The team of independent industry analysts at SemiAnalysis has unveiled the second iteration of their GPU Cloud ClusterMAX™ Rating System.
We are honored to share that Verda (formerly DataCrunch) has retained the bronze ranking from the previous evaluation in March 2025.
We are excited to unveil future improvements to our Instant Clusters as we continue working toward building the industry-leading product for self-service and large-scale model training.
ClusterMAX
The ClusterMAX Rating System by SemiAnalysis represents the most widely accepted evaluation framework for GPU infrastructure, particularly for model training at scale.
The second iteration of ClusterMAX has grown in both scale and requirements:
- The evaluation included over 200 cloud providers ranging from hyperscalers to neoclouds, as well as adjacent services.
- The list of requirements grew to include 10 criteria. We encourage you to learn more about the evaluation methodology directly from SemiAnalysis.
Instant Clusters
Instant Clusters is our product offering that the SemiAnalysis team evaluated in ClusterMAX v2. The key value proposition of Instant Clusters is to enable rapid provisioning to small- and mid-sized clusters (16x-128x GPUs).
Each GPU node contains:
- 8x B200 SXM6 GPUs
- 1440GB GPU VRAM
- 240 AMD Turin CPU cores
- 3200 Gbit/s InfiniBand interconnect
- 100 Gbit/s ethernet
- 5 Gbit/s uplink
One of the advantages of Instant Clusters is self-service and pay-as-you-go access, which enables you to end deployments at a 10-minute notice. This way, you can rapidly provision GPU clusters for as long as you need them, whether for days or mere hours.
To battle-test the Instant Clusters with a realistic workload, we allocated multiple nodes for the 1X World Challenge. Including two ML Engineers at Verda, the team topped both tracks: sampling and compression. Learn more about the challenge and the team’s submission from their technical report.
One of the highlights of the evaluation was the Slurm implementation. To quote the SemiAnalysis team: “We were also impressed by the completeness of their Slurm implementation, which stands in stark contrast to many other providers on the bronze or even silver tier of this list.”
You can read the evaluation in full from SemiAnalysis.
Looking ahead
We invite you to test our Instant Clusters with B200 GPUs and InfiniBand interconnect, all with self-service and pay-as-you-go access via our Cloud Dashboard.
If you are looking for other flavors of the Blackwell platform, you can:
- Deploy B200 SXM6 (and soon B300 SXM6) instances via our Cloud Platform
- Get early access to GB300 NVL72 by contacting us