Liquid Cooling for Bitcoin Mining: Why Air Is Becoming Yesterday's Technology
Air-cooled ASIC mining was fine when rack density was low and power was cheap. In 2026, with chips pushing 30kW per unit and AI data centers competing for the same infrastructure, the industry is moving to liquid cooling fast. Here is why and what the options actually look like.

Liquid Cooling for Bitcoin Mining: Why Air Is Becoming Yesterday's Technology
Walk into a large Bitcoin mining facility built before 2022 and what you will find is rows of ASIC miners blowing hot air through industrial fans into ductwork and out through vented walls. It works. It is cheap to set up. And it is increasingly inadequate for where the hardware is going.
The newest generation of mining ASICs — Bitmain's S21 XP Hydro at 5.7kW per unit, MicroBT's M66S class — see high-density rack configurations regularly pushing 20-30kW per rack, and full rows of hydro-cooled machines can exceed 50kW in a single bay. Air cooling those densities requires enormous airflow, significant facility footprint, and constant fan maintenance. More importantly, air cooling leaves performance on the table. Chips running cooler run at higher clockspeeds with better efficiency. Thermal management is now a competitive differentiator, not just a facility problem.
The industry is moving to liquid cooling. Understanding the options, the economics, and the tradeoffs is increasingly important for anyone evaluating mining operations or companies.
The Three Cooling Approaches
Air cooling remains dominant by installed base, particularly in facilities built before 2022. The hardware is standardized, the operational knowledge is widely distributed, and the upfront cost is low. The limitations are thermal density ceilings around 15-20kW per rack, significant noise, and poor thermal efficiency. For lower-density deployments with cheap electricity, air cooling remains economically viable.
Direct liquid cooling (hydro cooling) routes coolant through internal channels in modified ASIC chassis, replacing the standard air-cooled heatsink. The coolant absorbs heat directly from the chip package and carries it to a radiator or cooling tower outside the facility. Bitmain's Hydro series and similar products have made this approach increasingly accessible. Thermal performance improves significantly versus air — chips run 10-15°C cooler under load — and the noise profile drops to near-zero. Density increases to 30-50kW per rack. The limitation is hardware: you need hydro-specific miners, and retrofitting air-cooled units is not practical.
Immersion cooling submerges entire mining rigs in tanks of dielectric fluid — a non-conductive coolant that makes direct contact with all components. The fluid absorbs heat and is pumped to a heat exchanger. Immersion offers the best thermal performance of any approach, enabling sustained operation at higher clockspeeds, dramatically extended hardware lifespan, and the ability to overclock units that would throttle in air. Density can reach 100kW per rack. The economics require careful analysis: the dielectric fluid, tanks, and plumbing infrastructure add significant upfront cost, and servicing requires pulling machines from the tanks. For large-scale industrial operations, the total cost of ownership often favors immersion over air at the 5-year mark.
Why This Matters Beyond Mining
The liquid cooling transition in Bitcoin mining is happening in parallel with a nearly identical transition in AI data centers. Next-generation GPU clusters for AI training and inference draw 50kW or more per rack — densities that air cooling cannot handle without massive facility footprint and power overhead for fans.
The companies that have built liquid cooling expertise in mining — the plumbing configurations, the fluid chemistry, the heat exchanger sizing, the operational procedures — are building directly transferable knowledge. As miners reposition their infrastructure for AI hosting, that thermal management expertise is part of the value proposition.
ENGIE's Assu Sol solar mining evaluation is another example: co-located industrial mining rigs absorbing curtailed solar power would almost certainly use liquid cooling to achieve the density and efficiency needed to justify the infrastructure investment.
The Economics of Transitioning
The decision to switch cooling approach depends on three variables: hardware generation, power cost, and operational scale.
For operators running last-generation air-cooled hardware at high power cost, the case for liquid cooling is immediate. The efficiency gains from hydro or immersion cooling — measured in joules per terahash — can more than offset the conversion cost when power is the binding constraint.
For operators with cheap power and modern air-cooled hardware already deployed, the calculation is more nuanced. Replacing functional infrastructure before end of life is only justified if the efficiency gains or density improvements unlock revenue that the current setup cannot capture.
For greenfield deployments in 2026, the answer is increasingly clear: build for liquid cooling from the start. The hardware roadmap for ASICs, like the hardware roadmap for GPUs, points toward higher density and higher thermal output. Facilities that cannot handle 50-100kW per rack will be obsolete before the decade is out.
What the Best Operations Look Like in 2026
The most competitive mining operations in 2026 combine several elements:
Cheap, long-term power — ideally below $0.04/kWh — from renewable or stranded sources. Hydro or immersion cooling to maximize efficiency at high chip densities. Modern hardware at or near the efficiency frontier (sub-18 J/TH). Operational flexibility to dial hashrate up or down based on profitability, power price, and network difficulty.
The facilities that check all four boxes are the ones that survive full difficulty cycles, weather bear markets, and remain profitable when marginal operators are forced offline.
At Ultra Labs, infrastructure quality — from power to cooling to hardware selection — is core to how we think about the mining business. The same rigor applies to our Cardano staking operation, where uptime and efficiency directly translate to delegator returns.
Delegate your ADA to the ULTRA pool using Eternl or Lace and earn rewards from a team that treats infrastructure as a first-principles problem.
Sources: Bitcoin Mining in 2026: Air, Hydro, Immersion, and the Bitaxe Revolution · Bitcoin Miners Are Becoming AI Infrastructure Companies
