Summary

• Helsinki-based Verda has raised $189 million in new financing led by Emergence Capital.
• The company intends to expand compute capacity while investing further in inference and cloud-platform services.
• Verda says its annualised revenue run rate reached $165 million in July, against more than $450 million raised through equity and debt.

Helsinki-based Verda has raised $189 million to expand its full-stack AI cloud, adding another large financing round to Europe’s race to build GPU infrastructure outside the traditional hyperscale market.

The financing includes an oversubscribed Series B led by Emergence Capital, with participation from MUFG Innovation Partners, Supermicro, Varma Mutual Pension Insurance Company, Lifeline Ventures, 6 Degrees Capital, byFounders, Tesi, and angel investors. Verda says the round takes its combined equity and debt funding above $450 million.

The company plans to use the capital across physical compute and cloud services, including inference infrastructure, and says it intends to multiply its compute capacity during the coming year. The financing therefore carries more infrastructure risk than a conventional software round because the business has to secure accelerators, data-centre capacity, networking, and power before every unit of that compute has necessarily been sold.

Verda also says it reached a $165 million annualised revenue run rate in July and employs around 250 people. Founded in Helsinki in 2020 and previously known as DataCrunch, the company now operates across Europe, the US, and Asia.

AI funding becomes infrastructure funding

The scale of the round reflects how capital-intensive the AI market has become. Training and serving larger models requires expensive accelerator hardware and dense data-centre infrastructure, while the rise of production inference means demand increasingly continues after a model has been built.

That has created an opening for specialist GPU-cloud providers seeking to offer capacity more quickly or flexibly than general-purpose cloud platforms. Their problem is that growth requires buying or reserving infrastructure before all of it has been contracted, leaving providers exposed to hardware cycles, financing costs, and uncertain future utilisation.

Europe has already seen several companies raise substantial sums around the same opportunity. Eindhoven-based Euclyd recently raised more than €200 million for an AI infrastructure architecture focused on power consumption and inference economics, while governments and established infrastructure groups are developing sovereign-compute capacity of their own.

Verda operates across several layers, from physical data-centre and hardware capacity through to cloud services and internal AI research. Its offer includes GPU instances, clusters, storage, and serverless inference products built around several generations of Nvidia accelerators.

Inference changes the capacity equation

Training frontier models consumes conspicuous amounts of compute, but the commercial life of a deployed model can involve far more cumulative processing as users repeatedly query it in production. Agentic applications can increase that load further because a single user request may trigger several model calls, tool invocations, and intermediate steps.

For infrastructure providers, sustained inference can generate more predictable usage than one-off training jobs, although production workloads also raise expectations around latency, reliability, geographic availability, and unit cost. Businesses running live services cannot treat compute merely as experimental capacity when an outage can interrupt an operational process.

European location adds another commercial argument. Data-residency requirements, latency, procurement policy, and sector-specific controls can make regional capacity attractive to organisations that do not want every workload routed through infrastructure outside Europe.

Regional capacity does not remove dependence on the underlying accelerator supply chain, however, where Nvidia remains dominant and hardware availability can still determine how quickly providers can grow.

Expansion brings execution risk

The financing gives Verda more room to reserve hardware and infrastructure, but it also illustrates the financial burden of competing in AI cloud services. High-end GPUs lose relative value as new generations arrive, while power availability and data-centre construction can constrain deployment even when hardware has been secured.

The competitive field is formidable. Amazon Web Services, Microsoft, Google, Oracle, CoreWeave, and specialist GPU providers are all expanding, which leaves European operators needing to differentiate through price, availability, sovereignty, technical performance, or service rather than geography alone.

Verda’s reported $165 million annualised revenue run rate gives the expansion a more substantial base than an infrastructure strategy built solely around anticipated demand. Annualised run rate is not the same as audited annual revenue, however, and current utilisation does not guarantee that every newly added cluster will be equally productive.

The capital is now in place to test whether a specialist European provider can turn expensive compute into a durable cloud business after the current scramble for capacity begins to resemble a more conventional infrastructure market.