OpenAI和Broadcom正在建立定制的AI芯片,以减少对Nvidia的依赖,目标是到2029年达到10千兆瓦的电力。 OpenAI and Broadcom are building custom AI chips to cut reliance on Nvidia, aiming for 10 gigawatts of power by 2029.

OpenAI与Broadcom合作开发定制的AI芯片,目标是到2029年部署10千兆瓦的计算能力,预计到2026年底将安装最初的系统。 OpenAI has partnered with Broadcom to develop custom AI chips, aiming to deploy 10 gigawatts of computing power by 2029, with initial systems expected in late 2026. 合作涉及开放国际公司为其模型设计最佳的AI加速器,而Broadcom公司则负责制造和供应联网技术。 The collaboration involves OpenAI designing AI accelerators optimized for its models, while Broadcom handles manufacturing and supplies networking technology. 此举是OpenAI减少对Nvidia等外部供应商依赖的更广泛战略的一部分,也是在全球对AI的需求不断上升的情况下扩大其基础设施规模的战略的一部分。 The move is part of OpenAI’s broader strategy to reduce reliance on external suppliers like Nvidia and scale its infrastructure amid rising global demand for AI. 该项目始于18个月前,反映了行业内部AI硬件的增长趋势,尽管人们对大规模投资的可持续性仍然感到关切。 The project, which began 18 months ago, reflects a growing industry trend toward in-house AI hardware, though concerns persist about the sustainability of massive investments. Broadcom的股票在新闻上上涨了9%以上。 Broadcom’s stock rose over 9% on the news.