carbon footprint

The datacenter industry’s carbon footprint is enormous and growing rapidly, with 131 GW of capacity analyzed consuming an estimated 1,150 TWh of electricity annually—equivalent to 3% of global electricity consumption. However, the industry is simultaneously leading corporate decarbonization efforts, with 37.4% of capacity committed to renewable energy, PUE efficiency improving from 2.5+ (2007) to 1.06 (Meta Prineville leading), and pioneering pathways toward 24/7 carbon-free energy and net zero operations.

overview

scope 1, 2, 3 emissions framework

scope 1: direct emissions

Definition

• Direct emissions from owned or controlled sources
• On-site fuel combustion
• Company-owned vehicles
• Refrigerant leakage
• Direct control by datacenter operator

Datacenter Sources

• Backup diesel generators (testing and emergency use)
• Natural gas for heating (in cold climates)
• On-site power generation (rare in US, more common internationally)
• Refrigerant leaks from cooling systems (HFCs, very high GWP)
• Fleet vehicles for operations and maintenance

Magnitude

• Typically 1-5% of total datacenter emissions
• Diesel generator testing: 10-50 hours per year per generator
• Emergency use rare but high emissions when occurs
• Refrigerant leaks: small volume but high global warming potential
• Relatively small but directly controllable

Reduction Strategies

• Replace diesel generators with battery storage or fuel cells
• Green hydrogen for backup power (Data City Texas pioneer)
• Low-GWP refrigerants (HFO alternatives)
• Renewable natural gas for heating
• Electric vehicle fleet transition

scope 2: indirect emissions from purchased energy

Definition

• Indirect emissions from purchased electricity, steam, heating, cooling
• Most significant category for datacenters
• Where renewable energy commitments have impact
• Market-based accounting vs location-based accounting

Datacenter Impact

• Electricity consumption: 95%+ of datacenter emissions
• 24/7 operations, massive power consumption
• 131 GW capacity analyzed: approximately 1,150 TWh annually
• At average US grid carbon intensity (0.39 kg CO2/kWh): 448 million metric tons CO2 annually
• Equivalent to 97 million gasoline passenger vehicles

Accounting Methods

• Location-based: Uses average grid carbon intensity where facility located
• Market-based: Uses carbon intensity of contractual instruments (PPAs, RECs)
• Renewable energy commitments affect market-based accounting
• Debate over which better represents actual environmental impact

Reduction Strategies

• Renewable energy PPAs (primary strategy)
• On-site solar generation
• Renewable energy certificates (RECs) purchase
• 24/7 carbon-free energy matching
• Grid decarbonization advocacy
• Energy efficiency (PUE improvement)

scope 3: value chain emissions

Definition

• Indirect emissions in value chain
• Upstream and downstream activities
• Not owned or controlled
• Most comprehensive but most difficult to measure

Upstream Scope 3 (Datacenter Perspective)

• Embodied carbon in construction: Steel, concrete, equipment manufacturing
• IT equipment manufacturing: Servers, storage, networking equipment
• Supply chain: Component manufacturing, transportation
• Construction process: Equipment operation, worker transportation
• Infrastructure: Transmission lines, substations, fiber networks

Downstream Scope 3 (Datacenter Perspective)

• Tenant emissions: If colocation provider, tenant computing activities
• Data transmission: Network infrastructure beyond datacenter
• End-user devices: Devices accessing datacenter services
• Product lifecycle: Disposal, recycling, e-waste

Magnitude

• Can equal or exceed Scope 1+2 combined
• Embodied carbon in construction: 5-20% of lifetime emissions
• IT equipment manufacturing: significant component
• Difficult to measure comprehensively
• Increasing scrutiny and disclosure requirements

Reduction Strategies

• Low-carbon concrete (Meta DeKalb AI-developed concrete: 40% lower emissions)
• Steel with recycled content or low-carbon manufacturing
• Supplier engagement and requirements
• Circular economy approaches (equipment reuse, refurbishment)
• Extended equipment lifecycles
• Renewable energy in supply chain
• Transportation optimization

carbon neutral commitments

hyperscale operator goals

Google

• Carbon neutral since 2007 (via offsets and renewable matching)
• 24/7 carbon-free energy by 2030 (industry-leading goal)
• Operational carbon-free energy all datacenters and offices
• Eliminating carbon offsets over time
• Focus on additionality and hourly matching

Meta (Facebook)

• Net zero across value chain by 2030
• 100% renewable energy achieved 2020 (annual matching)
• Supporting grid renewable energy development
• Scope 3 reduction strategies
• Supply chain engagement
• Meta Kansas City: net-zero carbon emissions commitment

Microsoft

• Carbon negative by 2030
• Remove all historical emissions by 2050 (since 1975)
• 100% renewable energy by 2025 (achieved)
• Scope 3 reduction by 50% by 2030
• $1 billion Climate Innovation Fund
• Supply chain decarbonization requirements

Amazon Web Services (AWS)

• Net-zero carbon by 2040 (10 years ahead of Paris Agreement)
• 100% renewable energy by 2025 (on track)
• Climate Pledge co-founder
• $2 billion Climate Pledge Fund
• World’s largest corporate renewable energy buyer
• Talen Energy nuclear deal (Pennsylvania): 960 MW carbon-free

Apple

• Carbon neutral for corporate operations since 2020
• Carbon neutral products by 2030
• 100% renewable energy for datacenters
• Supply chain decarbonization (suppliers required renewable transition)
• Reno datacenters: 100% renewable

colocation provider commitments

Equinix

• 100% renewable energy achieved 2020
• Climate-neutral operations
• Science-based targets approved
• Long-term renewable coverage goal
• Global sustainability program

Digital Realty

• 100% renewable energy goal by 2030
• Sustainability-linked financing
• Science-based targets
• Carbon neutral in select markets
• Customer-specific renewable procurement

CyrusOne

• 100% renewable energy by 2040
• Water-free cooling reducing indirect emissions
• Phoenix-Chandler: first net-positive water datacenter
• ESG-focused development

Aligned Data Centers

• 100% renewable energy options all campuses
• Delta3 cooling: 85% water reduction
• Climate-focused design
• Renewable energy matching standard offering

regional provider commitments

QTS Data Centers

• Science-based emission reduction targets
• Renewable energy programs
• York County South Carolina: carbon-free power when feasible
• Customer sustainability partnerships

Vantage Data Centers

• Net zero carbon by 2030 commitment
• Phoenix campus: net zero carbon goal
• Renewable energy procurement
• Efficient cooling and operations

embodied carbon in construction

concrete and steel

Concrete Carbon Intensity

• Cement production: 8% of global CO2 emissions
• Typical datacenter: thousands of tons of concrete
• Foundation, structural elements, floor slabs
• Traditional concrete: 400-800 kg CO2/ton
• Datacenter construction: tens of thousands of tons CO2 from concrete alone

Low-Carbon Concrete Innovations

• Meta DeKalb Data Center (Illinois): AI-developed low-carbon concrete
• 40% lower carbon emissions than traditional
• Maintains structural performance
• Scalable to other projects
• Demonstrates machine learning for material optimization

Steel Production

• Traditional steel: 1.8-2.0 tons CO2/ton steel produced
• Blast furnace process carbon-intensive
• Recycled steel content reduces emissions
• Electric arc furnace with renewable energy: dramatic reduction
• Green steel (hydrogen-based) emerging but expensive

Datacenter Applications

• Structural steel framing
• Backup generator enclosures
• Cooling equipment
• Electrical infrastructure
• Thousands of tons per hyperscale facility

equipment manufacturing

Server Manufacturing

• Complex supply chain
• Semiconductor fabrication energy-intensive
• Rare earth mining environmental impact
• Transportation emissions
• Packaging materials
• Estimated 500-1,000 kg CO2 per server over lifecycle

Cooling Equipment

• Chillers, cooling towers, pumps, fans
• Manufacturing energy and materials
• Refrigerants (high GWP)
• Transportation (often international shipping)
• Installation energy

Electrical Infrastructure

• Transformers, switchgear, UPS systems, batteries
• Copper and aluminum production (energy-intensive)
• Lithium battery production (significant emissions)
• Generator manufacturing
• Cable production

construction process emissions

Heavy Equipment

• Excavators, cranes, concrete trucks, forklifts
• Diesel-powered equipment
• Hundreds of thousands of gallons diesel per large project
• Construction timeline: 18-36 months typical
• Electrification potential (electric construction equipment emerging)

Worker Transportation

• Peak construction: 1,000-3,000 workers on site
• Daily commutes over 18-36 months
• Significant cumulative emissions
• Remote locations exacerbate (longer commutes)
• Carpool programs and shuttles reduce impact

Material Transportation

• Steel, concrete, equipment shipped to site
• Often long distances (international for equipment)
• Heavy loads, high emissions
• Multiple deliveries over construction period
• Supply chain optimization reduces emissions

embodied carbon magnitude

Typical Hyperscale Datacenter

• 100 MW facility
• Estimated 50,000-100,000 metric tons CO2 embodied carbon
• Equivalent to 10,000-20,000 passenger vehicles one year
• Or 5-10% of 30-year operational emissions (traditional grid)
• Ratio improving as operational emissions decline (renewable energy)

Growing Significance

• As operational emissions decline (renewable energy), embodied carbon grows as percentage
• Potential to reach 20-30% of lifecycle emissions
• Increasing focus on construction phase
• Circular economy approaches critical

power usage effectiveness (pue) trends

pue definition and calculation

Formula

• PUE = Total Facility Energy / IT Equipment Energy
• Lower is better (closer to 1.0)
• PUE 2.0: 50% of energy goes to IT, 50% to infrastructure (cooling, power distribution, lighting)
• PUE 1.5: 67% to IT, 33% to infrastructure
• PUE 1.2: 83% to IT, 17% to infrastructure
• PUE 1.06: 94% to IT, 6% to infrastructure

Components

• IT equipment energy: servers, storage, networking
• Cooling energy: chillers, pumps, fans, cooling towers
• Power distribution losses: transformers, UPS, PDUs
• Lighting (typically minimal)
• PUE encompasses all non-IT energy

historical trends

2007 Baseline

• Industry average PUE: 2.5 or higher
• Inefficient cooling (raised floor, CRAC units)
• Poor airflow management
• Hot and cold air mixing
• Inefficient power distribution
• Oversized for reliability

2010-2015 Improvements

• Hot aisle/cold aisle containment
• Variable speed fans
• Free cooling (economizers)
• Higher temperature setpoints
• Improved airflow design
• Industry average: 1.8-2.0

2016-2020 Maturation

• Adiabatic cooling
• Liquid cooling for high-density
• Advanced containment
• AI-optimized cooling (Google DeepMind)
• Industry average: 1.5-1.7
• Hyperscale leaders: 1.2-1.3

2021-Present

• Waterless cooling technologies
• Direct-to-chip liquid cooling
• Immersion cooling pilots
• AI optimization standard
• Industry leaders: 1.06-1.2
• Average improving but still 1.4-1.6

industry leaders

Meta Prineville Data Center (Oregon)

• PUE 1.06 (industry-leading)
• Operational since 2012
• Free cooling leveraging Oregon climate
• Advanced airflow management
• Custom server design
• Continuous optimization

Prometheus Hyperscale WY-1 (Wyoming)

• PUE 1.08
• Evanston campus
• Leverages Wyoming’s cool climate
• High-altitude advantages
• Efficient design

TPC Data Centers Konterra 1 (Maryland)

• PUE 1.1
• Advanced cooling design
• Efficient power distribution
• Optimal for Maryland climate

Edged Data Centers (Multiple Locations)

• PUE 1.15 (Kansas City, Ankeny, Aurora)
• ThermalWorks waterless cooling
• Remarkably efficient despite no water
• AI-optimized operations
• Demonstrates waterless need not sacrifice efficiency

Aligned Data Centers Salt Lake City

• PUE 1.15
• Delta3 waterless cooling
• Low humidity climate advantage
• 352 MW capacity at high efficiency

factors affecting pue

Climate

• Cold climates enable lower PUE (free cooling)
• Hot climates more challenging
• Humidity affects cooling efficiency
• Altitude impacts air density and cooling

Cooling Technology

• Waterless cooling: typically 0.1-0.3 higher PUE in hot climates
• Evaporative cooling: lowest PUE in hot, dry climates
• Liquid cooling: enables high density, can improve overall PUE
• Free cooling: dramatic PUE improvement in cold climates

Design and Construction

• Containment (hot aisle/cold aisle)
• Airflow optimization
• Right-sizing equipment (not over-provisioning)
• Modular design
• Scalability allowing high utilization

Operations

• Temperature setpoints (ASHRAE allows higher, reducing cooling load)
• Humidity setpoints
• AI optimization (Google DeepMind: 40% cooling energy reduction)
• Continuous commissioning
• Maintenance and monitoring

IT Load Characteristics

• High-density computing more challenging to cool
• AI workloads: 50-200+ kW per rack
• Blade servers vs traditional
• Server refresh cycles
• Utilization rates

net zero pathways

operational net zero

Renewable Energy Matching

• 100% annual renewable energy matching
• Purchase RECs equal to consumption
• PPAs for renewable energy
• On-site generation
• Achieves net zero operational emissions (Scope 2, market-based)

Limitations

• Annual matching vs hourly reality
• Grid still uses fossil fuels some hours
• RECs may not represent additionality
• Does not address Scope 1 or Scope 3
• “Net zero” vs “carbon-free” distinction

24/7 carbon-free energy

Google’s Approach

• Carbon-free energy every hour, not just annual matching
• Requires baseload carbon-free (nuclear, geothermal, hydro)
• Battery storage for renewable intermittency
• Advanced grid management
• Significantly more challenging than annual matching

Technology Requirements

• Solar + battery storage
• Wind + battery storage
• Geothermal baseload
• Nuclear baseload (Talen Energy deal example)
• Hydrogen fuel cells potential
• Grid-interactive computing (shift load to renewable availability)

Status

• Google pilot projects underway
• Oklahoma demonstration projects
• Google Mayes County: 372 MW solar within one mile
• Industry following Google’s leadership
• Cost premium declining as technology matures

comprehensive net zero (all scopes)

Scope 1 Elimination

• Replace diesel generators with battery storage or fuel cells
• Green hydrogen for backup power
• Eliminate natural gas heating
• Low-GWP refrigerants
• Electric vehicle fleet

Scope 2 Excellence

• 24/7 carbon-free energy
• Not just renewable matching
• Hourly or sub-hourly tracking
• Additionality ensuring new renewable capacity
• Grid decarbonization support

Scope 3 Reduction

• Low-carbon construction materials
• Supplier renewable energy requirements
• Extended equipment lifecycles
• Circular economy approaches
• Transportation optimization
• Renewable energy in supply chain

Residual Offsets

• High-quality carbon offsets for unavoidable emissions
• Preference for removal (direct air capture, reforestation)
• Avoidance offsets (renewable energy, efficiency) less preferred
• Additionality, permanence, verifiability critical
• Microsoft: aiming to eliminate offsets over time

carbon negative

Microsoft’s Goal

• Carbon negative by 2030
• Remove more carbon than emit
• Remove all historical emissions by 2050 (since 1975)
• $1 billion Climate Innovation Fund
• Investment in carbon removal technologies

Carbon Removal Technologies

• Direct air capture (DAC)
• Bioenergy with carbon capture and storage (BECCS)
• Enhanced weathering
• Afforestation and reforestation
• Soil carbon sequestration
• Ocean-based approaches

Feasibility

• Current carbon removal costs: $100-600+ per ton
• Gigaton-scale removal needed
• Technology improving and costs declining
• Policy support growing (IRA, 45Q tax credits)
• Microsoft purchasing removal credits at scale

energy efficiency metrics

pue limitations

What PUE Doesn’t Measure

• IT equipment efficiency (servers, storage, networking)
• Carbon intensity of electricity
• Water consumption
• Renewable energy usage
• Total cost of ownership
• Workload efficiency

Complementary Metrics Needed

• Carbon Usage Effectiveness (CUE)
• Water Usage Effectiveness (WUE)
• Renewable Energy Factor (REF)
• Data Center Infrastructure Efficiency (DCiE) = 1/PUE
• Performance per Watt (IT equipment efficiency)

carbon usage effectiveness (cue)

Definition

• CUE = Total CO2 Emissions (kg) / IT Equipment Energy (kWh)
• Accounts for carbon intensity of electricity
• Better representation of climate impact than PUE alone
• Varies dramatically by grid and renewable energy usage

Example Calculations

• 100 MW datacenter, PUE 1.5, US average grid (0.39 kg CO2/kWh)
• Annual IT energy: 876,000 MWh
• Total facility energy: 1,314,000 MWh
• Total emissions: 512,460 metric tons CO2
• CUE: 0.585 kg CO2/kWh IT energy

With 100% Renewable Energy

• Same datacenter, 100% renewable matching
• Market-based Scope 2 emissions: 0
• CUE: 0 (operational only)
• Demonstrates renewable energy impact

Regional Variation

• Pacific Northwest (hydro-heavy): low CUE even without renewables
• Coal-heavy grids (West Virginia): very high CUE without renewables
• Renewable energy commitments override location-based CUE

performance per watt

IT Equipment Efficiency

• Computational performance per watt consumed
• Improving rapidly with Moore’s Law continuation
• Server refresh: 2x-3x performance per watt every 3-5 years
• SSD replacing HDD: dramatic energy savings
• Specialized processors (GPUs, TPUs, ASICs) for efficiency

Workload Optimization

• Software efficiency improvements
• Containerization and microservices
• Resource utilization optimization
• Decommissioning zombie servers
• Workload consolidation

Impact

• Faster server refresh reduces energy for same computing
• Offsets growing computing demand
• May allow datacenter capacity without proportional energy growth
• Carbon impact depends on embodied carbon vs operational savings

future pathways and innovations

short-term (2025-2027)

Renewable Energy Scaling

• 50%+ of new capacity with renewable commitments
• PPAs standard for hyperscale
• On-site solar increasingly common
• Battery storage integration growing
• Grid renewable percentage increasing

Efficiency Improvements

• PUE approaching 1.1 for leaders
• Waterless cooling mainstream
• Liquid cooling for AI workloads
• AI-optimized operations
• Higher temperature setpoints

Scope 3 Focus

• Low-carbon concrete and steel adoption
• Supplier engagement intensifying
• Circular economy initiatives
• Extended equipment lifecycles
• Embodied carbon disclosure

medium-term (2027-2030)

24/7 Carbon-Free Energy

• Google achieves 24/7 CFE goal
• Other hyperscalers follow
• Baseload carbon-free (nuclear, geothermal) resurgence
• Battery storage costs enabling hourly matching
• Grid-interactive computing demonstrations

Technology Breakthroughs

• Enhanced geothermal systems (EGS) deployments
• Small modular reactors (SMRs) first commercial deployments
• Green hydrogen for backup power pilots
• Advanced battery chemistries (solid-state, flow batteries)
• Waste heat recovery at scale

Net Zero Widespread

• Most hyperscale operators achieve operational net zero
• Scope 3 reduction strategies maturing
• Carbon removal purchases standard
• Industry-wide carbon accounting standards
• Regulatory requirements in some jurisdictions

long-term (2030+)

Carbon-Free Computing

• 100% carbon-free datacenter sector operational emissions
• 24/7 CFE standard for hyperscale
• SMRs widely deployed
• Fusion energy potential (2035+)
• Green hydrogen backup power mainstream

Comprehensive Decarbonization

• Scope 3 emissions reduced 80%+
• Circular economy mature
• Carbon-free construction materials
• Zero-emission construction equipment
• Supply chain fully decarbonized

Carbon Negative Sector

• Datacenter sector removes more carbon than emits
• Direct air capture integrated
• Afforestation and reforestation projects
• Biomass with CCS for backup power
• Industry leadership in climate solutions

policy and regulatory drivers

federal incentives (united states)

Investment Tax Credit (ITC)

• 30% credit for solar projects
• Energy storage eligible
• Supports on-site solar and battery storage
• Reduces PPA prices
• Extended by Inflation Reduction Act (IRA)

Production Tax Credit (PTC)

• Per-kWh credit for wind, geothermal, others
• Technology-neutral clean energy credits
• Lowers renewable energy costs
• 10-year credit period
• Supports datacenter renewable PPAs

45Q Carbon Capture Tax Credit

• Up to $85/ton for direct air capture with storage
• Up to $60/ton for other CCS applications
• Supports Crusoe/Tallgrass Wyoming project (carbon capture)
• Enables carbon removal economics
• Critical for net negative pathways

state policies

Renewable Portfolio Standards (RPS)

• Mandate percentage of electricity from renewables
• Examples: California 100% by 2045, New Mexico 100% by 2045
• Drives utility renewable procurement
• Benefits datacenter renewable access
• Varies widely by state

Clean Energy Standards

• Broader than RPS (includes nuclear, CCS)
• Supports 24/7 CFE goals
• Examples: Virginia, Washington
• Enables diverse low-carbon pathways

Carbon Pricing

• State and regional programs (RGGI in Northeast, California cap-and-trade)
• Increases fossil fuel electricity costs
• Improves renewable economics
• Creates incentive for efficiency
• Limited geographic scope in US

corporate disclosure requirements

SEC Climate Disclosure

• Proposed rules requiring Scope 1, 2, 3 disclosure
• Material climate risks
• Targets and goals disclosure
• Assurance requirements
• Timeline uncertain but momentum strong

CDP and TCFD

• Voluntary disclosure frameworks
• Many large corporations participate
• Investor pressure for transparency
• Renewable energy and efficiency disclosure
• Competitive differentiation

Science-Based Targets Initiative (SBTi)

• Corporate climate commitments aligned with Paris Agreement
• Scope 1, 2, 3 targets
• Third-party validation
• Growing corporate adoption
• Datacenter operators increasingly participating

conclusion

The datacenter industry faces a carbon paradox: enormous and growing emissions from 131 GW of capacity consuming 1,150 TWh annually, yet simultaneously leading corporate decarbonization with 37.4% renewable capacity, 58% PUE improvement (2.5+ to 1.06), and pioneering 24/7 carbon-free energy pathways. Scope 2 operational emissions dominate (95%+), driving renewable energy commitments from Google, Meta, Microsoft, and Amazon aggregating tens of gigawatts.

Embodied carbon emerges as growing concern, with construction-phase emissions reaching 5-20% of lifecycle totals. Innovations like Meta DeKalb’s AI-developed low-carbon concrete (40% reduction) and extended equipment lifecycles address Scope 3. PUE leaders—Meta Prineville (1.06), Prometheus Wyoming (1.08), Edged facilities (1.15 waterless)—demonstrate extraordinary efficiency gains.

Net zero pathways bifurcate: annual renewable matching achieves operational net zero (most hyperscalers by 2030), while Google’s 24/7 carbon-free energy goal represents next frontier requiring baseload carbon-free (nuclear, geothermal) and battery storage. Carbon-negative ambitions (Microsoft removing all historical emissions by 2050) push beyond net zero to climate repair.

Short-term (2025-2027) trajectories feature renewable scaling, PUE approaching 1.1, and Scope 3 disclosure. Medium-term (2027-2030) breakthroughs include 24/7 CFE achievement, enhanced geothermal and SMR deployments, and widespread operational net zero. Long-term (2030+) visions encompass carbon-free computing, comprehensive Scope 3 decarbonization, and sector carbon negativity.

Policy drivers—federal ITC/PTC/45Q credits, state RPS mandates, SEC disclosure requirements, and carbon pricing—accelerate transition. As computing demand grows exponentially (AI driving 50-200+ kW racks), efficiency and renewables must outpace consumption growth. The industry’s carbon trajectory will define climate impact of digital transformation, with datacenter decarbonization leadership potentially catalyzing broader economy-wide transition.