June 24, 2026
Too hot to ship?
OpenAI unveils its first custom chip, built by Broadcom
OpenAI’s new “Jalapeño” chip is here, and the internet is already roasting the name
TLDR: OpenAI has unveiled its first custom chip with Broadcom to make its AI responses faster and cheaper, a big step toward relying less on Nvidia. Commenters were split between calling it a smart power move, nitpicking who’s actually making it, and relentlessly clowning the name “Jalapeño.”
OpenAI just revealed Jalapeño, its first custom chip made with Broadcom, and the official pitch is simple: make its AI answers cheaper, faster, and less dependent on Nvidia’s famously in-demand hardware. In plain English, this is the silicon brain meant to help run ChatGPT-style responses more efficiently. OpenAI says early tests look promising, especially when it comes to getting more work done with less power — and yes, the company proudly noted that its own AI helped design it. Very AI builds AI energy.
But the real fireworks were in the comments. Some people saw a huge strategic flex, saying this is OpenAI finally following the path Google blazed with its in-house chips years ago. One commenter basically declared Google’s big chip bet “infinitely more prescient,” while others zeroed in on the bigger business story: this looks like OpenAI trying to loosen Nvidia’s grip on the AI world, even if the new chip is mostly for serving answers rather than training giant models.
And then came the comedy. One of the loudest reactions had nothing to do with performance and everything to do with the name. “Jalapeño” sparked instant eye-rolling, with one commenter calling it cringe and complaining that the “ñ” will be annoying in everyday use. Another person joked, “So this is where all the memory they bought is going,” while dreamers were already asking for a tiny super-powered version they could put on a desk at home. Even a manufacturing mini-mystery popped up, as commenters rushed to clarify that TSMC — not Intel — appears to be making it. In classic internet fashion, the chip may be cutting-edge, but the comment section immediately turned it into a naming war, a supply-chain detective story, and a wishlist thread.
Key Points
- •OpenAI unveiled Jalapeño, its first custom-built inference chip, developed with Broadcom.
- •The company said early testing shows significantly better performance-per-watt than current state-of-the-art alternatives.
- •The chip is aimed at reducing OpenAI’s dependence on Nvidia GPUs for some AI workloads.
- •Jalapeño is designed specifically for inference, with OpenAI highlighting low operating costs for real-time coding models.
- •OpenAI presented the chip as part of a full-stack infrastructure strategy spanning models, data centers, chip architecture, memory, networking, scheduling, and deployment systems.