Nvidia 发布 DSX POD 预制集装箱数据中心，部署周期有望从六个月压缩到14天
Nvidia's DSX POD could cut AI data center setup from six months to as little as 14 days
- 1Nvidia 发布围绕 Vera Rubin 系统构建的 DSX POD 模块化数据中心参考设计
- 2Nvidia 称集装箱预制方式可将 AI 数据中心部署周期从六个月缩短至最少 14 天
- 3Dell、HPE、联想、施耐德电气等合作伙伴在集装箱内预装计算、网络、供电与液冷
- 4Nvidia 称冷却水用量最高可减少100%，首批设备定于2027年初全球出货
Nvidia 把 AI 数据中心建设工业化成工厂预制的集装箱产品，将建设瓶颈转移到场地准备和水电接入。
- 影响：若官方数据成立，自建或租用 AI 算力的等待周期从六个月压缩到两周量级，选址和规划节奏可以前移；场地仍需预先备好电力、散热和网络，部分极热气候还要追加冷却。
- 参考：评估 AI 算力投入时，可考虑把这类工厂预制集装箱方案纳入比选，先算出自建与预制的交付周期差；值得提前盘点候选场地的供电与散热条件，判断能否衔接快速部署的前提。
- 依据：14 天部署和最高省100%冷却水均为 Nvidia 在 GTC Berlin 发布会上宣布的官方数字，暂无独立验证；首批设备定于2027年初出货。
Nvidia wants AI data centers to ship fully preassembled in containers. At GTC Berlin, the company unveiled DSX POD, a reference design for modular data centers built around Nvidia's Vera Rubin systems. Partners such as Dell, HPE, Lenovo, and Schneider Electric install the computers, networking gear, cabling, power supply, and liquid cooling in containers at the factory and test them there.
On site, the units only need to be hooked up to the power, cooling, and network of a prepared location. Nvidia says this cuts setup time from six months to as little as 14 days. The company argues that the sooner a data center runs AI models, the sooner its operator starts making money.
Nvidia also says the broader DSX platform can use up to 100 percent less cooling water because its closed-loop cooling systems release heat into the air without evaporation. The company admits that some climates will still need extra cooling during extreme heat. The first units are set to ship worldwide in early 2027.
AI News Without the Hype – Curated by Humans
Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.
来源：The Decoder：AI News · the-decoder.com