OpenAI and Broadcom Partner to Develop Custom AI Accelerator Chips for 10 GW of Infrastructure

OpenAI has announced a collaboration with Broadcom to develop its own custom AI accelerator chips and systems, aimed at powering 10 gigawatts of infrastructure. The partnership, which follows earlier deals with NVIDIA, AMD, and Oracle, is expected to begin deployment in the second half of 2026 and reach full implementation by 2029.

According to OpenAI, the collaboration involves a multi‑billion‑dollar contract spanning 10 gigawatts of computing capacity, a figure echoed by Broadcom CEO Hock Tan who previously disclosed a $10 billion order from an unnamed client now confirmed to be OpenAI. The customised AI accelerator chips will power OpenAI’s core infrastructure as well as the data centres of its strategic partners.

OpenAI’s move towards vertical integration and own silicon aims to reduce dependence on external suppliers while aligning hardware closely with its AI workloads, with potential long‑term commercial opportunities from these systems.