Sovereign by construction
Hardware, energy and operator in the EU. Open-weight models, and your data stays in the EU.
Inference doesn't need gigafactories. It needs a grid of micro-sites.
AI is short of power. Micro-sites have it: farms, industrial buildings, renewable producers. We put the GPUs where the energy is and wire every site into one grid, without compromising on three things.
Renewable power capacity across the AI-Grid, from biogas, solar and wind producers.
Hardware, energy and operator in the EU. Open-weight models, and your data stays in the EU.
Racks sit where the power is: biogas, solar, wind, or an existing grid interconnection. Capacity does not wait years for new lines.
Modular shells we design and build for our own needs. A site grows by adding modules.
Six layers between a site's energy and your API call. We build and run all of them.
Model endpoints, billed per token
Energy-aware orchestration across sites
AMD and NVIDIA, chosen per workload
Modular shell, liquid cooled, built in house
Biogas, solar, wind or grid on a direct line
Micro-sites: farms, industrial buildings, grid-connected land
Distributed and highly resilient.
We focus on micro-sites with spare power: farms, industrial buildings, renewable producers or a plain grid connection. We start in France and Belgium, then Germany. By December 2029 the AI-Grid reaches 248 sites, all feeding one API. Our 2029 plan targets 85% renewable power.
Start with an endpoint. Move to dedicated capacity when your volume justifies it. Same grid underneath.
Open-weight models behind an API. You send prompts, we bill tokens.
AMD and NVIDIA GPUs on bare metal or Kubernetes, for fine-tuning, batch jobs and evaluation.
A full module dedicated to you, on a site you can visit, on monthly or yearly terms.
Jobs run where power is available, and GPU power follows each site's energy budget.
Failover between sites, load balancing across the grid, and jobs placed where energy is available and cheapest.
The cooling loop hands GPU heat to the site and its neighbours. Selling that heat pays for the cooling.

We design and build our modules ourselves, for our own needs. Each one holds a GPU rack, its cooling loop and its power backup, next to the producer.
Inference from EU micro-sites running on renewable energy. Capacity goes to the first signed customers: tell us your model, your monthly token volume and when you need it.
contact@cairnode.aiFarms, industrial buildings, renewable producers. If you have 200 kW or more of spare power and a little land, we install a module on your site, buy your energy and give the heat back.
See the benefits and apply →Log in or join the waiting list to start using sovereign EU tokens.
Access your endpoints, keys and usage.
The EU Token Factory is open to current customers only while we bring new capacity online. Join the waiting list and we'll invite you as soon as a seat opens.
New here? Join the waiting list
We install a GPU module on your site, buy your energy at a premium price and give the heat back. Your kWh become sovereign AI tokens sold across Europe.
Inference is eating every kilowatt Europe can produce, and new grid connections take years. Your site already has the power. We bring the GPUs.
We buy the power our GPUs use at a fixed premium above the market price. Cooling energy is paid at market price.
Your kWh become AI tokens sold to European companies. Every site you power adds capacity to Europe's first AI-Grid.
We buy, install, own and run the module. You host it and supply the power.
About 70% of the server heat comes back as ≈45 °C water, free, for digesters, drying, barns or buildings.
A premium price agreed up front, on a steady load that runs day and night.
One 20-ft module on a 7 × 7 m pad, next to your existing installations.
Move the sliders to see how much energy your site could sell to the AI-Grid in a year, and how much heat comes back.
1,300 MWh at the 2025 market price, before our premium
Bought by Cairnode at a premium price
1,300 MWh at €20/MWh, free if your return water meets our server inlet limits
1 GPU rack
Like the numbers? Send us your site.
Indicative, not an offer. Revenue uses the 2025 market price (€85/MWh) before the Cairnode premium; one GPU rack per 200 kW of spare power, using about 1,300 MWh a year. Heat valued at €20/MWh, the price of a village heat network. Price agreed site by site.
Two minutes: your site, your spare power, your contact.
We study your power, land and connection with you.
Energy price, heat and land use, agreed up front.
We deliver, connect and run the module.
Your energy powers tokens. You get paid for every kWh.
Three short steps. We review every site and get back to you to plan a site check.
A member of our sales team will reach out to you shortly for the next steps.
We use these details only to contact you about your site. Privacy
We pay a fixed premium above the market price for the power our GPUs use. The energy used for cooling is paid at market price. The heat is yours, free.
Cairnode buys, installs, owns and operates the module. You host it on your site and supply the power.
You can join from 200 kW of spare power, the size of a 250 kVA transformer. Typical AI-Grid sites are 200 kW, 500 kW or 1 MW.
About 70% of the server heat comes back as water at around 45 °C: enough for digester heating, drying, barns or buildings.
France and Belgium first, then Germany. The AI-Grid plan reaches 248 sites by December 2029.
A member of our sales team reaches out to you for the next steps. Applying does not commit you to anything.