Verda raises $189M Series B to scale its AI cloud, becoming Europe's latest unicorn

Helsinki-based Verda, which builds cloud infrastructure for AI training and inference, has raised $189 million in an oversubscribed Series B. In an announcement published on 22 September 2026, the company described the round as making it Europe's latest unicorn and bringing its total funding, across equity and debt, to more than $450 million.

The round

Emergence Capital led the financing, with participation from MUFG Innovation Partners, server vendor Supermicro, Finnish pension insurer Varma, and Nordic investors Lifeline Ventures, 6 Degrees Capital, byFounders and Tesi (Finnish Industry Investment Ltd). Angel investors in the round include Ola Tørudbakken and Mark Saroufim.

Supermicro's involvement is worth noting: Verda designs and operates its own data centres, and a hardware supplier taking an equity position ties the startup's funding base to the equipment pipeline it needs to keep expanding.

Founder and CEO Ruben Bryon framed the moment bluntly: "There's a window right now to build one of the defining compute companies of this generation, and to do so from Europe. It won't be open for long."

What Verda is building

Verda describes a full-stack operation: it designs and builds its own data centres, the infrastructure inside them and the software platform that runs on top, and says it delivers capacity with a lower carbon footprint than the industry average.

The underlying thesis is that AI workloads have outgrown traditional capacity procurement. Training runs that once required months of advance planning now need infrastructure that already exists, production inference cannot compromise on latency or availability, and agentic applications add bursty demand as context accumulates across many turns. Decisions about how much compute to buy, the company argues, increasingly sit with individual engineers rather than procurement teams working annual cycles.

A core differentiator is the AI Lab, an internal team that develops compilers, serving software and kernel-level optimizations, and runs genuine research workloads on the same platform customers use. Findings on GPU utilization and inference efficiency feed back into the product roadmap.

Customers and revenue

The performance figures all come from the company itself. Verda says it reached a $165 million annualized revenue run rate in July 2026, employs more than 250 people across Helsinki, London, Taipei and San Francisco, and serves organizations in over 50 countries.

Named customers include Aleph Alpha, which runs its R&D infrastructure on Verda GPU clusters with joint engineering work on the underlying software stack; Magnific, which serves media generation at millions of requests per day; and Epsilon Health, which trains radiology AI models on a dedicated cluster at native image resolution.

Arjun Karpur, head of machine learning at Epsilon Health, said in the announcement: "What we didn't expect was a team of engineers who could help us design our own data streaming and cluster management, not just hand us a cluster and walk away."

Where the money goes

On capacity, Verda says it will have more than 250 MW of operations in 2027. Data centre capacity is live in Finland today, with additional sites planned across Europe, the UK, the US and Asia, and the company anticipates early deployments of NVIDIA's VR200 NVL72 systems in the coming months.

On the platform side, a container registry recently shipped and S3-compatible object storage is next on the roadmap. Verda's Instant clusters can now be deployed with Kubernetes scheduling through Kueue or Slurm pre-configured via Slinky, with a managed Kubernetes service to follow.

For enterprise buyers, the company points to newly added audit logs and single sign-on for identity and access management, confidential computing support for eight-GPU NVIDIA HGX B300 and B200 configurations, and SOC Type II and C5 certifications. The raise also funds inference services tuned to how models are actually used, faster provisioning of instances, clusters and storage, and expanded co-research between the AI Lab and outside labs, open-source projects and developer tool vendors.

Why it matters

This is one of the larger European growth rounds of the year, and a signal that investors still see room for new entrants in AI compute alongside the US hyperscalers. Verda is also an explicitly European play: a Finnish operator with its own facilities selling into a market where most capacity sits with American providers, and where AI teams are under pressure to keep data and workloads within the region.

The pairing of capital-intensive data centres with an in-house software and optimization layer, rather than reselling raw GPU time, is the company's bet that performance engineering decides who wins AI workloads, not just megawatts. The usual caveat applies: the revenue figures, customer claims and the unicorn framing all come from a single company announcement, with no valuation disclosed and no independent confirmation of the numbers.