FAQ
Questions, answered
At what point does air cooling stop being enough?
Around 20-30 kW per rack air starts to struggle, and roughly 40 kW is the practical ceiling — past that the airflow required becomes unworkable. For scale, a single Nvidia B200 draws about 1 kW, and a GB200 NVL72 rack pulls 120-130 kW, which is why new AI builds are liquid from day one. Standard air-cooled ASICs sit well below that line, which is exactly why air remains the cheaper answer for most mining fleets.
Liquid, immersion or air — how do I choose?
Below about 30 kW per rack, air is cheaper and simpler, and we will say so. Between roughly 35 and 100 kW, direct-to-chip liquid — hydro racks, cold plates, coolant distribution units — is the standard answer. Above 100 kW per rack, or where noise and floor area dominate the decision, immersion wins. Tell us your per-rack load and site constraints and we will point you at the cheapest option that actually works.
Can cooling hardware sold for mining be used with AI servers?
The infrastructure itself is workload-agnostic: a coolant loop, a manifold, a rack frame or a container does not care what generates the heat, and the price per kilowatt of dissipation in this market sits well below purpose-built datacenter equipment. Direct-to-chip liquid cooling of GPU nodes is well supported, and our hydro racks and distribution units serve it directly. Immersion of GPU servers is a different question — see below.
Does immersion cooling void the hardware warranty?
Often, yes, and this is worth checking before you spend anything. Submerging a standard air-cooled ASIC voids the Bitmain warranty; purpose-built immersion models keep coverage only in a manufacturer-approved fluid. On the AI side, Nvidia does not endorse immersion for Blackwell-generation GPUs, and most CPU and GPU warranties lapse once a part has been submerged. We will help you read the terms for your exact model — and we will tell you if the answer is no.
What coolant does a hydro system need, and how is it maintained?
Manufacturers specify deionised water or an approved inhibited antifreeze, typically held at pH 8.5-9.5, with flow around 8-10 L/min per unit, pressure at or below 3.5 bar and inlet temperature between 20 °C and 50 °C. Untreated water is the single most common cause of failure: it corrodes cold plates and blocks the microchannels they depend on. Budget for scheduled fluid checks and replacement — it costs a fraction of replacing hardware.
What does a site need before a hydro rack arrives?
Three things. A three-phase electrical feed sized to the load; somewhere to reject the heat, which below roughly 100 kW is usually a dry cooler rather than a chiller; and a floor rated for the weight of the system once the loop is filled. We size all three against your actual layout before you order and tell you plainly what has to change.
How quickly can a container be deployed?
Once power and network are live at the site, commissioning is measured in days rather than months, because cooling, distribution and monitoring are built in and tested before the unit ships. The long pole is almost never the container itself — it is the grid connection, the groundwork and the permits, and those should start well before you order.
Do I need permits?
Usually yes, and requirements vary sharply by country and often by municipality: building and electrical permits, environmental assessment, zoning approval. Noise limits are frequently the most restrictive constraint of all, and on-site gas generation adds emissions permitting on top. We will explain what generally applies in your jurisdiction, but the local authority has the final word — engage them early rather than after delivery.
Is on-site gas generation actually cheaper than the grid?
Where the gas is stranded or currently being flared, electricity typically lands at $0.02-0.05 per kWh against industrial grid rates that are usually well above that, with maintenance adding roughly $0.02 per kWh. But the stronger argument is often time rather than cost: grid connection queues across much of Europe now run to years, and generating behind the meter removes that queue entirely.
How loud is this equipment?
Air-cooled deployments are loud, and noise is what triggers most complaints and ordinance problems. Liquid cooling removes the fan wall and drops ambient noise sharply; immersion is close to silent, with only the pumps audible. If your site has neighbours or a noise limit, that difference frequently decides the technology on its own, ahead of any thermal argument.
Can the waste heat be reused?
Yes, and liquid is what makes it practical. A hydro loop returns water warm enough to feed building heating, greenhouses, drying processes or a district heating network, because the heat is captured in a fluid instead of being dispersed into a hall. Air-cooled sites can rarely do anything useful with it. Whether it pays depends entirely on having a heat consumer close by.
Do you only supply hardware, or design the deployment?
Both. We size cooling, power and network against your actual load profile, supply the hardware from EU stock with EU warranty, and support it afterwards. If you would rather not run the site at all, we host it across partner facilities in 9+ locations.
