tech
OpenAI and Broadcom Unveil 'Jalapeño' AI Chip
OpenAI, in partnership with Broadcom, has announced its first custom-designed AI processor, codenamed 'Jalapeño.' The new chip is an ASIC specifically optimized for large language model inference and is intended to reduce OpenAI's reliance on GPUs from suppliers like Nvidia. Initial deployment in data centers is expected by the end of the year.
TL;DR
- OpenAI and Broadcom unveiled Jalapeño, an ASIC designed specifically for LLM inference.
- The chip was developed in nine months and engineering samples are already running future models.
- Jalapeño aims for performance per watt substantially better than current state-of-the-art.
- The chip is intended to reduce OpenAI's reliance on Nvidia GPUs.
- Commercial use is expected by the end of the year, with real volume in 2027.
- Jalapeño is part of a trend of custom AI silicon development by major tech companies.
- Some observers suggest local and open-source AI might counterbalance vertically integrated giants.