Trane Technologies closed its acquisition of LiquidStack on March 3, marking the third time in four months that a legacy industrial giant swallowed a liquid cooling specialist whole. The deal, first announced February 10, gives the $21 billion HVAC conglomerate something it has never had: a proprietary immersion cooling platform. Financial terms were not disclosed. But the strategic math is obvious. Trane now owns every layer of data center thermal management, from the central plant to the chip.
Trane CEO Dave Regnery has been building toward this for years. The company made a minority investment in LiquidStack back in 2023, buying time and a front-row seat to the immersion cooling market while its competitors were still debating whether liquid would actually replace air at scale. That bet matured into something bigger. In December 2025, Trane acquired Stellar Energy Digital, a Jacksonville-based provider of turnkey liquid-to-chip cooling solutions with roughly 700 employees and two assembly operations in Florida. Then came LiquidStack. Three moves in rapid succession: modular cooling plants, direct-to-chip cold plates, immersion tanks. End-to-end.
The pattern reveals a CEO who decided the data center cooling market was not a side bet but a core business. Trane's commercial HVAC revenue jumped 20% year-over-year in 2025, and data centers were a key driver. But Regnery clearly saw that air-cooled chillers alone would not hold the line as GPU rack densities blow past 100 kW. You need liquid. And now, through Stellar Energy and LiquidStack, Trane has both cold plate and immersion capabilities in-house.
LiquidStack, headquartered in Carrollton, Texas, with engineering and R&D operations in Hong Kong, is one of the few companies on the planet with real production experience in two-phase immersion cooling. The company traces its roots to Allied Control Limited, a Hong Kong startup founded in 2012 that pioneered the technology before pivoting out of crypto mining and rebranding as LiquidStack under co-founder and CEO Joe Capes, who will stay on in a leadership role inside Trane.
Here is the technical detail that matters. LiquidStack's two-phase immersion system submerges servers in a dielectric fluid with a boiling point of roughly 50 degrees Celsius. As heat transfers from the chips, the fluid boils on contact, rises as vapor, hits a condenser, returns to liquid, and falls back into the tank. The cycle is largely passive. No pumps circulating coolant through cold plates. No fans. No air handlers. The company claims a partial PUE as low as 1.02, which is about as close to thermodynamic free lunch as the industry has ever gotten. LiquidStack's tanks support chips rated up to 1,000W and can handle 252 kW in a standard 48U rack footprint, while the company's MicroModular and MacroModular form factors scale further for hyperscale deployments. Compared to air cooling, the technology cuts energy use by 41%, reduces physical footprint by up to 90%, and can eliminate water consumption for heat rejection entirely.
Zoom out and the M&A picture looks like a land grab. Eaton closed its $9.5 billion acquisition of Boyd Thermal on March 12, just nine days after Trane completed its LiquidStack deal. Boyd brings cold plate and thermal component manufacturing with $1.5 billion in projected liquid cooling revenue for 2026. Eaton now calls its combined offering "grid-to-chip." Then on March 19, Ecolab announced a $4.75 billion deal for CoolIT Systems, snatching the direct-to-chip leader from KKR, which had bought CoolIT for just $270 million in June 2023. That is a 17x return in under three years. CoolIT projects roughly $550 million in sales over the next twelve months.
Three deals in March 2026. Three legacy industrials. Combined value north of $14 billion, and that only counts the two where terms were disclosed. The message from the buyer side is unanimous: organic R&D in liquid cooling is too slow. The AI infrastructure buildout is moving at a pace that demands you buy your way into the stack or get left selling chillers to office buildings.
The industry has spent years framing immersion cooling and direct-to-chip cold plates as competing philosophies. The Trane deal scrambles that narrative. By owning both Stellar Energy (cold plates, CDUs) and LiquidStack (immersion), Trane can offer whichever solution fits the workload. Hyperscalers running inference at moderate densities might prefer cold plates for their easier retrofit into existing air-cooled facilities. Training clusters pushing 100+ kW per rack, where every watt of cooling overhead cuts into compute budget, are where immersion starts to win on pure thermodynamics.
The data center liquid cooling market is on pace to hit roughly $7 billion by 2029, according to Dell'Oro Group. Cold plate dominates current deployments. But IDTechEx projects that two-phase solutions, both cold plate and immersion, will see large-scale adoption starting in 2026 and 2027 as next-generation accelerator chips from NVIDIA and AMD push thermal envelopes that single-phase systems struggle to manage. Trane is now positioned on both sides of that curve. So is Eaton, through Boyd. Ecolab, through CoolIT, is planted firmly in the cold plate camp.
The acquirers now face an integration problem that no press release can solve. LiquidStack has fewer than 200 employees. Trane has over 40,000. Bolting an immersion cooling startup onto a legacy HVAC sales machine means retraining field engineers, redesigning service contracts, and convincing enterprise customers that the same company selling rooftop units for their office campus can also keep their $200 million GPU cluster from melting. Joe Capes staying on helps. But the culture gap between a Hong Kong-born startup and a Dublin-headquartered industrial conglomerate is real.
The bigger question is pricing power. With three major liquid cooling independents now absorbed into industrial conglomerates, hyperscalers lose negotiating leverage. CoolIT, LiquidStack, and Boyd Thermal were all companies that could be played against each other. Now they sit inside Ecolab, Trane, and Eaton, respectively, bundled with power distribution, building management systems, and service contracts. The buyers are betting that bundling creates stickiness. The hyperscalers will test that theory fast.
Regnery has assembled the pieces. Chillers. Cold plates. Immersion tanks. Controls. Modular plants. The question is whether Trane can actually sell and deliver integrated thermal management at hyperscale speed, or whether this becomes another case of an industrial giant collecting assets and losing the engineers who made them valuable in the first place.