Trend analysis: Green IT & sustainable data centers — where the value pools are (and who captures them)

Global data center electricity consumption is projected to exceed 1,000 TWh by 2028, roughly equivalent to the total electricity demand of Japan. As AI workloads drive an unprecedented surge in computing demand, the race to decarbonize digital infrastructure has become one of the largest value creation opportunities in the energy transition, with an estimated $500 billion in cumulative investment flowing into sustainable data center infrastructure through 2030.

Why It Matters

Data centers already consume approximately 2-3% of global electricity, and that share is rising sharply. The International Energy Agency estimates that data center power demand could double between 2024 and 2028, driven overwhelmingly by generative AI training and inference workloads. A single large language model training run can consume as much electricity as 1,000 US households use in a year, and inference at scale multiplies that demand by orders of magnitude. For hyperscalers like Microsoft, Google, and Amazon, this creates a direct conflict with corporate net-zero commitments: Microsoft's emissions rose 29% between 2020 and 2024, largely due to data center expansion. The companies, investors, and technology providers that solve the sustainability equation for digital infrastructure will capture outsized returns. The stakes extend beyond corporate reputations. Governments are increasingly scrutinizing data center water usage, grid impact, and land use. Ireland, where data centers consume 21% of national electricity, has imposed planning moratoria. Singapore paused new data center construction for three years before introducing a green certification requirement. The regulatory pressure is creating a premium for operators who can demonstrate genuine environmental performance rather than just purchasing renewable energy certificates.

Key Concepts

Power Usage Effectiveness (PUE) is the ratio of total facility energy to IT equipment energy. A PUE of 1.0 means all energy goes to computing; a PUE of 2.0 means half is consumed by cooling and overhead. The global average PUE has improved from 2.0 in 2010 to approximately 1.58 in 2025, with leading facilities achieving 1.1-1.2. Further gains require fundamental architectural changes rather than incremental efficiency improvements.

Water Usage Effectiveness (WUE) measures liters of water consumed per kilowatt-hour of IT energy. Evaporative cooling systems are water-intensive, consuming 1.8-7.6 liters per kWh depending on climate. As water scarcity intensifies, WUE is becoming a critical site selection factor and regulatory constraint.

Carbon-free energy matching goes beyond annual renewable energy procurement to match hourly electricity consumption with carbon-free generation on the same grid. Google pioneered the 24/7 carbon-free energy (CFE) concept, achieving 64% hourly matching globally in 2023 and targeting 100% by 2030. This approach drives genuine grid decarbonization rather than relying on accounting offsets.

KPI	Current Benchmark	Leading Practice	Laggard Threshold
Power Usage Effectiveness (PUE)	1.5-1.7	<1.2	>2.0
Water Usage Effectiveness (L/kWh)	1.8-3.0	<0.5 (air-cooled)	>5.0
Renewable energy procurement (%)	60-80% annual	100% 24/7 hourly match	<40%
Scope 2 emissions intensity (kg CO2e/MWh IT load)	200-350	<50	>500
Waste heat recovery rate (%)	5-10%	>40%	0%
Server utilization rate (%)	30-50%	>65%	<20%

What's Working

Liquid cooling for AI workloads. Air cooling hits physical limits at rack densities above 30-40 kW, but AI GPU clusters now demand 80-120 kW per rack. Liquid cooling technologies, including direct-to-chip and immersion cooling, reduce cooling energy by 30-50% while enabling higher compute density. Equinix deployed liquid cooling across its xScale facilities, achieving PUE improvements of 0.15-0.2 compared to air-cooled equivalents. The technology also enables waste heat recovery at temperatures useful for district heating. Stockholm Data Parks, a collaboration between the city and data center operators, supplies waste heat to 30,000 apartments, turning a cost center into a revenue stream.

Modular and edge architectures. Prefabricated modular data centers from companies like Schneider Electric and Vertiv enable rapid deployment with optimized efficiency. Microsoft's Azure Modular Datacenter program has demonstrated 40% faster deployment with 15% lower PUE than traditional builds. Edge computing reduces the energy penalty of long-distance data transmission by processing workloads closer to end users. Vapor IO's Kinetic Edge platform, deployed across 36 US metro areas, has shown 25-35% reductions in total energy consumption for latency-sensitive applications.

Corporate PPA innovation for 24/7 matching. Rather than relying on annual renewable energy certificate purchases, leading operators are structuring power purchase agreements that pair solar, wind, and storage to match hourly demand profiles. Google's partnership with AES and Clearway Energy in Texas combines wind, solar, and battery storage into a single 24/7 CFE supply contract. Microsoft's agreement to purchase nuclear power from Constellation Energy for its data centers signals the scale of demand for baseload carbon-free power. These contracts are creating a new market for dispatchable clean energy products.

What's Not Working

Renewable energy certificate arbitrage. Many data center operators claim 100% renewable energy through the purchase of unbundled RECs that do not correspond to actual carbon-free electricity on their grid. A 2025 analysis by Ember found that the grid emissions intensity at the actual locations of "100% renewable" data centers often exceeded 400 kg CO2e/MWh during peak hours. The gap between accounting claims and physical reality undermines credibility and delays genuine decarbonization. Market pressure is building for hourly, location-based emissions accounting, but most operators have not adopted it.

Water consumption at scale. Despite efficiency improvements, the aggregate water footprint of data centers is growing. Google's US data centers consumed 5.6 billion gallons of water in 2023, a 17% year-over-year increase. In water-stressed regions like the American Southwest and parts of India, this creates direct competition with agricultural and municipal water needs. Air-cooled and advanced liquid-cooled alternatives exist but carry higher capital costs and are not yet deployed at scale in hot climates where cooling demand is greatest.

Embedded carbon in construction. The focus on operational energy overlooks the significant embodied carbon in data center construction. A typical hyperscale facility contains 30,000-50,000 tonnes of concrete and steel, with embodied carbon representing 20-30% of total lifecycle emissions for facilities powered by clean energy. Industry standards for measuring and reducing embodied carbon in data center construction remain fragmented, with no widely adopted benchmark equivalent to PUE for construction impacts.

Key Players

Established Leaders

• Equinix: Operates 260+ data centers globally with a commitment to 100% renewable energy and climate-neutral operations by 2030. Leads in liquid cooling deployment and waste heat recovery integration.
• Google Cloud: Pioneered 24/7 carbon-free energy matching methodology. Published open-source tools for hourly emissions tracking and invests in next-generation geothermal and nuclear for baseload supply.
• Microsoft Azure: Pursuing nuclear power agreements with Constellation Energy and investing in fusion energy through Helion. Deploying underwater data center prototypes and modular architectures.
• Schneider Electric: Leading supplier of power and cooling infrastructure for data centers. Its EcoStruxure platform provides integrated energy management and sustainability monitoring.

Emerging Startups

• LiquidCool Solutions: Develops single-phase liquid immersion cooling systems that eliminate the need for fans, compressors, and chillers, reducing cooling energy by up to 95%.
• ZutaCore: Offers HyperCool direct-on-chip dielectric liquid cooling technology enabling rack densities above 100 kW while maintaining PUE below 1.1.
• ECL (Lancium Clean Campus): Building behind-the-meter data centers co-located with renewable energy generation, enabling real-time carbon-free computing without grid dependency.
• Nautilus Data Technologies: Deploys water-cooled floating data centers that use natural water bodies for heat rejection, achieving PUE of 1.15 with zero water consumption.

Key Investors and Funders

• Brookfield Infrastructure Partners: Committed over $15 billion to digital infrastructure investments, including sustainable data center platforms across North America and Europe.
• DigitalBridge: Manages $80 billion in digital infrastructure assets, with sustainability requirements embedded in portfolio company operating frameworks.
• Global Infrastructure Partners: Major investor in data center platforms with increasing focus on energy-efficient and renewable-powered facilities.

Where the Value Pools Are

Cooling technology and thermal management. The data center cooling market is projected to reach $32 billion by 2028, driven by the transition from air to liquid cooling for AI workloads. Companies that provide end-to-end thermal management solutions, combining liquid cooling hardware with intelligent controls and waste heat recovery, command 2-3x the margins of traditional HVAC vendors. The shift to liquid cooling is not optional for AI-focused facilities, making it a structural growth market.

24/7 carbon-free energy supply. The gap between annual renewable procurement and hourly carbon-free matching creates a massive opportunity for energy developers who can structure firm, dispatchable clean power products. Nuclear restarts, advanced geothermal, and solar-plus-storage hybrids are all being pulled into the market by data center demand. Microsoft's willingness to sign a 20-year nuclear PPA at premium pricing signals the value that operators place on baseload carbon-free supply.

Sustainability software and monitoring. Operators need granular, real-time visibility into energy consumption, carbon emissions, water usage, and waste metrics across distributed facility portfolios. The market for data center infrastructure management (DCIM) software with integrated sustainability analytics is growing at 15-20% annually. Platforms that provide auditable data for regulatory compliance (EU Energy Efficiency Directive, SEC climate disclosures) earn recurring revenue with high retention rates.

Waste heat monetization. Data centers produce enormous quantities of low-grade heat, typically 30-50 degrees Celsius. In Northern European markets with established district heating networks, this heat has direct commercial value. Finland, Sweden, and Denmark have pioneered regulatory frameworks that enable data centers to sell waste heat, turning an operating cost into a revenue stream worth $5-15 per MWh of IT load. As heat pump technology improves, the addressable market expands to warmer climates where district heating has not traditionally been viable.

Action Checklist

• Benchmark current PUE, WUE, and carbon intensity against industry leaders and set improvement targets with specific timelines
• Evaluate liquid cooling deployment for current and planned AI workloads, with total cost of ownership analysis including waste heat recovery potential
• Transition from annual renewable energy certificate procurement to hourly, location-based carbon-free energy matching
• Assess water risk at all facility locations using tools like the WRI Aqueduct Water Risk Atlas and develop site-specific water reduction strategies
• Quantify embodied carbon in planned construction and set reduction targets through material substitution and design optimization
• Explore waste heat recovery partnerships with district heating networks, greenhouses, or industrial processes near facility locations
• Implement DCIM with integrated sustainability metrics to provide real-time operational visibility and regulatory compliance data

FAQ

What is the difference between PUE and carbon intensity as a sustainability metric? PUE measures energy efficiency within the facility boundary: how much overhead energy is consumed relative to IT load. A facility with a PUE of 1.2 is highly efficient, but if powered by coal-generated electricity, its carbon intensity remains extremely high. Carbon intensity (measured in kg CO2e per MWh of IT load) captures both efficiency and energy source. Leading operators now track both metrics, but carbon intensity is the more meaningful indicator of environmental impact.

How realistic is 100% 24/7 carbon-free energy for data centers? Google achieved 64% global hourly matching in 2023 and demonstrated over 90% in specific locations like Denmark and Finland where clean energy supply aligns well with demand profiles. Reaching 100% everywhere requires solving the "last 10%" problem: nighttime hours in regions with limited wind or hydro resources. Advanced storage, nuclear, and geothermal are the most likely solutions for closing the gap. Full 24/7 matching at global scale is likely achievable by the early 2030s for operators willing to pay a 10-20% premium.

Will AI growth overwhelm data center sustainability efforts? This is the central tension. AI workload growth could consume all efficiency gains and then some. The IEA projects that AI-related data center energy demand could increase 10x between 2023 and 2030 under aggressive adoption scenarios. The answer depends on whether efficiency improvements in chips (NVIDIA's Blackwell architecture delivers 4x the inference efficiency of its predecessor), cooling, and software optimization can keep pace. History suggests they partially will: the hyperscaler industry has grown compute output 6x since 2015 while only doubling energy consumption. But the absolute energy and resource demands will continue rising.

What regulatory risks should data center operators prepare for? The EU Energy Efficiency Directive now requires data centers above 500 kW to report PUE, WUE, renewable energy share, and waste heat utilization annually. Ireland and the Netherlands have imposed grid capacity constraints on new facilities. Singapore requires Green Mark certification for all new data centers. The trend is toward mandatory efficiency standards, water use limits, and grid impact assessments. Operators without proactive compliance strategies risk permitting delays and operational restrictions.

Sources

1. International Energy Agency. "Electricity 2024: Data Centres, AI and Cryptocurrency." IEA, 2024.
2. Uptime Institute. "Global Data Center Survey 2025." Uptime Institute, 2025.
3. Google. "24/7 Carbon-Free Energy: Methodology and Progress Report." Google Environmental Report, 2025.
4. Ember. "Data Centre Electricity: Matching Claims with Reality." Ember Climate, 2025.
5. BloombergNEF. "Data Center Sustainability Investment Outlook." BNEF, 2025.
6. European Commission. "Energy Efficiency Directive: Data Centre Reporting Requirements." EU Official Journal, 2024.
7. Carbon Tracker Initiative. "Digital Demand: The Carbon Cost of AI." Carbon Tracker, 2025.