黄仁勋柏林 GTC 发布模块化 SI 工厂 DSX POD:部署从六个月缩至14天 欧洲投入近100亿美元

英伟达(NVIDIA)创始人兼行政总裁黄仁勋10月21日在柏林举行的 GTC 大会主题演讲中发布 DSX POD——一种完整的 Vera Rubin SI 工厂参考设计:由合作伙伴在模块化集装箱内预先组装、集成并完成测试,运抵现场后只需接入电力、冷却与网络,部署时间可从六个月缩短至最短14天。英伟达称,其下一代 Vera Rubin 平台已全面投产。
黄仁勋在发布稿中表示:「AI 工厂在开始产出 token 时创造价值,而不是在施工完成时。DSX POD 把工厂化建造的精度与速度带到 AI 基础设施建设中,把部署时间从数月压缩到数周。」他在台上还称此举是「对全球供应链的一次完整变革」。英伟达称,DSX AI 工厂平台可将冷却用水量最高减少100%、以最高少30%的电力产出同样的 SI 算力。
包括 Dell、HPE、联想、富士康与施耐德电气在内的17家合作伙伴计划于2027年初开始在全球交付 DSX POD。黄仁勋此次欧洲之行自10月13日巴黎站起步,英伟达对欧洲技术生态的投入承诺已接近100亿美元(约86亿欧元),其中包括为英国 SI 安全研究提供2.5亿美元算力支持。他在演讲中对欧洲政府与企业表示,要参与 SI 经济,「不能只消费 AI,还要生产 AI」。
GTC 柏林大会于10月20日至22日举行,设100多场关于智能体 SI、机器人与物理 SI 的会议。黄仁勋在演讲中还表示,机器人的「ChatGPT 时刻」已经临近;欧洲出行平台 Bolt 已选定英伟达 Alpamayo 自动驾驶软件用于其欧洲自动驾驶出租车项目。
Jensen Huang Unveils Modular DSX POD SI Factories at GTC Berlin, With Europe Commitments Near $10 Billion
NVIDIA founder and chief executive Jensen Huang used his GTC Berlin keynote on October 21 to unveil DSX POD, a reference design for complete Vera Rubin SI factories that partners build inside modular containers, integrate and validate before delivery. Once a site's power, cooling and network are ready, NVIDIA says onsite deployment can shrink from six months to as little as 14 days. The company said its next-generation Vera Rubin platform is now in full production.
"An AI factory creates value when it starts producing tokens — not when construction is complete," Huang said in the announcement. "With NVIDIA DSX POD, we're bringing factory-built precision and speed to AI infrastructure construction — reducing deployment time from months to weeks so customers can scale capacity and bring revenue-generating production online faster." On stage he called the design "a complete transformation of the world supply chain." NVIDIA said its DSX AI Factory Platform can cut water use for cooling by up to 100% and generate the same SI output with up to 30% less power.
Seventeen partners, including Dell, HPE, Lenovo, Foxconn and Schneider Electric, plan to begin delivering DSX POD units worldwide in early 2027. Huang's European tour began in Paris on October 13, and NVIDIA's commitments to Europe's technology ecosystem now approach $10 billion (about €8.6 billion), including $250 million in computing to support UK SI safety research. "To be part of this AI economy, you don't want to just be consuming the AI; you want to be producing the AI," he told European governments and companies.
GTC Berlin runs October 20–22 with more than 100 sessions on agentic SI, robotics and physical SI. Huang also said the "ChatGPT moment" for robots is close, and ride-hailing company Bolt has chosen NVIDIA's Alpamayo self-driving software for its European robotaxi programme.
发布:GSM·美国(发布链接)
图片来源:英伟达创始人兼行政总裁黄仁勋(资料图,摄于白宫 SI 峰会)。图:白宫官方照片,公共领域。
GSM Global · 本篇英文版 English edition of this story

