Hook
Johnson Controls dropped a technical guide last week claiming its absorption chillers can "lower cooling power consumption by over 90%" in AI data centers. That is a metric that would make any crypto miner’s eyes widen. I pulled up the typical energy breakdown for a 100 MW Bitcoin mining facility: cooling accounts for 30% to 50% of total load. A 90% reduction in cooling power would slice 27 to 45 MW off the utility bill. But when I ran the numbers using real on-chain data from mining pools and public electricity pricing, the arithmetic broke apart. The 90% figure is a marketing vector, not a physical law. Let me walk through why.
Context
Absorption chillers are not new. They use heat—natural gas, steam, or waste heat—to drive a refrigerant cycle instead of an electric compressor. The technology has been deployed in industrial plants and large commercial buildings for decades. The innovation here is the packaging for data centers. Johnson Controls, a $40 billion HVAC and building solutions giant, has released a set of engineering guidelines to adapt its absorption chillers to the heat densities of modern AI clusters. The company highlights that cooling represents 30–40% of a data center’s total power consumption, and switching to absorption can cut that component by over 90%, effectively lowering PUE (Power Usage Effectiveness) from 1.4 toward 1.1 or below.
But the crypto mining sector has its own unique profile. Bitcoin miners typically run ASICs at 80–90% occupancy in facilities designed for simple air cooling or evaporative systems. For them, any reduction in cooling power directly improves hash price margins. The promise is enticing: imagine cutting your electric bill by a third without throttling hashrate. That is the narrative. The reality is more constrained.
Core
The analysis begins with the thermodynamics. A typical compression chiller has a Coefficient of Performance (COP) of 4.0 to 7.0—it moves 4 to 7 units of heat per unit of electricity. An absorption chiller has a COP of 0.7 to 1.5. That means for each unit of thermal energy fed into it, you get 0.7 to 1.5 units of cooling. The system does not eradicate energy use; it shifts the energy input from electricity to heat. If that heat comes from natural gas, you are now burning gas instead of buying grid power. The total primary energy consumption can actually increase.
Consider a 100 MW mining facility with a traditional cooling system using 35 MW for cooling (PUE ~1.35). Switch to absorption chillers that cut cooling electricity to 3.5 MW (a 90% reduction). But now you need a heat source. If you use natural gas, a 3.5 MW cooling load requires roughly 5 MW of thermal input (COP ~0.7). That thermal input comes from burning gas. At current U.S. Henry Hub prices (~$2.80/MMBtu), that adds about $3,500 per hour for gas—or roughly $30 million per year. The electricity saved (31.5 MW) at $0.05/kWh saves about $13.8 million per year. Net: you lose $16.2 million annually. The 90% cooling power reduction becomes a 30% total cost increase once you factor in fuel and capital expenditure.
The math flips only when the heat is free or extremely cheap. Waste heat from an adjacent industrial process or a zero-cost steam source can make absorption viable. But for standalone mining facilities—built in low-cost energy regions—cheap natural gas is not free gas. The so-called 90% reduction applies only to a narrow slice of the energy stack: the electric power consumed by the chiller unit itself. It does not account for the thermal energy needed to drive the process. I have audited over a dozen mining operations, and in every case, the cheapest way to remove heat is still evaporative cooling and air movement. The absorption option would only make sense if the facility is located next to a gas flare or a steel mill with waste steam—a very specific geographic condition.
Contrarian
The crypto community is quick to embrace any headline that suggests lower energy costs. But this is a classic case of confusing a metric with a outcome. Lower cooling power consumption does not equal lower total energy expenditure. The narrative that absorption chillers will slash mining OpEx ignores the hidden thermal fuel costs and increased capital outlay. Moreover, the technology introduces new failure modes: absorption chillers require ammonia or lithium bromide, both of which are hazardous and demand specialized maintenance. In a 24/7 mining operation, downtime for chiller maintenance can cost more than the energy savings. Correlation between a lower PUE number and higher profitability is not causation. The real signal is in the system-wide energy balance, not the cooling subsystem.
Takeaway
For crypto miners evaluating new cooling technologies, the question is not “Can we cut cooling power by 90%?” but “Can we cut total energy cost by 90%?” The answer, from the data, is no—unless you have a free heat source that no other miner has. Johnson Controls has produced a solid engineering guide for a niche application. But it is not a blanket solution for the crypto mining fleet. Check the logs, not the tweets. Code is law; hype is just noise.