Anthropic confirmed August 5 it is building an in-house silicon team to design custom AI chips for Claude, marking the first public acknowledgment of the effort after months of reporting. The company is hiring senior chip engineers in a newly created Custom Silicon Team with a salary range of $320,000–$485,000. Job listings seek candidates with direct experience having "shipped silicon" and ability to make "consequential calls without a large organization behind them"—signaling a lean, senior-focused program targeting chip tape-out and bring-up rather than evaluation-only work.
The move targets co-design of hardware and Claude models to run "faster and more efficiently at the scale our customers need," with potential 50% cost savings on per-token inference (benchmarked against OpenAI's Jalapeño chip). Anthropic maintains a deliberate "multi-chip strategy," affirming that AWS, Google, NVIDIA, and AMD hardware will remain central to its scaling efforts. The company has explored Samsung as a potential manufacturing partner (reported in July) but has not disclosed process node, timeline, or manufacturing commitments.
Developing an advanced AI chip carries an estimated $500 million cost before fabrication, driven by specialized engineering talent and defect-free production. The initiative positions Anthropic alongside OpenAI (Jalapeño, June 2026), Meta (next-gen chip in September), and others pursuing vertical hardware integration. For practitioners, this signals that frontier AI labs now see custom silicon as essential to cost and margin control—not optional. Anthropic's run-rate revenue surpassed $30 billion in April 2026, and each customer's growing usage patterns are forcing the vertical stack question: third-party dependency vs. in-house optimization.