OpenAI unveils “Jalapeño” AI inference chip with Broadcom

• OpenAI announced its first custom AI chip, “Jalapeño,” developed with Broadcom (NASDAQ:AVGO).
• The chip is designed for AI inference and is reportedly outperforming current leading chips in early testing.
• The move signals OpenAI’s push to reduce reliance on Nvidia (NASDAQ:NVDA) for AI compute infrastructure.

OpenAI has unveiled its first custom-designed AI chip, called Jalapeño, developed in partnership with Broadcom (NASDAQ:AVGO).

The chip is designed specifically for AI inference workloads, which involve running trained models to generate outputs for users and applications.

“Jalapeño is part of our long-term full-stack infrastructure strategy,” said OpenAI President Greg Brockman, describing the goal as making AI compute more efficient, accessible, and cost-effective.

OpenAI said early testing shows the chip outperforming current state-of-the-art processors, although detailed benchmarks have not been publicly disclosed.

OpenAI said the chip was developed in about nine months and represents the first stage of a broader multi-generation hardware roadmap for AI infrastructure.

The launch reflects a wider industry trend, with major cloud and AI firms including Amazon (NASDAQ:AMZN), Microsoft (NASDAQ:MSFT), and Google (NASDAQ:GOOG) building their own custom chips to reduce reliance on external suppliers.