French private equity giant Ardian is deploying serious capital into Nordic data centers through its Verne platform, positioning the region's cheap renewables and natural cooling as the structural answer to AI's energy problem.
The wires and watts underneath the AI boom have a geography, and increasingly that geography is Scandinavia and Iceland. Ardian, the Paris-based private equity firm managing over $170 billion in assets, has committed up to $1.2 billion to expand Verne, the green data center platform it acquired from Digital 9 Infrastructure, across Northern Europe. Bloomberg reported this month that the firm is planning to deploy over €3 billion into the region's digital infrastructure. The bet is straightforward: AI needs power, power is scarce and expensive in most of Europe, and the Nordics have both in abundance.
Verne already operates campuses in Iceland, Finland, Sweden, and Norway. The most concrete evidence of what Ardian is building sits in Keflavik, Iceland, where Verne has partnered with AI hyperscaler Nscale on a deployment of more than 4,600 Nvidia Blackwell Ultra GPUs. The first 7.5MW phase completed in roughly three months, with 64 liquid-cooled racks of GB300 GPUs installed. The entire campus runs on 100% renewable energy from geothermal and hydropower. In Finland, Verne broke ground on a 70MW data center campus in Mäntsälä, targeting high-density AI and machine learning workloads specifically.
None of this is accidental. Q1 2026 was the quarter, as analysts at Global Data Center Hub put it, when AI infrastructure stopped being constrained by capital and became constrained by power. The US is running into that wall hard. US construction spending on data centers eclipsed $50 billion as of June, per Bloomberg, and the grid in Northern Virginia, the world's largest data center cluster, is struggling to absorb the demand. Europe faces the same problem compounded: energy costs in Germany and the Netherlands are high, permitting is slow, and the regulatory environment adds friction that hyperscalers have little patience for.
The Nordic proposition is not subtle. Sweden's hydroelectric baseload produces spot prices of €40 to €60 per MWh with low volatility, according to market data compiled by researchers tracking the sector. Norway can theoretically cover 100% of its domestic data center industry's power needs through long-term contracts with local renewable producers. Iceland, sitting on geothermal reserves, offers some of the cheapest and most consistent green power on the planet. The ambient climate does the rest: you don't need to spend much on cooling when the outside air is doing half the job for free.
That combination has been pulling capital for a while, but 2026 has accelerated the pace noticeably. Microsoft, CoreWeave, and Brookfield collectively committed over $15 billion to Nordic AI infrastructure in 2025. Stargate Norway has plans for 230MW of capacity with an additional 290MW expansion targeted. A startup called Fossefall launched in Norway this year with a stated target of 500MW. The institutional money is no longer testing the thesis; it's executing on it.
Ardian's angle is slightly different from the hyperscalers, and that difference matters for how you should think about where the durable money in AI infrastructure is going. Hyperscalers like Microsoft or Amazon build data centers to run their own workloads and sell capacity on top. Ardian is a private equity firm: it built Verne into a platform, will lease capacity to whoever needs it (Nscale being the latest example), and collects long-duration infrastructure returns without touching a single model weight. It's a toll road play on AI, not an AI bet itself. The firm doesn't care which foundation model wins. It cares that training and inference require electricity and cooling, both of which it controls.
This is the structure European institutional capital is gravitating toward. Sovereign wealth funds and pension funds, the LP base for firms like Ardian, want exposure to the AI buildout but not the model risk. A 25-year data center lease backed by a hyperscaler or an AI cloud company is a bond with an infrastructure premium attached. The Nordics make that instrument cheaper to construct and more defensible to hold, because the energy inputs are renewable, contractually stable, and politically uncontroversial in a way that a gas-powered campus in Central Europe is not.
Ardian is also, it should be noted, not putting all its European chips in the north. The firm and Verne announced a €5 billion digital infrastructure campus in Île-de-France at the Choose France summit, targeting 500MW with an initial 200MW phase by 2030. France has its own advantages: nuclear baseload, proximity to financial centers, and an aggressive government courting AI investment. But the Nordic platform is where Ardian has the deepest existing position and the clearest cost advantage.
For founders and VCs watching the infrastructure layer, the practical takeaway is this: the cloud costs you pay to train and serve models are, in part, a function of where the underlying compute sits and what it costs to power. The firms now locking up Nordic capacity at scale are betting those geography-driven cost differentials will persist for decades. Given how long it takes to permit and build a 100MW data center campus versus how fast AI compute demand is growing, they're probably right.
Also read: Proception settles Tesla's trade secret lawsuit and closes an $11 million seed round on the same day • OpenAI scrambles to fix Codex as coding agent usage blows past its own capacity models • Asian AI rivals fill the gap left by Anthropic's export ban and match frontier performance for a fraction of the cost
