Anthropic has begun early-stage work on a custom AI inference chip and is in talks with Samsung Electronics as a potential manufacturing partner, according to reporting from multiple outlets. The company is exploring Samsung's advanced 2-nanometer process and packaging facilities to reduce reliance on Nvidia hardware, with the goal of running Claude faster and more efficiently at the scale customers demand. Samsung participated in Anthropic's $65 billion Series H funding round in May alongside SK Hynix and Micron.
Anthropic hired Clive Chan in June, the second engineer to join OpenAI's custom chip team, and had worked on OpenAI's Broadcom-designed Jalapeño inference processor for two and a half years. The project remains in early exploratory stages with no final design or manufacturing work begun. If successful, the partnership could mirror OpenAI's playbook: custom silicon optimized for a specific company's models can reportedly achieve 50% cost savings versus standard GPU inference.
This move positions Anthropic among the growing roster of AI labs developing proprietary silicon—Google (TPUs), Amazon (Trainium/Inferentia), Microsoft (Maia), and Meta (MTIA) all have custom chip programs. The shift reflects the AI industry's drive to reduce Nvidia dependence and optimize total cost of ownership for massive inference deployments as demand accelerates. A Samsung win would also strengthen the South Korean foundry's position against TSMC's dominance in cutting-edge AI chip manufacturing.