Lambda vs CoreWeave
Both appear in 1 section: gpus. CoreWeave is cheaper on 3 of 3 comparable measures.
Both of these sell GPU capacity as proper infrastructure rather than a marketplace, and on published on-demand pricing Lambda is the cheaper of the two on most models. An H100 is $3.29/hr at Lambda against $6.16/hr at CoreWeave, which is 87% more. An A100 is $2.79 against $2.70, essentially level. An H200 is not listed by Lambda in this data; CoreWeave asks $6.31. A B200 is $9.86 against $8.60, where CoreWeave is cheaper. CoreWeave's H100 price is the highest in Pikkly's on-demand table, and the reason is that its published on-demand rate is not really the product. CoreWeave sells reserved capacity at negotiated rates to companies committing to large fleets for months or years, with InfiniBand fabric, Kubernetes-native scheduling and the kind of multi-node topology that distributed training at scale requires. The on-demand number is the price of not committing. Lambda sits between that and the marketplaces: its own data centres, a maintained ML environment, multi-GPU nodes, and hourly pricing you can actually use without a contract. If you are renting a handful of GPUs for days or weeks, Lambda's published price is the one that applies and it is substantially lower. If you are planning a cluster and will sign for capacity, CoreWeave's list price tells you almost nothing and the conversation is the real comparison.
GPUs
All gpus →Neither provider runs an affiliate programme; Pikkly earns nothing from this page.