OpenAI to launch its first AI chip in 2026 with Broadcom

OpenAI is building its own custom AI chip with Broadcom, aiming to roll it out in 2026 for internal use only.

‍

- The chip is designed to reduce reliance on Nvidia’s GPUs and give OpenAI tighter control over costs and supply

- Broadcom disclosed over $10B in AI infrastructure orders from a new customer — widely believed to be OpenAI

- Mass production is expected to begin in 2026, with chips optimized for OpenAI’s own workloads

- OpenAI now joins Google, Amazon, and Meta in the race to develop in-house silicon for AI

‍

Will custom chips give OpenAI the edge it needs — or just spark another costly arms race?