Anthropic Launches In-House Chip Development Amid Rising Compute Costs

On August 5, 2026, Anthropic, the AI company behind the Claude model series, officially revealed its plans to establish an in-house custom silicon team. This strategic move aims to design proprietary chips for training and inference workloads, aligning Anthropic with tech giants like Google, Apple, and Microsoft, who have long invested in custom silicon.

The motivation behind this initiative is clear: Anthropic reportedly spends around $19 billion on compute in 2026, a staggering figure that has raised concerns among investors. By developing its own chips, Anthropic hopes to reduce inference costs to levels that match or surpass those of its closest competitors.

Building advanced AI chips is estimated to cost around $500 million, a significant investment but increasingly necessary for leading AI labs striving for cost efficiency. To spearhead this effort, Anthropic has recruited Clive Chan, a former OpenAI engineer, who joined the company in June 2026.

Anthropic's recruitment strategy is aggressive, with engineering roles in the silicon program offering salaries between $320,000 and $485,000. Additionally, the company has initiated exploratory discussions with Samsung Electronics to potentially collaborate on chip manufacturing, with talks reportedly beginning in July 2026.

Importantly, Anthropic is not abandoning its existing hardware partnerships. The company is pursuing a multi-chip strategy, continuing to collaborate with Nvidia, AMD, AWS Trainium, and Google TPUs while developing its proprietary silicon. This approach allows Anthropic to leverage Google's expanding TPU capacity, which is set to increase by approximately 3.5 gigawatts starting in 2027.

The implications of Anthropic's move extend beyond its own operations. By potentially partnering with Samsung, Anthropic could enhance its negotiating leverage and supply chain resilience in a market where TSMC dominates leading-edge AI chip production. This strategic diversification is crucial as the company aims to manage its substantial compute demands more effectively.