OpenAI Debuts Jalapeño AI Inference Chip and Previews Astra Models

OpenAI has unveiled its first custom AI inference chip, named Jalapeño, which reportedly offers significantly higher throughput per kilowatt and lower end-to-end latency compared to NVIDIA's GB200 and GB300 rack systems in initial benchmarks. This development could make Jalapeño a major new contender for HBM4 supply. Additionally, OpenAI previewed its upcoming Astra family of AI models, demonstrating capabilities such as multi-agent AI for complex math problems and navigation of desktop software. This move comes as OpenAI aims to regain its lead in the AI race, having reportedly fallen behind rivals like Anthropic.