OpenAI chip roster: Cerebras, NVIDIA, AMD, and now Broadcom
OpenAI has finalized supply commitments across four major silicon partners — Cerebras (announced January 2026), NVIDIA (existing), AMD (existing), and now Broadcom for custom inference ASICs reportedly in design at TSMC.
The four-way diversification is unprecedented at this scale. Most hyperscalers run two or three substrates concurrently; OpenAI's four-vendor posture suggests they anticipate workload growth that would saturate any single supplier.
The Broadcom partnership is the most informationally novel — it confirms long-running rumors of OpenAI custom silicon. Like the Google TPU program before it, this would put OpenAI in the rare position of designing its own training-and-inference accelerators.