CoreWeave, a specialized cloud provider built for large-scale GPU-accelerated workloads, is among the first to offer cloud instances with NVIDIA HGX H100 supercomputing.
NVIDIA claims its HGX H100 enables up to seven times better efficiency in high-performance computing (HPC) applications, up to nine times faster AI training on the largest models and up to 30 times faster AI inference than the NVIDIA HGX A100. That speed, combined with the lowest NVIDIA GPUDirect network latency in the market with the NVIDIA Quantum-2 InfiniBand platform, reduces the training time of AI models to “days or hours instead of months.” Such technology is critical now that AI has permeated every industry.
CoreWeave says its Kubernetes-native infrastructure and networking architecture produce performance advantages, including industry-leading spin-up times and responsive auto-scaling capabilities that allow clients to use compute more efficiently. CoreWeave competitors charge for idle compute capacity to maintain access to GPUs and use legacy-networking products that degrade performance with scale.
“This validates what we’re building and where we’re heading,” said Michael Intrator, CoreWeave co-founder and CEO. “CoreWeave’s success will continue to be driven by our commitment to making GPU-accelerated compute available to startup and enterprise clients alike. Investing in the NVIDIA HGX H100 platform allows us to expand that commitment, and our pricing model makes us the ideal partner for any companies looking to run large-scale, GPU-accelerated AI workloads.”
