6.7 Tech OpenAI与博通联合发布Jalapeño AI芯片，专为LLM推理设计，成本较GPU降低50%，2026年起大规模部署，未来将持续扩展。
OpenAI与博通（Broadcom）联合推出了Jalapeño，这是一款专为大型语言模型（LLM）推理任务（如ChatGPT和Codex）设计的定制AI芯片。Jalapeño采用ASIC架构，针对特定AI工作负载进行了优化。博通表示，该芯片相较于传统AI GPU可节省约50%的成本。 Jalapeño由台积电（TSMC）制造，开发周期为九个月，预计将于2026年底开始大规模部署。双方还规划了未来芯片迭代的路线图，下一代产品计划于2028年推出，并计划到2029年在OpenAI数据中心部署10吉瓦定制AI加速器。 消息发布后，博通股价上涨2%，而英伟达（Nvidia）股价略有下跌。OpenAI与博通强调，这一合作标志着战略性转向，逐步减少对Nvidia的依赖，致力于提升AI计算的速度、可靠性和性价比。
STOCKS | U.S. Stocks Open Mixed as Broadcom Rises on OpenAI AI Chip Announcement
OpenAI unveils its first custom inference chip, built with Broadcom
Sam Altman takes on Nvidia with OpenAI’s Jalapeño chip
Sam Altman's OpenAI Unveils First Intelligence Chip As Global AI Race Intensifies
Broadcom unveils custom chip for OpenAI, challenging Nvidia’s dominance
OpenAI Turns Up the Heat With Jalapeño, Its First Custom AI Chip
OpenAI and Broadcom (AVGO) Unveil ‘Jalapeño’ AI Chip to Challenge Nvidia (NVDA) Dominance
Broadcom (AVGO) and OpenAI New Custom AI Chip “Jalapeño” Takes Aim at Nvidia
OpenAI and Broadcom unveil LLM-optimized intelligence processor in 10-gigawatt chip partnership
OpenAI and Broadcom Release AI Chip Jalapeño, Cutting Inference Costs by Up to 50%
