OpenAI and Broadcom Unveil Jalapeño, a Chip Built to Halve Inference Costs (opens in new tab)

OpenAI and Broadcom unveiled Jalapeño on Tuesday, the company's first custom inference chip, built in nine months on TSMC 3nm with AI-assisted design. The chip targets 50% cheaper inference than Nvidia GPUs — a claim that, so far, rests on internal benchmarks alone.