Anthropic confirmed Wednesday (August 5) it is assembling an in-house silicon team to design custom chips for its Claude AI models, targeting co-optimization of hardware and software to run Claude faster and more efficiently at scale. The company is recruiting engineers across hardware and software with direct semiconductor shipping experience, offering salaries ranging from $320,000 to $485,000. No timeline for first-chip production has been disclosed, nor has Anthropic confirmed whether it will manufacture chips in-house or outsource fabrication.
The move mirrors OpenAI (Jalapeño via Broadcom), Meta (MTIA accelerators), and Google DeepMind (TPUs), signaling that leading AI labs view custom silicon as essential infrastructure amid persistent chip shortages. Industry estimates place the R&D cost of a leading-edge (2nm/3nm) AI chip at $500-750 million. Anthropic has held exploratory talks with Samsung Electronics about using its 2nm SF2P node, though no manufacturing agreement has been signed yet. Samsung became a Series H investor in May 2026, suggesting potential supply-chain leverage.
Critically, Anthropic is NOT abandoning existing suppliers—AWS, Google, Nvidia, and AMD will remain central to its multi-chip strategy. The company is architecting this as a hedge on long-term supply: it will continue accessing outside hardware while developing custom silicon for workloads where the query volume justifies the $500M+ design investment. Anthropic's annualized revenue run rate exceeds $30 billion (per reports), making the economics of a custom ASIC compelling if inference costs can be cut by ~50%.
For infrastructure architects, the strategic signal is clear: frontier-lab reliance on Nvidia continues, but the highest-margin players are now building escape routes via custom silicon. This diversification of the chip supply curve will likely reshape how hyperscalers bid out capacity contracts over the next 24 months. Anthropic's custom-chip timeline remains unknown—the company deliberately offered no production date—but the hiring pace and Samsung discussions suggest materialization within 18-36 months.