OpenAI and Broadcom on Wednesday unveiled Jalapeño, OpenAI's first homegrown AI accelerator — a purpose-built chip designed specifically to run large language model inference, the process that powers every ChatGPT conversation and Codex coding request. Engineering samples of the chip, manufactured by TSMC, were delivered to OpenAI CEO Sam Altman and President Greg Brockman this morning by Broadcom President and CEO Hock Tan and Semiconductor Solutions President Charlie Kawwas at OpenAI's San Francisco headquarters. For any developer building on the OpenAI API, or any user whose cost of access to ChatGPT scales with usage, the arrival of a purpose-built inference chip with a credible cost reduction claim is the single development that could most directly change what AI services cost to run — and therefore what they cost to use.
To read more, click here.