Verda Raises $189M Series B to Expand European AI Neocloud

Helsinki-based neocloud provider Verda Cloud Oy, formerly known as DataCrunch, has secured $189 million in early-stage Series B funding to expand its dedicated artificial intelligence compute infrastructure. Led by Emergence Capital with participation from Supermicro, MUFG Innovation Partners, and Finnish sovereign investor Tesi, the transaction elevates Verda’s valuation above $1 billion, establishing the European neocloud as the continent’s newest infrastructure unicorn. The capital injection brings Verda’s total equity and debt financing to more than $450 million as it scales operations across Europe, the United States, and Asia.

Capital injection and European unicorn valuation

Founded in 2020 by Chief Executive Officer Ruben Byron, Verda has experienced sharp commercial growth amid persistent global demand for dedicated GPU clusters. According to Verda’s company disclosures, the firm reached an annualized revenue run rate of $165 million in July 2026. The company now employs more than 250 personnel distributed across operating hubs in Helsinki, London, Taipei, and San Francisco.

The oversubscribed Series B round drew substantial institutional and strategic backing. Investors include Varma Mutual Pension Insurance Company, Lifeline Ventures, 6 Degrees Capital, and byFounders, alongside prominent angels such as Ola Tørudbakken, director of AI systems at Meta Platforms, and Mark Saroufim, co-founder of Core Automation. As reported by SiliconANGLE, Verda operates live data center capacity in Finland and plans to scale beyond 250 megawatts of capacity in 2027.

Verda operates within the emerging “neocloud” segment—specialized infrastructure providers engineered exclusively for accelerated computing. Unlike legacy public clouds whose software layers accommodate diverse general-purpose web services, neoclouds configure physical facilities, networking fabrics, and virtualization stacks solely around intensive model training and high-throughput inference.

Full-stack optimization for agentic and multi-turn inference

The architectural requirements of production AI have evolved beyond isolated batch generation. Enterprise workloads increasingly involve multi-turn conversational agents, long-context reasoning chains, and persistent execution loops. These operational profiles introduce distinct platform challenges: dynamic memory fragmentation, prolonged cluster reservations, and latency spikes across sequential tool invocations.

To address these constraints, Verda pairs bare-metal accelerator provisioning with dedicated systems engineering. The company maintains an internal research lab that executes real-world research workloads directly on its platform. Rather than relying entirely on generic upstream libraries, Verda develops custom compilers, specialized serving software, and low-level kernel optimizations. These software modules improve GPU utilization rates, accelerate batch scheduling, and minimize inference latency when handling complex model architectures.

This co-design methodology mirrors wider enterprise shifts toward purpose-built hardware and monitoring stacks, such as enterprise AI infrastructure architectures that integrate compute and telemetry. By working directly with open-source project maintainers and developer tooling teams, Verda aims to help platform engineers extract maximum sustained performance from high-density silicon.

Enterprise adoption across healthcare, media, and sovereign AI

Verda reports active workload deployments across more than 50 countries, serving startups, sovereign initiatives, and enterprise institutions. Its customer portfolio illustrates how organizations navigate cloud infrastructure choices when building domain-specific models:

• Sovereign foundation models: German sovereign AI developer Aleph Alpha utilizes Verda clusters to power core research and development infrastructure, aligning with European data governance priorities.
• High-volume digital media: Visual content platform Magnific runs managed inference on Verda to sustain millions of daily image generation requests, demanding tight uptime SLAs and consistent serving latency.
• Clinical medical imaging: Healthcare technology firm Epsilon Health trains bespoke radiology models on dedicated Verda clusters. By collaborating with Verda’s systems engineers on custom data streaming pipelines and cluster management, Epsilon processes radiology scans at native image resolution rather than downsampling to fit standard hardware limits.

While specialized neoclouds offer granular performance optimizations, enterprise buyers face structural trade-offs. Moving datasets between hyperscale object storage and specialized GPU environments can incur network transfer costs and latency overhead. Furthermore, leveraging custom compiler features and proprietary serving configurations requires engineering teams to balance throughput gains against potential lock-in should workloads require relocation.

What happens next

Verda plans to direct its Series B capital toward expanding international facility footprint and accelerating hardware rollouts. The company is preparing early deployments of Nvidia VR200 NVL72 rack-scale supercomputing clusters in the coming months, providing the interconnect bandwidth and memory density necessary for next-generation frontier models.

Geographic diversification will also accelerate. Beyond its existing installations in Finland, Verda is preparing compute capacity across the United Kingdom, continental Europe, the United States, and Asia to support regional data sovereignty requirements. As European technology policymakers advocate for domestic infrastructure independence, Verda’s emergence as an infrastructure unicorn positions the Helsinki outfit as a significant regional competitor against legacy cloud providers.