Night aerial view of the Dutch glasshouse belt glowing under a dark sky

Energy-producing AI data centers

Co-generating power, flexible compute, and heat.

As energy makers, we access power and permits in Europe's most constrained metro regions, where other data centers can't.

Partners & supporters

NVIDIAVertivSchneider ElectricArmadaRaffayEfficientEcoDigitalJohnson ControlsFrigelUnica DatacentersRoyal HaskoningDHVJenbacher

What we do

One platform. Three outputs.

Every Digital Energy Center produces AI capacity, heat and power for the place it stands.

AI

compute capacity
  • Colocation capacity for AI, built to your requirements
  • Single tenant: one customer per site
  • Sites of 5-20 MW, interconnected into clusters
  • Liquid-cooled, NVIDIA NCP reference architecture
Secure AI capacity →

Heat

refined, usable heat
  • Baseload heat, delivered year-round
  • Long-term supply contracts
  • Greenhouses, industry and district heating
Get heat →

Power

flexible power
  • Flexibility that relieves a congested grid
  • Peak power when the network needs it
  • Emergency power from on-site generation
Talk to us →

Why we exist

From energy taker to energy maker.

Building the neo-utility for energy and compute.

AI compute can be the fuel: energy built from AI, and for it. To get capacity and scale at speed, AI has to become an energy producer. That means it cannot be an extractive facility that exports tokens. It has to produce local value: power for the grid, heat for the community, capacity for the tenant. When all three win, capacity comes fast.

A Digital Energy Center is not extractive. It produces energy, and it strengthens the place it stands.

For the grid

Flexible, on-site power that relieves constrained networks instead of adding load to them.

For the community

More affordable, cleaner heat for greenhouses, industry and homes, replacing gas.

For Europe

Local AI infrastructure for research, schools, governments and startups, so Europe can compete.

Why Digital Energy

Speed. Scale. Purpose-built.

What an AI buyer or an energy consumer gets from us that a conventional data center cannot give.

01

Speed

Live grid connections and permits already in place. RFS as fast as 30 weeks from financial close, indicative, where others queue for years.

02

Scale

Start at one site, grow across a campus and a cluster, on one platform that scales with your demand.

03

Purpose-built

Designed for AI from day one: liquid-cooled, highest density, power-flexible, deployable in prefabricated modules, and an asset to the grid, not a load on it.

The platform

Distributed sites, not one mega-campus.

Sites of 5-25 MW, interconnected into clusters. No single dependency, no giant commitment: start small, scale in steps, on a platform that grows with you instead of a site you outgrow.

Interconnecting 5 to 10 adjacent sites, each with its own grid connection, is how you reach hundreds of megawatts in Europe at AI speed. The Andijk Campus is the first: three sites, one campus, each with its own heat user.

The vision: 300 distributed nodes by 2030, powering and generating 5 GW across Europe, interconnected into one inference network that runs each workload where it is best served.

Map of the Netherlands with unlabeled site markers across the Amsterdam metro region
18sites secured
300 MWin development
600 MWunderlying energy infrastructure run by our partners today, to be made multipurpose
5 GWidentified today

The product

Build-to-suit AI Factories.

Optimized for NVL72 and beyond. Liquid-cooled native.

  • N+1, Tier 3 equivalent design: 99.99% SLA, power, shell and cooling to your racks
  • Liquid-cooled from day one on the NVIDIA NCP reference architecture
  • Tailor-built: your whitespace, your facility design
  • Enterprise-grade security: 5 layers, 1 entry
  • Turnkey setup and build-out with the partners who build for hyperscalers and leading neoclouds
  • Managed services and smart hands, if wanted

Full technical specification in the site dossier, under NDA.

40-600 kW rack densities99.99% SLAN+1, Tier 3 equivalent designDirect liquid coolingAs fast as 30 weeks from financial close, indicative
Row of NVIDIA Vera Rubin class GPU racks: the systems our liquid-cooled halls are built for

Built for your AI, not for everyone's compute.

A colocation hall built for every kind of compute is optimal for none. We build each AI Factory to your requirements: liquid-cooled, highest density, the environment GPUs excel in.

More tokens from less energy.

PUE 1.03, measured at Kirkenes. Every watt drawn also heats a greenhouse and steadies the grid.

Infrastructure for inference

Built for inference, close to the people it serves.

Training happens in a few giant campuses. Inference happens wherever people are. We build for the second: a network of sites in the metro regions where users, data, connectivity and talent already are.

European metros first

Amsterdam first: 350 million+ people within 10 ms, over half of Europe's GDP. Then Europe's other metros, one platform.

Then the global hubs

The same model in the world's other hubs, with partners already building in North America and France.

At the edge, for the community

Inference next to the people who use it, in facilities that also heat homes and greenhouses and steady the grid. One watt, three jobs.

On-prem, without the burden of on-prem

Your own dedicated facility, your hardware, your access. Neutral and yours, without building or running it yourself.

AMS-IX: Europe's #2 exchange, one of the three busiest on Earth, plus NL-IX.

Recognition

What the industry says.

“Digital Energy demonstrates how AI infrastructure can contribute beyond compute, by reusing heat, supporting food production and delivering sovereign AI capacity.”
Michael Winterson, Secretary General, European Data Centre Association

In the news and on stage

Next step

Power intelligence, fast.

Tell us what you're building. We'll tell you how fast we can power it.

jonathan@digital-energy.group