Verda

What Is Verda?

For most of its history, this company was called DataCrunch — a Helsinki-founded GPU cloud provider that quietly built a reputation among ML engineers for cheap H100 access without the AWS markup. Ruben Bryon founded it in 2020, and for five years it stayed a well-regarded but relatively niche player, the kind of provider you found through a Hacker News thread rather than a marketing budget. On 17 November 2025, that changed: DataCrunch became Verda, migrated to verda.com, and repositioned itself from a GPU rental service into what it calls a "full-stack AI cloud."

The rebrand landed alongside a $64 million Series A in September 2025, led by byFounders with Skaala, Varma, and Tesi, plus debt financing from Nordea, Danske Bank, and other Nordic lenders. Seven months later, in April 2026, Verda raised a further $117 million led by Lifeline Ventures, pushing total funding past $180 million across the two disclosed rounds. Its annualised revenue run rate reportedly doubled past $60 million in the same quarter. Researchers should expect to find both names scattered across search results, comparison sites, and older technical blog posts. Verda and DataCrunch are the same entity, and the old name isn't disappearing from the internet's memory any time soon.

What hasn't changed is the underlying infrastructure story. Verda owns and operates its own data centres in Finland and Iceland rather than reselling colocation capacity, and runs them entirely on renewable power. It has also built a Python SDK, a Terraform provider, and a CLI that infrastructure teams actually reach for. Today it sits alongside CoreWeave and Lambda as one of the specialist GPU clouds competing for AI training and inference workloads outside the big three hyperscalers.

Key Features

GPU instance range

Verda offers 15 GPU types across 55 configurations, spanning the NVIDIA A100 up through the newest B200, B300, and GB300 SXM6 accelerators. On-demand pricing starts at $1.79/hour for A100 80GB. It reaches $8.62/hour for GB300, with spot instances available at roughly half the on-demand rate across the range. For teams training frontier-scale models, that pricing sits well below AWS or Azure list prices for comparable NVIDIA hardware, and is broadly competitive with CoreWeave's published rates.

Serverless inference and storage

Beyond raw GPU rental, Verda runs auto-scaling serverless containers for inference workloads, priced from $0.06/hour for CPU-only AMD EPYC instances. Prices scale up to $8.25/hour for B300-backed containers. Storage — NVMe block storage, a shared filesystem, and a container registry — is priced uniformly at $0.20 per GiB per month across all three types. That flat rate simplifies cost forecasting compared to providers that tier storage pricing by IOPS or throughput class.

Infrastructure-as-code tooling

This is where Verda separates itself from many boutique GPU clouds. An official Python SDK, a Terraform and OpenTofu provider, and a CLI let infrastructure teams manage clusters the same way they manage any other cloud resource, rather than clicking through a dashboard. Multi-node training runs can be orchestrated via Slurm or Kubernetes, which matters for teams running distributed training jobs across dozens of nodes rather than single-instance experiments.

Renewable power and waste heat

Verda's data centres in Finland and Iceland run entirely on renewable energy, and the company has built a district heating programme that repurposes waste heat from its facilities to warm thousands of nearby homes. For organisations under increasing pressure to report AI infrastructure emissions, this is a genuine differentiator against hyperscaler regions still reliant on grid mixes with substantial fossil generation.

EU Gigafactory ambitions

Verda has submitted a proposal, with backing from the Republic of Latvia, to host one of the European Commission's AI Gigafactories — dedicated facilities designed to support training and inference at up to 100,000 accelerators. Whether that bid succeeds is out of Verda's hands, but it signals the company's ambition to move from a mid-sized specialist GPU provider toward genuine hyperscaler-adjacent status within Europe.

Pricing

Verda prices in USD across on-demand, spot, and reserved tiers rather than fixed monthly plans, which is standard for GPU cloud infrastructure. On-demand H100 SXM5 runs $3.25/hour. The same instance on spot pricing drops to $1.63/hour but can be reclaimed with short notice, so it suits fault-tolerant training jobs rather than latency-sensitive inference. Reserved instances with a 2-year commitment cut up to 25% off on-demand rates, which is meaningful for teams with predictable, sustained GPU demand.

There's no published free trial or credits programme visible on the current pricing page, which puts Verda behind some competitors that offer trial credits to lower the barrier for first-time evaluation. Enterprise customers negotiating multi-node clusters or dedicated capacity work through a custom quote process rather than self-serve checkout.

EU Compliance and Privacy

Verda is headquartered in Helsinki, Finland, and its GPU data centres sit in Finland and Iceland — both squarely inside the EU/EEA data protection framework. That keeps default data processing under Finnish and EU jurisdiction. It's a meaningful distinction from CoreWeave and Lambda, both US companies whose infrastructure and legal entity sit under US law regardless of which region a customer selects.

The company's own funding announcements increasingly describe expansion into the UK and US markets alongside its EU sovereignty pitch, including new offices in London and San Francisco planned for 2026. That's a reasonable growth strategy for a scaling infrastructure company. EU buyers who chose Verda specifically for EU-only positioning should still watch how that expansion affects where new capacity gets built and which entity ultimately processes their workloads.

Who It's Best For

ML engineering teams training or fine-tuning large models who want frontier-tier NVIDIA hardware without hyperscaler pricing. If your workload tolerates spot interruption, Verda's spot pricing is genuinely competitive.

European organisations prioritising data sovereignty for AI infrastructure. Choose Verda over CoreWeave or Lambda if keeping GPU workloads under EU jurisdiction and on renewable Nordic power matters as much as raw price.

Infrastructure teams that live in Terraform and Kubernetes will find Verda's tooling genuinely usable rather than an afterthought bolted onto a web dashboard.

Teams needing an established enterprise SLA track record should ask pointed questions first. Verda's uptime claims aren't yet backed by a long, independently audited public history at hyperscaler scale.

The Verdict

Verda earns its place on the strength of what DataCrunch built quietly for five years: real, owned infrastructure at genuinely competitive prices, backed by infrastructure-as-code tooling that ML platform teams actually want. The rebrand and the fresh $117 million round signal real momentum, and the renewable-power, EU-jurisdiction story gives it a distinct pitch against CoreWeave. What buyers are trading for that price and sovereignty advantage is a thinner managed-services layer than the hyperscalers offer and a support organisation still scaling alongside a headcount that only recently passed 165. Teams whose priority is GPU price-performance under EU law will find that trade easy to make. Buyers wanting AWS-style breadth of managed services will find it a real gap. Verda sits well alongside EU infrastructure options like Scaleway, OVHcloud, and Hetzner in the wider cloud hosting category.

Frequently Asked Questions

Is Verda the same company as DataCrunch?

Yes. DataCrunch rebranded to Verda on 17 November 2025 and migrated its site to verda.com. It's the same Helsinki-founded, Finland- and Iceland-based GPU cloud provider under a new name and a broader AI-cloud positioning.

Is Verda GDPR compliant and EU-hosted?

Verda is headquartered in Helsinki, Finland, and its GPU data centres are located in Finland and Iceland, keeping default data processing under EU jurisdiction. The company is also bidding, with the Republic of Latvia, to host a European Commission AI Gigafactory.

How much do Verda's GPUs cost?

On-demand pricing starts around $1.79/hour for A100 80GB. It runs up to $8.62/hour for the newest GB300 SXM6. Spot instances cut those prices roughly in half, and 2-year reserved instances save up to 25% versus on-demand.

How does Verda compare to CoreWeave?

CoreWeave is a US-based GPU cloud that scaled through Microsoft and OpenAI capacity deals. By comparison, Verda is a smaller, Finland-headquartered alternative built on owned, renewable-powered Nordic data centres, with published on-demand prices competitive against CoreWeave's for equivalent NVIDIA hardware.

What funding has Verda raised?

A $64 million Series A landed in September 2025, led by byFounders, Skaala, Varma, and Tesi, followed by a $117 million round in April 2026 led by Lifeline Ventures. Total funding now stands at more than $180 million across the two disclosed rounds.