CoreWeave's Nvidia supply advantage

This reveals that CoreWeave won the first phase of the AI infrastructure race by getting scarce Nvidia chips faster than the hyperscalers, then wrapping them in a GPU cloud that was cheaper than AWS for the same H100 class workloads. In practice, that let customers reserve large GPU clusters when AWS, Azure, and GCP often had limited supply, while still buying a cloud product with autoscaling, networking, and Kubernetes support instead of raw rented servers.1 2 3 4 5

• The Nvidia relationship was concrete, not just symbolic. Nvidia invested in CoreWeave, CoreWeave became one of Nvidia's largest customers in 2023, and by early 2025 CoreWeave said it had more than 250,000 deployed GPUs and was among the first cloud providers to roll out H100, H200, GH200, and GB200 systems.1 2 5
• The price advantage was real for production buyers. A CoreWeave customer described AWS H100 instances as costing two to three times as much as CoreWeave, while CoreWeave still delivered the production features that made it usable for live model serving. The same customer used cheaper Lambda for experiments, but kept production on CoreWeave because it behaved more like AWS for GPU workloads.3
• This also explains why CoreWeave scaled faster than most GPU clouds. The company went from an estimated $229M of revenue in 2023 to $1.9B in 2024 and expanded from 3 data centers in 2023 to 28 by the end of 2024, while peers like Lambda and Crusoe were smaller and more segmented by use case, price point, or energy angle.2 4 6 7

Going forward, the supply edge turns into a software and power race. As hyperscalers and rivals catch up on GPU access, CoreWeave's advantage will depend less on getting Nvidia hardware first and more on whether it can stay the default place to run large AI clusters with the best uptime, fastest provisioning, and enough power capacity to keep adding the next wave of Nvidia systems.2 3 5