Heat that would be wasted, put to work.
Data centres spend money getting rid of heat; direct air capture plants spend money getting hold of it. Nordvar designs one site that does both jobs at once, built in phases: the data centre first, capture as its own permitted phase, alongside or after it depending on the financing of each phase. The five steps below show how, and the numbers that follow show what it saves.
Compute
Tenants run their workloads on servers in a compact, modular facility. All of the electricity those servers consume leaves them as heat; a server does no mechanical work, so there is nowhere else for the energy to go.
Collect the heat
The servers are liquid-cooled. Cold plates on the chips and heat exchangers on the rack doors together deliver about 85% of the rack's heat into a closed water loop; the rest escapes to the room air. It is the design choice that makes everything downstream possible - and it cannot be added later.
Raise the temperature
Server heat is warm, not hot. A heat pump lifts the water to around 100 °C, the temperature the capture process needs. The lift costs some power, which is why the site is designed around renewable electricity.
Remove carbon
A direct air capture unit draws air through a material that binds carbon dioxide. Heated to around 100 °C, the material releases it as a concentrated stream. This is where the data centre's heat does its second job.
Store it permanently
The concentrated carbon dioxide is dissolved and injected into basalt, where it reacts with the rock and becomes stone. Once mineralised, it does not return to the atmosphere. Storage is the default route. Where producers of synthetic fuels or carbon-based products exist near a site, captured CO₂ can also be supplied to them. Only stored tonnes are counted as removal; supplied tonnes are reported separately.
- AC-EAir-cooled exhaust - too cool to use
- LC-WLiquid-cooled water - warm enough to work with
- CNWhat capture needs - the heat pump bridges the gap
Energy, in numbers.
Every kilowatt-hour of electricity a server uses ends up as heat, and most data centres pay to get rid of it. Not all of it can be caught: liquid cooling collects the heat from the chips, and rear-door exchangers catch most of the rest, but some still escapes to the room air. A heat pump then raises the water in the loop to the temperature the capture process needs. The figures below follow one rack through a year.
What is saved
The saving is the heat that no longer has to be made from scratch. Delivering the same 112,500 kWh of process heat without server heat would take about 112,500 kWh of electricity with electric heating, or roughly 56,000 kWh with a heat pump drawing on cold outdoor air, because lifting heat from a few degrees is much harder than lifting it from fifty. Against that fairer comparison, one 10 kW rack saves in the order of 18,000 kWh of electricity a year; against electric heating, about 75,000 kWh.
The gain
The fair measure is electricity per tonne of CO₂ captured. The comparison below runs the same capture unit on different heat sources, using 2 MWh of heat and 0.5 MWh of capture electricity per tonne.
Reusing server heat cuts the electricity needed per tonne by about 20% compared with the same unit on a conventional heat pump, and by about half compared with electric heating. The server heat itself is energy the tenant has already paid for; the capture process gets it for the cost of upgrading it. Where geothermal heat is piped in directly, it uses less electricity still, which is why a Nordvar site in a geothermal region uses both.
Where the larger gains are
- One site, two products
- The plot, the grid connection, the power contract, the fibre and the operations are paid for once and serve both the data centre and the capture unit.
- The heat is already there
- A standalone capture plant has to find heat, a data centre has to get rid of it. Putting them together removes a problem from each.
- The customer pays for the heat
- The tenant's electricity bill covers the server heat, so the removal is financed partly by compute revenue rather than entirely by credit sales.
- Attribution is the product
- Removal tied to a tenant's own energy use is something neither a standalone capture plant nor an ordinary data centre can offer.
One caveat: it is not lower-energy than capture running on geothermal heat. The energy gain of the method is real and modest; the gains above are what make a small site possible at all.
One loop, two jobs.
The water that carries heat away from the servers is the same water that feeds the heat pump. It leaves the racks at 45 to 60 °C, gives up its heat to the heat pump, and returns to the racks cooled to the temperature they need. When capture is running, the heat pump is therefore doing the data centre's cooling as a by-product, and the dry cooler that would otherwise reject the heat to the outside air can rest.
The saving on the data centre side is the dry cooler's fan energy: typically 1 to 3% of the heat rejected. It is small, and we say so. It is also a further reason the two systems are designed as one loop rather than two: the compressor runs for the capture side, and the cooling comes free.
- The dry cooler stays
- Sized for the full load: capture will pause for maintenance, sorbent cycles or grid curtailment, and the servers must be cooled in those hours.
- Control follows the servers
- The heat pump is controlled on the temperature of the water returning to the servers, the dry cooler trims, and a buffer tank between them smooths the difference.
- The balance flexes
- When the capture unit needs more heat than the racks produce, the heat pump draws the balance from outside air or, where available, geothermal water; when it needs less, the dry cooler rejects the surplus.
Scaling
The arithmetic is linear. A 50 kW module supports roughly 170 to 375 tonnes a year on its recovered heat alone; a 1 MW site roughly 3,400 to 7,500 tonnes. These are heat-side figures. The capture unit and the heat pump draw electricity of their own, so a site's total power use rises when capture is added; what falls is the energy spent making heat that computing had already produced.
These are illustrative figures from published ranges and labelled estimates, not measurements. Nordvar will publish measured values once a site is operating.
Where the figures come from.
- 01Deutz, S. and Bardow, A. (2021). Life-cycle assessment of an industrial direct air capture process based on temperature–vacuum swing adsorption. Nature Energy 6, 203–213.DAC energy demand - measured and 2030 target
- 02International Energy Agency (2022). Direct Air Capture 2022: A key technology for net zero. IEA, Paris.Commercial DAC energy ranges
- 03Climeworks. Direct air capture technology: innovations in CO₂ removal. Official technology pages.Solid-sorbent capture, release at about 100 °C, mineral storage - operating in Iceland
- 04Jiang, J., Hu, B., Wang, R.Z. et al. (2023). High-temperature heat pump prototype, 50 °C source to 100 °C supply. Energy 263, Part D.Heat pump COP at the server-heat lift
- 05TLK Energy. Industrial heat pump guide.COP against temperature lift
- 06European Environment Agency (2024). Greenhouse gas emissions per capita, EU-27, 1990–2022.EU per-capita emissions, 2022
- 07Eurostat (2025). Greenhouse gas footprint per capita, EU, 2023. Dataset cli_gge_foot.EU consumption footprint, 2023
- 08Terlouw, T., Pokras, D., Becattini, V. and Mazzotti, M. (2024). Assessment of Potential and Techno-Economic Performance of Solid Sorbent Direct Air Capture with CO₂ Storage in Europe. Environmental Science & Technology.DAC potential and cost in Europe
- 09Schneider Electric / DCD (2024). Direct Liquid Cooling System Challenges in Data Centers. White Paper 210.Heat capture ratio - share to liquid vs air
- 10Reviews of absorption and compression heat pumps for industrial heating. City University of Hong Kong; Heat Pumping Technologies.Heat pumps at large temperature lifts
Designed as one system
Each part of the chain exists today. Nordvar does not build the capture unit or the storage; both come from established operators - the capture technology itself was developed by a prominent company and is already operating at commercial scale. What Nordvar adds is the decision to design a data centre from the start as the heat source for carbon removal. A facility built this way can operate as a data centre first and add capture as a second phase, without rebuilding.
What it is not
Nordvar does not offset. The carbon removed is measured, verified and stored permanently, and it is counted once. CO₂ supplied for reuse is not counted as removal. The data centre itself runs on renewable electricity; the capture process removes carbon from the atmosphere in addition to that, it does not cancel emissions from elsewhere.