Nvidia

OpenAI and Broadcom Unveil Jalapeño, a Custom AI Inference Chip

The processor is meant to cut the cost and energy of running large language models and give OpenAI more control over the hardware that serves ChatGPT and other products.