OpenAI unveils Jalapeno, 1st custom chip to reduce NVIDIA reliance
Technology
OpenAI just dropped Jalapeno, its first-ever custom AI chip. Built with Broadcom, this move is all about boosting OpenAI's tech game and cutting back on dependence on NVIDIA GPUs.
Jalapeno is made for handling inference tasks and tackling the limits of current computing power.
Jalapeno designed in 9 months
Jalapeno was designed in only nine months (thanks to help from AI tools) and Broadcom's CEO says it rivals NVIDIA's Blackwell chips and Google's TPUs.
Manufactured by TSMC in Taiwan, it's tuned for big language models like GPT-5.3-Codex-Spark.
OpenAI plans to start using it by the end of 2026, kicking off a multigeneration chip program, with prototypes already running in labs.