AWS将英伟达GPU路线图扩至200万颗,覆盖Blackwell
AWS expanded its NVIDIA roadmap by 2M GPUs across Blackwell Ultra, Rubin and Rub…
AWS expanded its NVIDIA roadmap by 2M GPUs across Blackwell Ultra, Rubin and Rubin Ultra.
The added Blackwell Ultra, Rubin and Rubin Ultra GPUs are scheduled for 2027-2028, on top of AWS's earlier 1M-plus plan.
AWS announced that earlier batch in March for deployment starting in 2026.
NVIDIA's role now reaches across Vera CPUs, Spectrum networking, Nemotron models, CUDA-X libraries and Amazon Robotics' physical AI platform.
Future Trainium chips will also work with NVIDIA NVLink Fusion and NVHBM, tying Amazon's accelerators to NVIDIA interconnect and memory technology.
That lets AWS keep expanding custom silicon while using NVIDIA components around it at the rack level.
更进一步:量化金融体系
看懂新闻只是起点——沿量化金融路径,把它变成能交付的工程能力