DataCrunch vs Lambda
Both appear in 1 section: gpus. DataCrunch is cheaper on 4 of 4 comparable measures.
DataCrunch is materially cheaper than Lambda on every GPU both of them list, and the gap is not small. An on-demand H100 is $3.25/hr at DataCrunch against $3.29/hr at Lambda, which is level, but the A100 tells the real story at $1.29 against $2.79, where Lambda is 116% more expensive. A B200 is $6.11 against $9.86. DataCrunch also runs a spot tier that goes well below its on-demand rates, with H100s reaching $1.63/hr and A100s $0.65/hr. DataCrunch is a Finnish provider running its own hardware in the Nordics, which is where some of the cost advantage comes from: cheap hydroelectric power and cool air are real inputs to a GPU hosting bill. It offers on-demand and spot instances, NVLink multi-GPU nodes, and an API, and it is smaller than Lambda in every dimension including capacity, region count and support depth. Lambda's case is scale and maturity: more regions, a longer track record, a maintained ML stack many teams already build against, and enough capacity that a large request is likely to be met. For a team in or near Europe renting a handful of cards, DataCrunch's prices are hard to argue with and the spot tier is cheaper than most marketplaces. For a workload that needs guaranteed capacity in North America, or a vendor with a longer history behind procurement, Lambda is the safer answer at roughly double the hourly rate.
Pikkly earns nothing from Lambda, and the ranking above is by price regardless.