Are Traditional Data Centers Hurting Our Climate?

Every time you stream a video, send an email, or ask an AI a question, a data center is burning energy to make it happen. That's not inherently a problem. But the way most of those facilities manage heat — a byproduct as unavoidable as exhaust from an engine — has quietly become one of the infrastructure sector's most pressing environmental liabilities.

The numbers are hard to ignore. Data centers globally consume roughly 200 terawatt-hours of electricity per year, accounting for about 1% of worldwide electricity demand. Cooling systems alone are responsible for nearly 40% of that consumption in traditionally designed facilities. When you trace that power back through the grid to its source, you often land on natural gas turbines and coal plants emitting carbon dioxide at scale. Some large hyperscale campuses generate more annual CO₂-equivalent emissions than mid-sized industrial manufacturing plants — a comparison that rarely makes the headlines but absolutely should.

Understanding Data Center Cooling Methods

Walk into a traditional data center, and you're immediately hit by the scale of the thermal problem. Server racks run hot — sometimes exceeding 40°C at the chip level — and that heat must be continuously extracted or the hardware fails. The industry's default answer for decades has been computer room air conditioning units, or CRACs, which push chilled air through raised floors and across server cabinets in a cycle that runs 24 hours a day, 365 days a year.

These systems work. But "works" and "efficient" are very different things.

The fundamental flaw in traditional air-based cooling is that you're using one of the least thermally conductive substances on earth — air — to move heat from some of the most power-dense equipment ever manufactured. Chillers must work harder, fans spin constantly, and the mechanical infrastructure required to maintain cold-aisle/hot-aisle separation adds complexity and capital costs that compound over the facility's lifetime.

Cooling towers are another common component. They reject heat to the atmosphere through evaporation, which introduces a secondary problem: water consumption. A mid-sized data center can consume millions of gallons of water annually just for cooling — a resource constraint that's increasingly relevant in drought-prone regions of the American Southwest, where much of the nation's data center capacity is concentrated.

The Environmental Cost of Traditional Cooling

The carbon math gets uncomfortable quickly. When cooling systems drive 40% of a facility's power load, and that facility sits in a grid region with a carbon intensity of 0.4 to 0.6 kg CO₂ per kilowatt-hour (typical for many U.S. utility territories), the emissions stack up fast. A 100-megawatt data center campus — not unusually large by hyperscale standards — could be responsible for hundreds of thousands of metric tons of CO₂ annually from cooling alone.

To put that in perspective: the EPA estimates the average American passenger vehicle emits about 4.6 metric tons of CO₂ per year. That same 100MW facility's cooling load could offset the benefit of taking tens of thousands of cars off the road.

The comparison to other industries is instructive, if often misused. Critics sometimes point out that data centers consume less total electricity than, say, global steel production or cement manufacturing. True. But data center demand is growing at 15-20% annually in many markets, driven by AI workloads, cloud migration, and the proliferation of connected devices. Steel demand isn't doubling every few years. Data center power demand effectively is. The trajectory matters as much as the current snapshot.

There's also the indirect emissions problem. When a major data center cluster comes online in a region, it can single-handedly strain local grid capacity, forcing utilities to bring peaker plants — typically natural gas turbines — online to meet demand. Those turbines could emit more carbon dioxide annually than the headline renewable energy commitments of the tech companies running those data centers.

Innovative Cooling Solutions for Sustainable Data Centers

The good news is that the engineering community has been working on this problem seriously, and several approaches are moving from pilot projects to commercial scale.

Liquid cooling is the most significant near-term shift. Direct liquid cooling (DLC) routes chilled water or dielectric fluid directly to server components — CPUs, GPUs, memory — rather than blasting cold air at an entire cabinet. The thermal efficiency improvement is dramatic: water conducts heat roughly 25 times better than air. For AI training clusters running NVIDIA H100 or H200 GPUs, where rack densities can exceed 100 kW per cabinet, air cooling is becoming physically inadequate. Liquid isn't optional for these deployments — it's an engineering necessity.

Immersion cooling takes this further. Servers are submerged entirely in thermally conductive but electrically inert dielectric fluid. Companies like Submer, GRC (Green Revolution Cooling), and LiquidStack have deployed this at commercial scale, with Power Usage Effectiveness (PUE) ratings approaching 1.03 — meaning nearly all power drawn goes directly to computing, with almost none wasted on cooling overhead. Compare that to a traditional air-cooled facility running a PUE of 1.5 or higher.

Location strategy is another lever. Microsoft's Project Natick experiment placed a sealed data center pod on the seafloor off the coast of Scotland, using the ambient cold of the North Atlantic as a passive cooling medium. The retrieved servers showed lower failure rates than land-based equivalents. The project was a research effort, not a commercial rollout, but it demonstrated that the industry's geographic assumptions about where data centers can operate are far more flexible than originally assumed.

Free cooling — drawing in ambient outdoor air during cold months rather than running mechanical chillers — has also matured significantly. Facilities in Scandinavia, Iceland, and the northern United States increasingly run without chillers for substantial portions of the year. Facebook's data center in Luleå, Sweden, uses outside air for cooling year-round, contributing to a facility-level PUE of around 1.07.

Financial Benefits of Sustainable Data Center Practices

Sustainability and financial performance aren't in tension here — they're aligned, which makes the transition more likely to accelerate than slow.

Cooling represents the largest variable operating cost in most data centers after raw power. Improving cooling efficiency directly compresses operating expenditure. A facility that drops its PUE from 1.6 to 1.15 reduces its non-IT power spend by nearly 40%. At industrial electricity rates of $0.06-$0.09 per kWh and loads of 50+ megawatts, that translates to millions of dollars per year — per facility.

Water consumption reductions carry financial weight too, particularly in regions where water rights are increasingly scarce and expensive. In states like Arizona and Nevada, where data center development continues despite serious drought conditions, future water pricing and regulatory risk could make water-intensive cooling approaches a stranded-cost liability.

For investors and asset owners, there's also a growing valuation differential. Data center operators who can demonstrate low PUE, credible emissions accounting, and resilient cooling infrastructure are commanding premium valuations in the M&A market. Buyers — particularly institutional capital with ESG mandates — are increasingly pricing sustainability performance into acquisition multiples.

The Future: Regulations Are Coming, Whether Operators Are Ready or Not

The policy environment is shifting in ways that will force the issue for operators who haven't yet prioritized cooling sustainability.

The European Union's Energy Efficiency Directive now requires data centers above 500 kW to report PUE, water usage effectiveness (WUE), and renewable energy factors, with the data feeding into an EU-wide register. Ireland, home to a massive concentration of hyperscale capacity, has imposed moratoriums on new data center connections to the Dublin grid pending infrastructure upgrades. Singapore paused new data center construction entirely for nearly two years before implementing strict efficiency requirements for new approvals.

The U.S. is moving more slowly, but the direction is the same. The CHIPS Act and subsequent federal data infrastructure funding increasingly tie grants and incentives to energy efficiency benchmarks. State-level mandates — particularly in California — are tightening disclosure and performance requirements.

For operators, the risk isn't just regulatory compliance. It's site selection and permitting. A data center project that can't demonstrate best-in-class cooling efficiency may simply not get permitted in high-demand markets where grid and water capacity are constrained.

The operators who will win the next decade aren't necessarily the ones with the most megawatts under management today. They're the ones treating cooling infrastructure as a strategic asset — investing in liquid cooling retrofits, co-locating with renewable generation, and building in regions where the physics of climate actually work in their favor.

The data center industry has an opportunity to get ahead of a problem that's becoming impossible to ignore. The technology exists. The economics increasingly support the transition. The remaining variable is urgency — and regulations have a way of solving that problem for operators who lack it on their own.

[INTERNAL LINK: data center cooling solutions]

[INTERNAL LINK: sustainability in data centers]

[INTERNAL LINK: energy efficiency regulations]

###