Anthropic made its first public admission that it is building a dedicated silicon team to create custom chips for Claude, the company’s flagship large‑language model. The announcement came through a spokesperson who told Business Insider that the new group will co‑design hardware and models, enabling Claude to operate "faster and more efficiently at the scale our customers need."
Details emerged from a job listing posted by Anthropic that describes a "custom silicon team" and seeks engineers who have "shipped silicon." The posting lists a compensation range of $320,000 to $485,000 and emphasizes that candidates must have a "realistic relationship with schedules" and be able to make consequential decisions without a large organization behind them. The language suggests Anthropic is looking for veterans who have taken a chip from design to production.
Anthropic’s strategy does not abandon its existing multi‑chip approach. The company currently runs Claude on hardware from Amazon Trainium, Google TPUs, Nvidia GPUs and AMD accelerators. By adding its own silicon, Anthropic would gain a fifth option that it controls end‑to‑end, potentially reducing reliance on external suppliers and optimizing performance for its specific workloads.
The move follows months of industry chatter. Reuters reported in April that Anthropic was exploring custom chips as Claude’s run‑rate revenue topped $30 billion, though at the time the firm had no dedicated team. The Information later noted that Anthropic had discussed manufacturing with Samsung. The hiring of Clive Chan, a veteran who helped launch OpenAI’s chip program, further hinted that the company was moving from speculation to execution.
Anthropic joins a growing list of AI firms investing in proprietary hardware. OpenAI unveiled Jalapeño, a custom inference chip built with Broadcom, slated for deployment in late 2026. Meta announced its "Iris" chip will enter production in September, while French startup Mistral’s CEO has floated the idea of building its own silicon. Industry analysts estimate that designing an AI chip costs roughly $500 million, but economies of scale improve at the size of Anthropic’s operations.
By pursuing its own silicon, Anthropic hopes to improve both performance and cost efficiency. Custom chips can be tuned to the exact computational patterns of Claude, reducing wasted cycles and energy consumption. At the same time, owning the hardware stack gives the company greater flexibility in pricing and service-level agreements for enterprise customers.
Despite the ambitious push, Anthropic is not abandoning its partnerships. The company’s continued use of AWS, Google, Nvidia and AMD hardware indicates a pragmatic approach: retain proven, high‑throughput platforms while developing a proprietary alternative that can eventually complement or replace them. This hybrid model mirrors the strategies of rivals who also balance external cloud resources with in‑house silicon.
Investors and analysts will be watching closely to see how quickly Anthropic can move from design to silicon shipment. The high salary range in the job posting underscores the scarcity of talent capable of delivering a production‑ready chip on schedule. If Anthropic succeeds, it could set a new benchmark for AI startups seeking to control more of the compute stack, potentially reshaping the competitive dynamics in the fast‑moving generative‑AI market.
Este artigo foi escrito com a assistência de IA.
News Factory APP - notícias agênticas para impulsionar seu SEO e AEO.