
Verda
5.9/10Helsinki-based GPU cloud, formerly DataCrunch, that rents NVIDIA GPU instances, instant clusters, serverless containers and inference APIs to AI teams.
Data last updated November 2026 · Our methodology
Overview
Verda is a European AI cloud. It was founded in 2020 as DataCrunch and renamed Verda in November 2025. It rents NVIDIA GPU instances ranging from Tesla V100 and RTX A6000 up to H100, H200, B200, B300 and GB300. It also offers CPU instances, instant multi-node clusters with Slurm or Kubernetes, serverless containers, a token-priced inference API in preview, and storage. Storage covers NVMe block volumes, a shared filesystem, object storage in beta and a container registry. Confidential-computing variants of Blackwell GPUs and AMD EPYC CPUs are also available. Its data centres are in Finland and Iceland, and the Finnish sites are described as running on renewable energy. That makes it a good fit for EU-focused ML teams, startups and research groups that care about data residency. Pricing is pay-as-you-go by the hour. For example, an H100 SXM5 costs about $5.19 per hour on demand or $2.60 on spot, and a B200 costs $9.41 per hour. Reservations take 2% off for one month and 25% off for two years. Storage is $0.20 per GiB-month. You can automate it through a REST API, a Verda CLI, Python and Go SDKs, and a Terraform provider. It holds SOC 2 and GDPR claims and NVIDIA Preferred Partner status. Its main limitation is reliability. SemiAnalysis's ClusterMAX 2.0 rated it Bronze after customers reported hardware failures, site-level outages and, at the time of testing, no written SLA with penalties. Verda has since committed to downtime credits. Its footprint is also small outside the Nordics.
Our Rating
AI Software6.1/10
Hosting & Infrastructure5.7/10
Who Verda is best for
Best software for DevelopersGood fit
An API, CLI, SDKs and Terraform provider let ML engineers script their GPU infrastructure.
Self-serve, hourly GPU rental with spot pricing suits AI startups that train and serve models without long contracts.
Less suited
Serverless containers and an inference API support AI features in production, at mid-maturity.
Large GPUs and reserved clusters are available, but independent testing found reliability and RBAC gaps.
Pros & Cons
Pros
- Offers current NVIDIA Blackwell hardware, including B200, B300 and GB300, with SemiAnalysis confirming B200 availability.
- Spot instances cost roughly half the on-demand price, and two-year reservations cut prices by 25%.
- SemiAnalysis found its Slurm instant clusters more complete than many peers, with pyxis, enroot, NCCL and DCGM health checks.
- Data centres in Finland and Iceland, plus SOC 2 and GDPR claims, suit teams that need EU data residency.
- Full automation is available through a REST API, a CLI, Python and Go SDKs, and Terraform.
Cons
- SemiAnalysis reported hardware failures, sites going offline and, at the time, billing during outages. Verda later committed to downtime credits.
- There was no written SLA with penalties when ClusterMAX 2.0 tested it.
- There is no RBAC at the storage or Slurm level, and its Kubernetes offering is still maturing.
- Its regional footprint is limited to the Nordics, so latency is higher for users in the US and Asia.
- Very little public review data exists since the rebrand. RFP.wiki cites a Trustpilot score of 3.4 from only 5 reviews.
Key Features
Not included
- Managed WordPress
- Shared Hosting
Verda Pricing
Pay-as-you-go (on-demand)
- B200 SXM6 at $9.414/h
- H100 SXM5 at $5.189/h
- CPU instances from $0.048/h
- Hourly GPU and CPU instances
Reserved
- 8% off for 1 year
- 2% off for 1 month
- 25% off for 2 years
- Reserved clusters by custom quote
Spot
- H100 SXM5 at $2.595/h
- Interruptible capacity
- About 50% below on-demand
Frequently Asked Questions
Yes. DataCrunch, a Helsinki company founded in 2020, rebranded as Verda in late 2025, and datacrunch.io now redirects to verda.com.
In Finland and Iceland. Expansion into the UK, US and Asia was announced alongside its 2026 funding round.
Pay-as-you-go by the hour, with cheaper spot pricing and reserved discounts from 2% for one month up to 25% for two years. A third-party guide says usage is rounded up to 10-minute increments.
SemiAnalysis found no written SLA with penalties in late 2025. Verda has since committed to downtime credits of at least twice the instance cost.
Yes. Verda documents a Terraform provider, a CLI with an MCP server, a REST API, and Python and Go SDKs.
Yes. Instant clusters run Slurm or Kubernetes, and there is a managed Kubernetes page in the docs. Independent testers rated Slurm as strong and Kubernetes as still maturing.
Yes. Confidential versions of the B300, B200 and RTX PRO 6000 GPUs and AMD EPYC CPU instances are offered at a small premium.
Similar Tools
See all alternatives
Amazon's public cloud platform, with over 200 pay-as-you-go services for compute, storage, databases, networking and AI, used by startups through large enterprises and governments.

DigitalOcean is a developer-focused cloud platform offering Droplet VMs, managed Kubernetes, databases, object storage, App Platform and GPU compute for startups and small businesses.

Luxembourg-based edge and cloud provider offering CDN, virtual machines, bare metal, managed Kubernetes, object storage, AI GPU cloud, and DDoS/WAF security for businesses.
Ranked near Verda in AI Software
See Verda in the listVerda is #519 in AI Software by our score.




