OpenAI and Broadcom Unveil Custom 'Jalapeno' AI Chip to Run Models Faster and Cheaper

OpenAI has unveiled its first custom-designed AI chip, developed in partnership with semiconductor giant Broadcom, in a landmark move toward vertical integration of its AI infrastructure. The chip, named 'Jalapeno', is built to run AI models faster and more cheaply than existing solutions, and signals OpenAI's ambition to own the full technology stack — from model development down to the silicon.

The announcement, confirmed across Reuters, Bloomberg, CNBC, and Firstpost, positions OpenAI alongside Google, Amazon, and Meta as one of the few AI companies to develop proprietary chips. For years, the AI industry's explosive growth has been almost entirely dependent on Nvidia's GPUs, creating supply bottlenecks and soaring costs. Jalapeno is designed to ease that reliance for OpenAI's own inference workloads — running existing models at scale rather than training new ones from scratch.

OpenAI has framed the chip as part of a broader strategy to 'build the full stack', meaning the company wants end-to-end control over hardware, software, and AI models. The Broadcom partnership provides the manufacturing and chip design expertise that OpenAI, as a software-first company, does not have in-house. The chip is expected to be deployed in OpenAI's own data centres to serve its products, including ChatGPT.

The development comes as the AI infrastructure arms race intensifies globally. Custom silicon typically offers significant efficiency advantages for specific workloads, and cheaper inference could eventually translate into lower costs for API customers and potentially end users. For the broader AI industry — including Indian startups and enterprises that build on OpenAI's platform — a more cost-efficient OpenAI could mean more competitive pricing and greater accessibility.