OpenAI and Broadcom to build custom inference chip Jalapeño amid industry shift away from Nvidia

OpenAI disclosed plans to develop Jalapeño, a custom inference chip built in collaboration with Broadcom, as part of a broader industry trend toward reducing reliance on Nvidia. The company described the initiative as a hedge against single-supplier dependency rather than a complete departure from existing suppliers.

The effort is framed as a move to gain more control over hardware and achieve performance gains tailored to specific AI needs, drawing a parallel to Apple’s transition from Intel processors to its own silicon. OpenAI’s announcement places it among a growing list of companies—including Google, Apple, and SpaceX—that are developing custom silicon to address performance and supply chain concerns.

The disclosure was made during a TechCrunch Equity podcast segment, where hosts discussed the implications of custom chip development for the AI industry and highlighted recent deals of the week.