropic is using Claude to design its own AI chips and touting it to IPO investors, The Information reports

Anthropic's in-house chip team is using Claude to design the company's own AI chips, and Anthropic is pitching that work to IPO investors, The Information reported on November 3. The article is paywalled. We read its headline and public summary ourselves; the team-size and speed details below come from a summary of the full article posted on X, and we could not check them against the text.

WHAT THE REPORT SAYS

- Per The Information's public summary: after its models helped uncover serious cyber vulnerabilities and changed how software gets written, Anthropic is now seeing how much of an advantage they give it in designing its own AI chips, as it heads toward its IPO. - Per a summary of the full article posted on X by @kimmonismus: the chip design team is about 15 people, and it has found that Claude speeds up the work compared with its estimates of traditional chip design workflows. - The story is by Qianer Liu, Valida Pau and Phoebe Liu.

WHAT IS SOURCED ELSEWHERE

- August 5: Anthropic confirmed to TechCrunch that it is building a team to design custom chips, saying it plans to co-design hardware and models so its technology runs faster and more efficiently. The job listing calls it a "custom silicon team". Business Insider reported the team first. - Anthropic rents its compute from AWS, Google, Nvidia and AMD today, per the same TechCrunch story, which also notes that OpenAI unveiled its own Broadcom-built inference chip, Jalapeño, in June. - Mid-October: Reuters reported that investors were weighing a valuation of about $1.5 trillion for Anthropic's November share sale.

WHY IT MATTERS

A frontier lab using its own model to shorten the design of the chips that will run its next models would cut one of the slowest, most expensive steps in AI hardware. It is also an IPO pitch, so treat the speed claim as Anthropic's case to investors, not a measured result.

WHAT WE DO NOT KNOW

- How much faster, and against what: the comparison is against Anthropic's own estimates of a traditional workflow. - Whether any chip has taped out, who would manufacture it (The Information reported in July that Anthropic was scouting Samsung), and when one would run Claude. - Which Claude model the team uses.

Posts on X flagged the report. Anthropic has not commented on it publicly, as far as we can find.