NVIDIA and Elon Musk’s xAI are ramping up their colossal supercomputer, Colossus, to an unprecedented 200,000 Hopper GPUs—doubling the current 100,000 GPU setup in a scale never before seen. Leveraging NVIDIA’s advanced Spectrum-X Ethernet networking for seamless data throughput and zero packet loss, Colossus has already set records, becoming operational in just 122 days. As xAI’s Grok large language models expand, the additional 100,000 GPUs will enable faster, more efficient AI model training, accelerating xAI’s generative AI solutions for consumers and enterprises alike.
https://nvidianews.nvidia.com/news/spectrum-x-ethernet-networking-xai-colossus