Anthropic confirms it's building custom AI chips to break free from Nvidia

Anthropic isn’t just building frontier models. It’s now officially building the silicon they run on.

The company confirmed to TechCrunch and Business Insider that it is assembling a “custom silicon team,” with job listings up for a senior silicon engineer and a technical program manager focused on semiconductor design. A spokesperson framed the move as part of a “multi-chip approach” — meaning Anthropic plans to keep using hardware from existing partners while pursuing its own designs in tandem. This puts meat on the bones of a rumor The Information reported earlier, which pointed to Samsung as a potential manufacturing partner.

If that partnership materializes, it would be a significant signal. Samsung’s foundry business trails TSMC but offers a viable path for companies needing advanced nodes without the brutal competition for TSMC’s limited capacity. For Anthropic, which has been vacuuming up billions in funding — including a reported $2.75 billion from Amazon alone — a custom chip investment is a logical hedge against the Nvidia tax that every AI lab is paying right now.

Anthropic is late to this party, but not embarrassingly so. OpenAI already announced Jalapeño, an inference chip developed with Broadcom, and has a broader plan to use TSMC for manufacturing. Google’s TPUs have been a strategic moat for years. Meta deployed its own inference silicon, and Mistral is reportedly kicking the tires on custom hardware too. The real question isn’t whether Anthropic should be doing this. It’s whether a company that just killed its latest model release on X with a single post can manage the multi-year slog of chip development without losing focus. That’s the tension nobody’s talking about.