Case study: how a hyperscale operator achieved carbon-free energy matching and what it learned

Why It Matters

Global data center electricity consumption reached an estimated 460 TWh in 2025, roughly 1.8 percent of worldwide electricity demand, and the International Energy Agency projects that figure could exceed 800 TWh by 2030 as artificial intelligence workloads scale (IEA, 2025). For hyperscale operators, traditional renewable energy procurement through annual matching of certificates and power purchase agreements no longer satisfies stakeholder expectations. Google pioneered the concept of 24/7 carbon-free energy (CFE) matching, setting the ambitious target of running every data center on carbon-free electricity every hour of every day by 2030. By mid-2025, Google reported achieving 64 percent average hourly CFE matching across its global fleet, up from 41 percent when it first began disclosing the metric in 2021 (Google, 2025). This case study examines what Google and other hyperscale operators have learned about the technology, procurement structures, and grid partnerships required to move from annual renewable matching to genuine around-the-clock clean electricity.

The shift matters because the gap between annual and hourly matching is substantial. A data center can claim 100 percent renewable energy on an annual basis while still drawing heavily from fossil-fueled grids during nighttime or low-wind periods. Hourly CFE matching forces operators to confront the temporal mismatch between variable renewable generation and constant compute demand, driving investment in energy storage, firm clean power, and advanced grid integration that benefits the broader electricity system.

Key Concepts

Annual matching vs. hourly matching. Under annual matching, an operator purchases enough renewable energy certificates (RECs) or unbundled guarantees of origin to cover total annual electricity consumption. This approach, used by RE100 signatories, does not guarantee that clean electrons flow to the facility at any given hour. Hourly CFE matching requires demonstrating that carbon-free sources supply the data center in every hour of the year, measured on a regional grid basis. The difference is analogous to paying off a credit card balance once per year versus settling every transaction as it occurs.

Carbon-free energy percentage (CFE%). Google defines CFE% as the fraction of each hour's electricity consumption covered by carbon-free generation on the local grid or through directly contracted supply. A facility with 90 percent CFE% still relies on fossil generation for 10 percent of hours. The metric captures not just solar and wind but also nuclear, geothermal, and battery storage discharge (Google, 2024).

Time-matched procurement instruments. Moving beyond standard PPAs, operators use 24/7 energy attribute certificates (EACs), hourly shape-matching contracts, and dispatchable clean capacity agreements. Organizations like EnergyTag and the 24/7 Carbon-Free Energy Compact (convened by the United Nations and Sustainable Energy for All) are developing standardized frameworks for hourly certificate issuance and retirement (EnergyTag, 2025).

Grid carbon intensity. The achievable CFE% at any location depends heavily on the local grid's generation mix. A data center in Quebec (where hydropower supplies over 94 percent of generation) can achieve near-100 percent CFE with minimal incremental procurement. A facility in Poland or Indonesia, where coal dominates, faces a far steeper challenge. Hyperscale operators increasingly use grid carbon intensity as a factor in site selection, load shifting, and new capacity investment decisions.

Demand-side flexibility. Some workloads are deferrable: batch processing, model training, data replication, and backup jobs can be shifted to hours when clean energy is abundant. Google's Carbon-Intelligent Computing platform, deployed since 2021, shifts flexible workloads across time zones and hours, reducing the effective carbon intensity of compute without affecting latency-sensitive services (Google, 2023).

What's Working and What Isn't

Measurable progress in favorable grids. Google's data centers in Denmark, Finland, and Oregon achieved CFE% scores above 90 percent in 2024, driven by combinations of local wind, hydropower, and contracted solar-plus-storage (Google, 2025). Microsoft reported that its Swedish data center operations achieved 95 percent hourly clean energy matching through a combination of Nordic hydro and wind PPAs (Microsoft, 2025). These results demonstrate that hourly matching is technically achievable where the grid mix is favorable and procurement is well structured.

Storage and firm clean power remain gaps. Reaching the final 10 to 20 percent of hourly matching in less favorable grids requires significant investment in battery storage, long-duration energy storage, or firm clean power such as geothermal or advanced nuclear. Google signed a 500 MW agreement with Kairos Power for small modular reactor capacity to supply its data centers starting in the early 2030s (Google, 2024). The timeline for such deployments remains uncertain, and cost premiums are substantial. Grid-scale lithium-ion batteries currently provide 4 to 8 hours of storage, insufficient for multi-day lulls in renewable generation.

Standardization is emerging but incomplete. EnergyTag published its first Granular Certificate Standard in 2023, and pilot markets in Europe and the United States began issuing hourly EACs in 2024 (EnergyTag, 2025). However, most electricity markets still lack the metering infrastructure, registry systems, and regulatory frameworks to support granular certificate trading at scale. Without standardized hourly instruments, operators rely on bespoke bilateral contracts that are expensive to structure and difficult to audit independently.

Cost premiums and additionality tensions. Hourly CFE procurement costs 20 to 40 percent more than equivalent annual matching strategies, depending on location and technology mix (BloombergNEF, 2025). This premium raises questions about capital allocation: would the same dollars achieve greater emissions reductions if spent on adding new renewables to carbon-intensive grids rather than perfecting hourly matching in already-clean grids? Hyperscale operators argue that 24/7 CFE drives systemic grid decarbonization by creating demand for storage and firm clean power that benefits all grid users, not just data centers.

Water consumption draws scrutiny. Cooling is the second-largest environmental impact of data centers after electricity. Google disclosed that its global data centers consumed 6.1 billion gallons of water in 2024, a figure that grew 20 percent year-over-year driven by AI workload expansion (Google, 2025). Microsoft reported similar increases. Operators are investing in air-cooled and closed-loop cooling systems, but water intensity remains a growing concern, particularly in water-stressed regions. A comprehensive sustainability strategy must address both energy and water footprints.

AI workload growth threatens progress. Even as CFE% improves, absolute electricity consumption is rising sharply. Google's total data center electricity consumption grew 17 percent in 2024, and Microsoft's grew 22 percent (Google, 2025; Microsoft, 2025). If workload growth consistently outpaces efficiency gains and clean energy procurement, the absolute carbon footprint can increase even as the intensity metric improves. This tension makes demand-side efficiency, workload optimization, and hardware efficiency critical complements to clean energy procurement.

Key Players

Established Leaders

• Google (Alphabet) — Pioneered 24/7 CFE matching methodology, operates 40+ data centers globally, targeting 100 percent hourly CFE by 2030 across its fleet.
• Microsoft — Committed to 100/100/0 clean energy goal (100 percent clean electricity, 100 percent of the time, zero carbon) by 2030 and is the world's largest corporate buyer of renewable energy.
• Amazon Web Services (AWS) — Largest corporate purchaser of renewable energy by capacity (over 28 GW contracted), transitioning from annual matching toward time-matched procurement.
• Equinix — Largest colocation provider globally with over 260 data centers, pursuing 100 percent renewable energy coverage and piloting hourly matching in European markets.

Emerging Startups

• Turntide Technologies — Develops AI-driven building management systems optimizing data center cooling and energy consumption with measurable efficiency gains.
• ZeroAvia / Kairos Power — Kairos Power developing small modular reactors contracted by Google for firm, carbon-free baseload supply to data centers.
• Lancium — Builds "clean campuses" colocating data centers with renewable generation and battery storage in Texas, offering flexible demand to the ERCOT grid.
• EnergyTag — Nonprofit developing the global standard for granular, time-stamped energy certificates enabling hourly CFE accounting.

Key Investors/Funders

• Breakthrough Energy Ventures (Bill Gates) — Invested in advanced nuclear, long-duration storage, and geothermal technologies critical to 24/7 CFE supply.
• Generate Capital — Infrastructure investor deploying capital into distributed energy and storage projects serving data center loads.
• TPG Rise Climate — Climate-focused fund investing in clean energy infrastructure at scale, including data center renewable supply.

Examples

Google's journey to 64 percent global CFE%. Google began tracking its hourly CFE% in 2021 at 41 percent across its global data center fleet. Through a combination of location-optimized wind and solar PPAs, battery storage procurement (including a 140 MWh system at its Nevada facility), and the Carbon-Intelligent Computing load-shifting platform, Google raised its fleet-wide CFE% to 64 percent by mid-2025 (Google, 2025). The company's highest-performing regions (Northern Europe and the Pacific Northwest) exceed 90 percent, while facilities in Asia and parts of the United States remain below 50 percent. Google's approach demonstrates that aggressive progress is possible in favorable grids, but reaching universal 24/7 coverage will require technologies and markets that do not yet exist at scale.

Microsoft's Sweden deployment. Microsoft's data center region in Gavle and Sandviken, Sweden, achieved 95 percent hourly CFE matching in 2024. The company leveraged the Nordic electricity market's high share of hydro and wind generation, supplemented by dedicated wind PPAs and battery storage contracts (Microsoft, 2025). Microsoft also partnered with the Swedish grid operator to provide demand response services, flexing non-critical compute loads to absorb excess renewable generation. The Swedish model illustrates how hyperscale operators can function as grid assets, providing flexible demand that supports renewable integration and grid stability.

Equinix's European hourly matching pilot. Equinix launched a pilot in 2024 to track and match electricity consumption at its Amsterdam and Frankfurt data centers on an hourly basis using EnergyTag-compliant granular certificates. Early results showed that standard annual PPA coverage translated to only 62 percent hourly CFE matching, revealing a 38 percentage-point gap that required incremental procurement of shaped wind and solar products and 50 MWh of co-located battery storage (Equinix, 2025). The pilot highlighted how annual matching overstates actual clean energy usage and provided Equinix with a roadmap for targeted investments in storage and dispatchable clean supply.

AWS and the 28 GW renewable portfolio. Amazon Web Services surpassed 28 GW of contracted renewable energy capacity globally by the end of 2025, making it the world's largest corporate renewable buyer by installed capacity (Amazon, 2025). While AWS primarily uses annual matching today, the company announced a phased transition to hourly matching in its European regions starting in 2026, partnering with grid operators and certificate registries to implement granular tracking. AWS's scale gives it considerable market power to shape procurement standards and drive down the cost of time-matched instruments.

Action Checklist

• Baseline your CFE%. Calculate hourly carbon-free energy matching for each data center location using local grid data and contracted supply. Identify the gap between annual and hourly coverage to prioritize investments.
• Adopt granular certificates. Transition from annual RECs or guarantees of origin to hourly energy attribute certificates where market infrastructure supports them. Engage with EnergyTag and regional registries to pilot time-stamped tracking.
• Invest in storage and firm clean power. Procure battery storage co-located with or dedicated to data center loads. Evaluate long-duration storage (iron-air, compressed air, thermal) and firm clean generation (geothermal, advanced nuclear) for baseload gaps.
• Implement demand-side flexibility. Classify workloads by latency sensitivity and shift deferrable tasks (training, batch analytics, replication) to hours of high CFE% using carbon-aware scheduling platforms.
• Integrate water stewardship. Measure and disclose water consumption per MWh of compute. Evaluate air-cooled and closed-loop cooling alternatives for water-stressed locations and set water-positive targets alongside energy goals.
• Engage with grid operators. Offer flexible demand response capabilities to local grid operators, converting data center load flexibility into a grid stability resource that supports higher renewable penetration.
• Set location-specific targets. Recognize that global averages mask regional variation. Set per-facility CFE% improvement targets and prioritize capital allocation toward the most carbon-intensive locations where marginal investment yields the greatest emissions reduction.

FAQ

What is the difference between 100 percent renewable and 24/7 carbon-free energy? A company claiming 100 percent renewable energy typically purchases enough renewable certificates annually to match total consumption. This means fossil-fueled electricity may still power the facility during many hours. 24/7 carbon-free energy requires matching consumption with clean generation in every hour, ensuring the facility actually runs on clean electricity around the clock. The latter is substantially more demanding and drives investment in storage and firm clean supply.

How much does 24/7 CFE matching cost compared to annual matching? BloombergNEF estimates that hourly CFE procurement costs 20 to 40 percent more than equivalent annual strategies, varying by grid region and technology availability (BloombergNEF, 2025). The premium stems from the need for storage, shaped supply contracts, and investments in grids with limited clean baseload. Costs are expected to decline as granular certificate markets mature and storage costs continue falling.

Can smaller data center operators pursue 24/7 CFE? Yes, though at smaller scale. Colocation providers like Equinix are piloting hourly matching, and the development of standardized granular certificate markets will make time-matched procurement accessible to mid-size operators. Smaller operators can also improve their effective CFE% by selecting sites with favorable grid mixes, investing in on-site solar and storage, and participating in aggregated procurement programs.

What role does AI play in data center energy management? AI optimizes cooling systems (reducing energy overhead by 10 to 30 percent in documented deployments), predicts renewable generation for scheduling, and shifts flexible workloads to low-carbon hours. Google's Carbon-Intelligent Computing and DeepMind-based cooling optimization are examples of AI directly reducing data center carbon intensity (Google, 2023). However, the growth of AI workloads is also the primary driver of rising data center electricity demand, creating a dual-edged dynamic.

Why is water consumption an emerging concern for data centers? Evaporative cooling is the most energy-efficient method for rejecting heat, but it consumes significant freshwater. Google's global data centers used 6.1 billion gallons of water in 2024 (Google, 2025). In water-stressed regions, this creates competition with agricultural and municipal users. Operators are transitioning to air-cooled and hybrid systems and setting water replenishment targets, but the trade-off between water-efficient and energy-efficient cooling remains an active design challenge.

Sources

• International Energy Agency. (2025). Data Centres and Data Transmission Networks: Tracking Report and Electricity Demand Projections. IEA.
• Google. (2025). 2025 Environmental Report: Carbon-Free Energy, Water Stewardship, and Data Center Sustainability. Alphabet / Google.
• Google. (2024). 24/7 Carbon-Free Energy: Methodology, Progress, and Kairos Power SMR Agreement. Google Cloud Blog.
• Google. (2023). Carbon-Intelligent Computing: Shifting Workloads to Reduce Carbon Intensity. Google Research.
• Microsoft. (2025). 2025 Sustainability Report: Data Center Energy, Water, and Waste Performance. Microsoft.
• Amazon. (2025). Amazon Sustainability Report: Renewable Energy Portfolio and Climate Pledge Progress. Amazon.
• BloombergNEF. (2025). 24/7 Clean Energy Procurement: Cost Premiums, Market Readiness, and Technology Gaps. BNEF.
• EnergyTag. (2025). Granular Certificate Standard: Framework for Hourly Energy Attribute Certificates. EnergyTag.
• Equinix. (2025). Hourly Carbon-Free Energy Matching Pilot: Amsterdam and Frankfurt Results and Methodology. Equinix Sustainability.
• Cambridge Centre for Alternative Finance. (2025). Cambridge Blockchain Network Sustainability Index and Data Center Electricity Benchmarks. University of Cambridge.