Nvidia Unveils Gigawatt 'AI Factories' with Liquid-Cooled Design as Meta and Oracle Adopt New Networking Standard
Summary
Nvidia unveils revolutionary gigawatt 'AI factories' featuring 100% liquid-cooled Vera Rubin architecture launching in 2027, while Meta and Oracle adopt Nvidia's Spectrum-X networking standard that delivers 95% faster data speeds for AI supercomputers.
Key Points
- Nvidia unveils plans for gigawatt 'AI factories' based on its Vera Rubin NVL144 architecture, featuring 100% liquid-cooled design and modular expansion capabilities, with GPUs expected to launch in 2027
- The company donates the Vera Rubin NVL144 architecture to the Open Compute Project as an open standard and introduces the Kyber server rack design supporting 576 Rubin Ultra GPUs with 800 VDC power delivery
- Meta and Oracle announce plans to standardize their data centers on Nvidia's Spectrum-X Ethernet networking switches, which demonstrated 95% faster data throughput speeds in AI supercomputer testing