SAN FRANCISCO, June 24 — OpenAI Group PBC today revealed a custom chip called Jalapeño that it will use to power its large language models. The processor is the fruit of a collaboration with Broadcom Inc., which is no stranger to designing custom silicon for some of the biggest names in tech. Here’s what it means, in plain terms: OpenAI is no longer willing to rely entirely on off-the-shelf chips from companies like Nvidia to run its AI models.
The company has designed its own chip, tailored specifically for the kind of math that makes ChatGPT and its successors tick. Think of it like a restaurant that got tired of using a standard oven and decided to build one that cooks its signature dish perfectly every time.
The Jalapeño chip is designed for inference — the process of running a trained AI model to generate an answer, create an image, or analyze text. Training models is a different beast, and OpenAI is still using other hardware for that. The practical upshot for users could be faster responses, lower costs, or both.
OpenAI has been spending heavily on computing infrastructure, and owning its own inference silicon gives it more control over that bill. Broadcom handled the physical design and manufacturing logistics.
The company has a long track record here; it has built custom chips for Google’s Tensor Processing Units and other hyperscale projects. For OpenAI, the partnership means it gets to skip the years-long process of building a chip design team from scratch. Broadcom brings the engineering muscle, OpenAI brings the knowledge of exactly what its models need.
Why the name Jalapeño? The company hasn’t said, but it fits a pattern. OpenAI has used spicy food names for internal projects before.
The chip is not a one-off experiment. Sources familiar with the matter indicate that OpenAI plans to deploy Jalapeño chips at scale in its data centers, gradually shifting inference workloads off general-purpose hardware.
This move puts OpenAI in a club that already includes Amazon, Google, and Microsoft — all of which have designed custom chips for AI workloads. The difference is that OpenAI is primarily an AI company, not a cloud provider. Its decision to invest in custom silicon signals that it sees hardware as a strategic advantage, not just a cost to be managed.
The timing matters. Demand for AI inference is exploding as more companies and consumers use AI-powered products.
Every query to a large language model requires computation, and that adds up fast. A chip that can do that work more efficiently than a general-purpose GPU gives OpenAI a competitive edge. It also reduces its dependence on a single supplier, which has been a concern for the entire industry.
So what’s next? The Jalapeño chip will need to prove itself in production. OpenAI will have to show that it can deliver on the promises of better performance and lower cost.
If it works, expect other AI companies to follow suit. The era of generic chips for AI may be giving way to a world where every major player builds its own.
{“@context”: “https://schema.org”, “@type”: “NewsArticle”, “headline”: “OpenAI, Broadcom debut custom Jalapeño chip for AI inference”, “description”: “SAN FRANCISCO, June 24 — OpenAI Group PBC today revealed a custom chip called Jalapeño that it will use to power its large language models….”, “datePublished”: “2026-06-24 20:30:46T09:30:00+00:00”, “author”: {“@type”: “Person”, “name”: “Erika Johnson”}, “publisher”: {“@type”: “Organization”, “name”: “InfoPulse Today”, “url”: “https://newsfeed.infopulsetoday.com/”}}
