Research Finder
Find by Keyword
Is Finland the New Silicon Valley of European AI Infrastructure?
Nebius plans a 310 MW AI factory in Finland, targeting 3 GW global capacity by 2026.
4/01/2026
Key Highlights
- Nebius is scaling its Finnish footprint with a new 310 MW AI factory in Lappeenranta.
- The company is aggressively pursuing a global power target of 3 GW by the end of 2026.
- New infrastructure aims to deliver Nvidia Blackwell and Rubin GPU architectures to European builders.
- Sustainable design features include closed-loop liquid cooling and heat recovery for local districts.
The News
Nebius has announced plans to construct a massive 310 MW AI factory in Lappeenranta, Finland, with operations expected to begin in 2027. This development follows a recent expansion of their Mäntsälä facility and forms a central part of its strategy to secure 3 GW of contracted power globally by 2026. Find out more about the announcement here.
Analyst Take
We see a clear pattern emerging in the European cloud landscape where the race for compute supremacy is moving north. The announcement by Nebius regarding its Lappeenranta facility is not just about raw power; it represents a calculated bet on Finland as a primary hub for specialized AI infrastructure. While the hyperscalers have long dominated the general-purpose cloud, we believe that the rise of "AI factories" indicates a shift toward highly specialized, power-intensive environments that the traditional data center model was never quite architected to handle.
The scale here is significant. A 310 MW site is substantial by any metric, and when combined with their French and American projects, it suggests a company moving at a pace that few outside the Big Tech circle can match. This specialized focus is increasingly aligned with market demand; HyperFRAME Lens data reveals that 35% of enterprises now prioritize performance as their top criterion when selecting a vendor, a shift that directly benefits purpose-built AI clouds over legacy providers. Furthermore, this industrialized approach is a necessary response to the current "Execution Gap," where only 23% of AI/ML projects launched in the last year successfully reached production and met original ROI objectives, largely due to underlying infrastructure limitations.
What Was Announced
The Lappeenranta AI factory is designed to be a multi-building campus situated on a 100-acre industrial site. Technically, the facility aims to deliver capacity for the latest Nvidia Blackwell and Rubin generations of accelerated compute. Specifically, Nebius has indicated that while the Mäntsälä site houses the NVIDIA GB300 NVL72 platform, the new roadmap includes the NVIDIA Vera Rubin NVL72 platform starting later this year. The cooling infrastructure is architected to use a closed-loop liquid cooling system, which is intended to eliminate the need for local water intake. Furthermore, the system is designed to integrate heat recovery technology, which aims to divert excess thermal energy from the servers into the local district heating network, potentially reducing carbon emissions and local heating costs.
In our view, the technical specifications reflect a transition from air-cooled racks to sophisticated liquid-cooled environments. This is a necessity rather than a luxury when dealing with the thermal design power of next-generation GPUs. The decision to build in Lappeenranta also leverages the local talent pool from nearby universities, which we see as a move to secure the specialized operational expertise required to run such high-density clusters. It is a smart play. The European market has been crying out for sovereign, high-performance compute that sits outside the jurisdiction of US-based hyperscalers, and this Finnish expansion aims to fill that vacuum.
The broader strategy of Nebius appears to be focused on achieving massive scale quickly. Securing 3 GW of contracted power by 2026 is an ambitious target that places them in a direct competitive line with other specialized GPU cloud providers. However, the European focus gives them a distinct flavor. By embedding themselves in the local Finnish economy and utility grid, it's making it difficult for competitors to displace Nebius, although they hope so. The use of clean energy is a prerequisite in the current regulatory environment, but the heat recovery aspect is a particularly clever way to gain local political and social favor. It turns a waste product into a community asset.
We also observe that the move toward 310 MW in a single location suggests a belief in the continued growth of large-scale model training. While some in the industry are beginning to pivot their focus toward edge inference, Nebius is doubling down on the massive centralized factory model. This suggests they anticipate a long tail of demand for training foundational models and large-scale fine-tuning. The infrastructure they are building is architected to handle the massive interconnect speeds and power densities that these workloads require. It is a heavy-duty approach to a heavy-duty problem.
Looking Ahead
Based on what we are observing, the center of gravity for European AI is shifting toward regions that can offer both political stability and massive, green energy reserves. The Finnish expansion is a testament to this trend. The key challenge we are watching is how quickly Nebius can convert this massive power capacity into operational chips, especially since only 14% of enterprises currently classify their core data architecture as "fully modernized" for AI workloads, creating a massive market of organizations that must look to external AI factories to bridge their internal technical debt.
Our perspective is that Nebius is positioning itself as the primary alternative to American hyperscalers for European enterprises and startups that value geographic proximity and specialized AI stacks. Going forward, we will be closely monitoring how the company performs on its 2027 delivery timeline. According to HyperFRAME lens data 79% of organizations anticipate having multiple foundation models concurrently deployed, the need for the high-density, high-bandwidth environments Nebius is constructing will only increase, as multi-model architectures become the emerging enterprise standard.
HyperFRAME will be tracking how the company does in competing with other high-growth providers like CoreWeave and Lambda, who are also racing for global footprint. The competitive advantage may ultimately come down to the efficiency of the "AI factory" design and the ability to integrate into local energy ecosystems. Nebius has a head start in the Nordics, but the capital intensity of this race remains a significant factor to watch.
Steven Dickens | CEO HyperFRAME Research
Regarded as a luminary at the intersection of technology and business transformation, Steven Dickens is the CEO and Principal Analyst at HyperFRAME Research.
Ranked consistently among the Top 10 Analysts by AR Insights and a contributor to Forbes, Steven's expert perspectives are sought after by tier one media outlets such as The Wall Street Journal and CNBC, and he is a regular on TV networks including the Schwab Network and Bloomberg.