Analysis
Anthropic confirmed this week that it's building an in-house team to design custom AI chips for its Claude models, offering salaries up to $485,000 to recruit senior chip engineers, according to TechCrunch.
A Multi-Chip Strategy, Not a Replacement
The company frames this as the latest step in a deliberately diversified hardware strategy rather than a move away from its existing partners -- Anthropic says it will keep relying on chips from Nvidia, AWS, Google and AMD alongside whatever custom silicon its new team designs, co-designing hardware and models together to run more efficiently rather than chasing full vertical independence.
Following OpenAI and Google's Playbook
The move follows a path already set by larger competitors: OpenAI has been developing custom silicon with Broadcom, and Google has run its own in-house TPU program for years specifically to reduce dependence on Nvidia GPUs and manage the enormous cost of AI infrastructure at scale. The Information had previously reported Anthropic was scouting Samsung as a potential manufacturing partner, suggesting the custom-chip effort has been in motion well before this public confirmation.
What to watch: how quickly Anthropic's chip team can ship anything resembling a production design given custom silicon typically takes several years from team formation to deployed hardware, and whether Samsung or another foundry partner gets formally confirmed as the manufacturing relationship matures.