OpenAI and Broadcom have announced a chip designed for LLM adoption at scale

OpenAI, the company behind ChatGPT and Codex and the prototypes these tools use, and Broadcom, a leading silicon vendor, are involved. he announced A new chip called Jalapeño, designed specifically for the main language of data centers.

The chip is designed to be deployed in big data centers, both companies say this is the first generation of a long-term project that will see the chips developed over time.

Broadcom says that this ASIC (Application-Specific Integrated Circuit) was designed from the beginning for LLM inference, based on “details” in discussions between companies and researchers at OpenAI, and that the development of the chip was informed by OpenAI’s roads for models and future products. Designing and manufacturing the chip took nine months.

The promise is that this chip is more specific to the current needs of LLMs than the display systems used in existing environments.

OpenAI says that “initial testing shows that Jalapeño will provide more performance per watt than current devices,” but says that it has not been tested, and that “a detailed technical report will be provided in the coming months.”