CoreWeave Achieves Major AI Infrastructure Milestone with Large-Scale Nvidia GPU Cluster
CoreWeave (CRWV), an AI cloud provider, has successfully linked hundreds of Nvidia's newest Rubin GPUs into a single working cluster using Spectrum-X Ethernet networking. This marks a significant step beyond its June milestone when it first deployed a single rack containing 72 Rubin GPUs and 36 CPUs. The new multi-rack system supports up to approximately 128,000 GPUs on one non-blocking network, enabling large-scale AI training without performance bottlenecks. The achievement is seen as critical for agentic AI systems that require continuous learning capabilities. While the stock rose about 3% following the announcement—less than its previous jump of nearly 14% in June—analysts remain divided on CoreWeave's long-term prospects due to concerns over its heavy debt load, which stands at $51.6 billion against only $5.5 billion in cash. In addition to GPU scaling, CoreWeave also announced three upgrades to its AI Object Storage system aimed at improving data throughput and reducing latency for customers using the GPUs.
Topics: AI cloud computing CoreWeave Nvidia GPUs