SemiAnalysis released its ClusterMAX™ 3.0 ratings this week, and Verda was rated Bronze. Of the 77 neoclouds put through full testing, only 19 landed a ‘recommended AI cloud provider’ medallion rating this cycle, down from 41 last time. We're one of them.
Since our last rating, we shipped Slinky, giving you Slurm on top of Kubernetes, and Kueue for native Kubernetes job queuing, the same architecture SemiAnalysis tested. We also kept shipping through the testing window itself: SemiAnalysis notes a further round of changes we pushed after their testing had already closed. The cluster in this report is already a version behind what's live today.
Our commitment to security
SemiAnalysis gave particular attention to security this cycle, following on from an earlier piece examining security practices across the neocloud industry. Their criteria go beyond certifications on a wall. They test tenant isolation at the VM level, patch currency against known CVEs, and the security of the network fabric itself. Testing Verda's cluster, SemiAnalysis found our security posture to be ‘solid’. It's a distinction we care about.
"Security is integral to how we build. We invest in it continuously, not just as an afterthought. We're glad to see that reflected in SemiAnalysis's testing,"
- Michael Champion, CISO, Verda
We continue to build
In September, we layered in continuous health monitoring across both scheduler types: passive checks that run alongside your workloads at no measurable overhead, and active per-node diagnostics every six hours on idle nodes.
Our instant clusters keep scaling as well. You can now deploy up to 144 NVIDIA HGX™ B300 GPUs, 128 B200, or 64 H200 per cluster on InfiniBand, alongside GB300 NVL72 configurations from two trays up to eighteen. Instant clusters are provisioned in as little as 15 minutes and available with pay-as-you-go contracts, giving you maximum flexibility for rapid deployments.
The team behind it
SemiAnalysis credited our engineers directly during testing. They were responsive, quick to explain the reasoning behind design choices, and willing to adjust course based on what testing surfaced. It's how we work with every customer, every day.
Writing about Verda's trajectory more broadly, SemiAnalysis said they remain "optimistic about their engineering team and roadmap." That roadmap runs deeper than the cluster layer. The same week this rating was published, we announced a $189M Series B to extend that capacity across Europe, the US, and Asia and to go deeper on the platform itself including object storage, managed kubernetes, and networking.
Spin up your own cluster today:
- Log onto our console
- Or learn more about our CLI, API, and other provisioning methods in our docs
