SAN FRANCISCO — OpenAI is set to co-design and produce its own artificial intelligence chips in partnership with Broadcom, in a deal reportedly exceeding $10 billion. The initiative marks a strategic pivot to reduce reliance on NVIDIA and secure long-term control over its compute infrastructure.
The chips, internally referred to as XPUs, are tailored for OpenAI’s proprietary workloads and will not be sold commercially. Production is scheduled to begin in 2026.
Broadcom, already a key partner for Google, Meta, and ByteDance in custom chip development, now counts OpenAI as its fourth major client. The announcement triggered a surge in Broadcom’s stock, while shares of NVIDIA and AMD dipped amid concerns over shifting market dynamics.
OpenAI CEO Sam Altman has emphasized the urgent need for scalable compute resources to support models like GPT-5. The Broadcom collaboration is expected to significantly bolster OpenAI’s hardware autonomy and performance capabilities.
