OpenAI unveils Jalapeño inference processor designed with Broadcom

According to TechCrunch and The Verge, OpenAI announced Jalapeño, a custom-built inference processor designed and manufactured in collaboration with Broadcom. OpenAI said the chip is designed specifically for its inference systems, with early results showing significantly better performance-per-watt than current alternatives. The partnership reduces OpenAI's dependence on Nvidia GPUs, following similar moves by Google and Amazon to build custom AI accelerators.

Update (27 June 2026): TechCrunch further described Jalapeno as a custom ASIC built with Broadcom to power current and future OpenAI models, part of a wider industry shift toward proprietary silicon that also includes Google, Apple, and SpaceX, aimed at reducing single-supplier dependence on Nvidia. TechCrunch compared the potential performance gains to those Apple achieved after moving away from Intel processors.