As artificial intelligence (AI) workloads consume more electricity than ever, data center operators are turning to liquid cooling technology to manage extreme heat and keep power bills under control. Trane Technologies, a global HVAC (heating, ventilation, and air conditioning) giant, recently acquired LiquidStack, a liquid cooling specialist, to integrate direct-to-chip and immersion cooling into its portfolio. The move reflects a broader industry recognition that traditional air cooling alone cannot handle the thermal demands of modern graphics processing units (GPUs) and other AI infrastructure.

What Is Liquid Cooling and Why Does It Matter for Data Centers?

Liquid cooling works by circulating coolant directly to the hottest components in a server, such as GPUs and processors, rather than relying solely on fans and air circulation. This approach is far more efficient at removing heat because liquid absorbs and transfers thermal energy much faster than air. For data centers running thousands of GPUs simultaneously, the difference is substantial. Direct-to-chip cooling can reduce energy waste, lower operational costs, and allow operators to pack more computing power into the same physical space.

The timing of Trane's acquisition is significant. Data centers worldwide are racing to expand capacity to meet explosive demand for AI model training and inference. Meta, for example, is deploying "millions" of Nvidia chips, including the first large-scale rollout of Nvidia's standalone Grace CPUs, across AI-optimized data centers. This vertical integration of AI infrastructure means companies are designing entire systems around GPU-intensive workloads from the ground up, and thermal management has become a critical competitive advantage.

How Are Data Centers Managing Extreme Power Demands?

Beyond cooling innovations, data centers are pursuing multiple strategies to handle surging electricity consumption:

• Energy Procurement Partnerships: Meta, Amazon, Google, and Microsoft accounted for 49% of all corporate clean-energy procurement in 2025, securing renewable power through long-term contracts to offset AI workload emissions.
• On-Site Power Generation: Iron Mountain partnered with Calibrant Energy to deploy a 23-megawatt-hour battery energy storage system at its New Jersey data center, integrated with rooftop solar, demonstrating how facilities are moving toward self-sufficiency.
• Grid Infrastructure Investment: The PJM Interconnection, which serves 15 states, approved an $11.8 billion transmission upgrade and expansion plan aimed squarely at meeting surging data center demand, with roughly $4.8 billion allocated to "Data Center Alley" in Northern Virginia.
• Demand Management: Anthropic pledged to cover 100% of grid-upgrade costs tied to its data center growth, invest in new generation capacity, and deploy demand-curtailment systems during peak periods to reduce strain on regional electricity networks.

These efforts underscore how power availability and thermal efficiency have become the primary constraints on data center expansion. Utilities across North America and Europe are struggling to keep pace. American Electric Power added 28 gigawatts of new large-load demand to its interconnection pipeline, with roughly 80% tied to big tech companies. Meanwhile, the UK's Ofgem launched a sweeping review of connection rules with a specific focus on data centers, signaling that regulators are taking the infrastructure challenge seriously.

Where Are Data Centers Being Built to Support AI Growth?

Texas is emerging as the dominant hub for new data center construction. JLL projects Texas will become the world's largest data center market by 2030, overtaking Virginia. Of the more than 35 gigawatts of data center capacity under construction in North America, Texas alone accounts for 6.5 gigawatts. CyrusOne advanced a 760-megawatt Texas project colocated with a natural gas plant, while Meta announced a 1-gigawatt Indiana campus and quietly expanded its massive Hyperion site in Louisiana.

Internationally, the expansion is equally ambitious. India's Adani Group unveiled a massive $100 billion plan for renewable-powered, AI-ready data centers by 2035, anchored by hyperscale partnerships with Google and Microsoft. Yotta Data Services committed more than $2 billion to deploy over 20,000 Nvidia Blackwell GPUs in Noida, India, while Australia's Firmus secured a $10 billion debt facility to roll out Nvidia-based AI factories nationwide. Goodman Group added a massive 1 gigawatt to its global data center power bank, bringing available capacity to 6 gigawatts, primarily across Australia and Europe.

Not every project succeeds. Brennan Investment Group withdrew a $2 billion proposal outside Chicago after resident opposition, while Vantage Data Centers was denied permission to build a data center near Frankfurt. Local communities are increasingly scrutinizing data center development due to concerns about power consumption, water usage, and environmental impact.

What Role Will Thermal Innovation Play in the Future?

Trane's acquisition of LiquidStack positions the company to offer end-to-end thermal management solutions for AI workloads, combining its existing HVAC expertise with cutting-edge cooling technology. This vertical integration mirrors the broader industry trend toward purpose-built AI infrastructure. As GPU power consumption continues to climb, innovations in liquid cooling, battery storage, and grid management will determine which operators can scale efficiently and which will face bottlenecks.

The data center industry is at an inflection point. Power is no longer just an operational expense; it is a strategic asset. Companies that secure reliable, affordable electricity and deploy efficient cooling systems will have a decisive advantage in the race to build the next generation of AI infrastructure.