OpenAI Unveils Custom Jalapeño AI Inference Chip, Reportedly Outperforming NVIDIA's GB200/GB300 Systems

OpenAI has introduced its first custom AI inference chip, codenamed 'Jalapeño,' which reportedly delivers significantly higher throughput per kilowatt and lower latency compared to NVIDIA's GB200 and GB300 rack systems in initial benchmarks. Samsung is reportedly supplying HBM4 for the new chip, marking a major development in AI hardware and the semiconductor industry.