Data Centers and the 24/7 Carbon-Free Energy Challenge: Beyond Annual RECs

The data center industry is facing a carbon paradox. The same artificial intelligence workloads that promise to optimize energy systems, accelerate drug discovery, and transform industrial operations are simultaneously driving an unprecedented surge in electricity demand. Global data center energy consumption is projected to exceed 1,000 terawatt-hours annually by 2028, roughly equivalent to the entire electricity consumption of Japan. This growth trajectory is colliding with corporate carbon commitments, regulatory scrutiny, and an uncomfortable truth: the way most data centers account for their carbon footprint is fundamentally misleading.

This article examines why annual Renewable Energy Certificate procurement creates a false picture of data center carbon performance, what 24/7 carbon-free energy actually requires, and how continuous energy and emissions monitoring transforms data center carbon accounting from an annual reporting exercise into a real-time operational capability.

The AI-Driven Energy Surge

The relationship between artificial intelligence and energy consumption is not linear -- it is exponential. Training a large language model can consume as much electricity as hundreds of American homes use in a year. But training is the smaller portion of the total energy footprint; inference -- the ongoing computation required to serve billions of queries, generate images, write code, and power AI assistants -- consumes far more energy cumulatively because it runs continuously at scale.

The major hyperscale operators -- Google, Microsoft, Amazon, Meta -- have each announced data center construction programs measured in tens of billions of dollars. Microsoft alone has committed to approximately $80 billion in data center investment for fiscal year 2025. These facilities require gigawatts of power capacity, equivalent to small cities, and they need that power to be available 24 hours a day, 365 days a year, with near-perfect reliability.

This growth is happening in a context where all of these companies have made ambitious climate commitments. Google has pledged to run on 24/7 carbon-free energy by 2030. Microsoft has committed to being carbon negative by 2030. Amazon Web Services has targeted 100 percent renewable energy by 2025. But the mechanisms these companies use to account for their energy-related carbon emissions deserve careful scrutiny.

The Inadequacy of Annual REC Procurement

The dominant approach to data center renewable energy accounting relies on the purchase of Renewable Energy Certificates, also known as Guarantees of Origin in Europe or International RECs (I-RECs) in other markets. Under this framework, a data center operator purchases certificates equivalent to the total electricity consumed over a year. If a data center consumes 1,000 GWh annually, the operator purchases 1,000 GWh worth of renewable energy certificates. On paper, the facility is "100% renewable."

The problem is that this annual matching approach bears little relationship to the actual carbon emissions associated with the facility's electricity consumption. A data center in Virginia that purchases wind energy certificates from Texas is not running on wind power. It is running on whatever the Virginia grid delivers at any given moment -- which at midnight in January might be 70 percent natural gas and coal. The certificates are an accounting instrument, not a physical energy flow.

The temporal mismatch is the most obvious flaw. Solar certificates are generated during daylight hours; data centers operate around the clock. Wind certificates are generated when the wind blows; data centers run regardless. When a data center claims 100 percent renewable energy based on annual matching, it is almost certainly running on fossil-generated electricity during periods when renewable generation is low. The annual aggregate may balance, but the actual emissions do not.

The geographic mismatch is equally problematic. Most certificate markets allow cross-regional trading, meaning a data center in one grid zone can claim renewable energy generated in an entirely different grid zone, hundreds or thousands of kilometers away. The renewable electrons never reach the data center. The local grid from which the data center actually draws power receives no additional renewable energy. The certificates function as a financial instrument, not an emissions reduction mechanism.

Google, Microsoft, and the 24/7 CFE Commitment

To their credit, both Google and Microsoft have recognized the limitations of annual REC matching and have committed to a more rigorous standard: 24/7 Carbon-Free Energy (CFE). This commitment requires matching electricity consumption with carbon-free energy generation on an hourly basis, within the same grid region where the data center operates.

Google's 24/7 CFE program, launched in 2021 and expanded since, aims to match every hour of electricity consumption at every data center with carbon-free energy by 2030. The company reports a CFE percentage for each data center campus, measuring the fraction of hours in which carbon-free energy supply meets or exceeds the facility's demand. As of their most recent reporting, Google's global fleet averages approximately 64 percent 24/7 CFE, with individual campuses ranging from over 90 percent (in regions with abundant hydro and wind) to below 30 percent (in regions dominated by fossil generation).

Microsoft has taken a complementary approach, committing to 100 percent renewable energy by 2025 (on an annual basis) while simultaneously pursuing hourly matching for its largest campuses. Microsoft has also invested heavily in next-generation clean energy sources, including nuclear (through agreements with nuclear developers and the restart of Three Mile Island Unit 1) and advanced geothermal, recognizing that solar and wind alone cannot deliver 24/7 carbon-free energy.

From Annual to Hourly: The Granular Certificate Revolution

The shift from annual to hourly energy matching requires a corresponding evolution in the certificate and tracking infrastructure. Annual RECs are generated for each megawatt-hour of renewable energy produced over a settlement period, typically monthly or annually. They carry minimal temporal and spatial granularity. Hourly matching requires certificates that specify exactly when and where the energy was generated, at hourly or sub-hourly resolution.

The EnergyTag initiative has emerged as the leading framework for granular energy certificates. EnergyTag defines a standard for time-stamped energy certificates that specify the generation time period (hourly or sub-hourly), the generation location (grid zone), the energy source, and the carbon content. These granular certificates enable the precise temporal and spatial matching that 24/7 CFE commitments require.

Several European countries have begun implementing or piloting granular certificate systems. Denmark, the Netherlands, and the UK have all taken steps toward hourly certificate issuance. The EU's revised Renewable Energy Directive (RED III) includes provisions that lay the groundwork for temporal matching requirements, particularly for green hydrogen production (where hourly matching is required to claim "renewable hydrogen" status under the Delegated Acts).

For data center operators, the shift to granular certificates changes procurement strategy fundamentally. Instead of purchasing the cheapest available certificates from any location and vintage, operators must procure certificates that match their consumption profile hour by hour, in their specific grid region. This typically requires a portfolio approach combining solar (daytime), wind (variable), baseload clean energy (nuclear, geothermal, or hydro), and potentially battery storage to cover gaps.

Scope 2: Market-Based vs. Location-Based Accounting

The GHG Protocol Scope 2 Guidance provides two methods for accounting for electricity-related emissions. The location-based method uses average grid emission factors to calculate emissions based on where the electricity is consumed. The market-based method allows companies to claim reduced emissions by purchasing renewable energy certificates or through direct procurement contracts.

This dual-reporting framework creates a significant disclosure gap for data centers. A facility in a coal-heavy grid region that purchases wind certificates from a clean grid region can report near-zero Scope 2 emissions under the market-based method while having substantial location-based emissions. The GHG Protocol requires companies to report both methods, but many companies emphasize the market-based figure in their public communications, creating a misleading impression of their actual impact on the electricity system.

The 24/7 CFE approach partially resolves this tension by requiring temporal and geographic alignment between clean energy procurement and actual consumption. When a data center achieves a high 24/7 CFE percentage, the gap between its market-based and location-based Scope 2 emissions narrows, because the clean energy is actually displacing fossil generation in the same grid during the same hours. This alignment is what makes 24/7 CFE a more credible carbon claim than annual matching.

Scope 3: The Embodied Carbon of Digital Infrastructure

While energy-related Scope 2 emissions dominate the data center carbon conversation, Scope 3 embodied emissions represent a growing and increasingly scrutinized portion of the total footprint. Scope 3 for data centers includes the embodied carbon of servers, networking equipment, storage devices, cooling systems, electrical infrastructure, and the building structure itself.

A modern hyperscale data center may contain hundreds of thousands of servers, each of which has a manufacturing carbon footprint that includes semiconductor fabrication (one of the most energy-intensive manufacturing processes in existence), printed circuit board assembly, housing and chassis production, and global logistics. The embodied carbon of a single server is estimated at 500 to 1,500 kg CO2-equivalent, and these servers are typically replaced on a 3 to 5 year refresh cycle.

For a hyperscale campus with 500,000 servers refreshed every four years, the annual embodied carbon from server procurement alone can reach 100,000 to 200,000 tonnes of CO2-equivalent -- a figure that rivals or exceeds the Scope 2 emissions of the same facility if it is located in a clean-grid region. As operators succeed in decarbonizing their electricity supply, embodied carbon becomes a proportionally larger share of the total footprint and demands corresponding attention.

The construction materials themselves -- steel, concrete, aluminium, copper cabling -- carry significant embodied carbon. A large data center campus may require 50,000 tonnes of steel, 200,000 tonnes of concrete, and thousands of tonnes of copper and aluminium. As the building products sector develops green premiums for low-carbon materials (as discussed in our companion article on mining and building products), data center operators have the opportunity to reduce embodied emissions through procurement specifications that require verified low-carbon materials.

EU Energy Efficiency Directive and SEC Climate Disclosure

Regulatory requirements for data center energy and carbon reporting are tightening on both sides of the Atlantic. The EU Energy Efficiency Directive (EED), as amended in 2023, introduced specific reporting requirements for data centers above 500 kW of installed IT capacity. These requirements include annual reporting of total energy consumption, PUE (Power Usage Effectiveness), use of renewable energy, water consumption for cooling, waste heat utilization, and temperature set points.

The EU requirements go beyond simple disclosure. Member states are directed to ensure that data centers implement energy efficiency measures and explore the utilization of waste heat. The Directive's reporting framework is designed to create a standardized, comparable dataset across EU data centers, enabling regulators to identify underperforming facilities and set future efficiency requirements. Data centers that cannot meet reporting deadlines or provide accurate data face regulatory action under national implementing legislation.

In the United States, the SEC's Climate Disclosure Rules (as implemented following litigation and revision) require publicly traded companies to disclose material climate-related risks and, for larger registrants, Scope 1 and Scope 2 GHG emissions. For hyperscale operators like Google (Alphabet), Microsoft, Amazon, and Meta, this creates a federal disclosure obligation for their data center energy emissions. For major colocation providers, the rules apply to their own operational emissions and create pressure from tenants to provide facility-level emissions data.

PUE as a Carbon Proxy vs. Actual Measured Emissions

Power Usage Effectiveness -- the ratio of total facility power to IT equipment power -- has been the dominant metric for data center energy efficiency since its introduction by The Green Grid in 2007. A PUE of 1.0 would indicate that all power goes to IT equipment with zero overhead; modern hyperscale facilities achieve PUEs of 1.1 to 1.2, while older enterprise facilities may have PUEs of 1.5 to 2.0.

PUE is a useful operational efficiency metric, but it is a poor proxy for carbon emissions. A facility with a PUE of 1.1 powered by coal-generated electricity has vastly higher carbon emissions than a facility with a PUE of 1.5 powered by hydroelectricity. PUE measures how efficiently a facility uses power, not how clean that power is. Yet PUE remains the most commonly reported data center environmental metric, partly because it is simple to measure and partly because it has been the default indicator for nearly two decades.

The transition from PUE-centric to emissions-centric reporting requires actual measurement of the carbon intensity of electricity consumed, hour by hour, combined with measured energy consumption at the facility, IT equipment, and cooling system levels. This is a fundamentally different data infrastructure requirement than simply measuring total power at two points (facility and IT) and computing a ratio.

Agentic AI for Real-Time Energy and Emissions Monitoring

The data infrastructure required for accurate data center carbon accounting -- hourly energy consumption by system, real-time grid carbon intensity, granular certificate tracking, and embodied carbon accounting -- exceeds what manual processes or traditional BMS (Building Management System) reporting can deliver. This is where agentic AI transforms the operational model.

Denominator's approach deploys autonomous monitoring agents that connect directly to the data center's electrical distribution infrastructure, BMS, and energy procurement systems. These agents operate continuously, performing several functions simultaneously:

• Energy monitoring agents track electricity consumption at the facility, UPS, PDU, and rack level, producing a granular consumption profile that captures temporal variations in load across the facility.
• Grid carbon agents ingest real-time grid marginal emission factor data for the facility's grid zone, computing the actual carbon intensity of consumed electricity for every hour of operation.
• Certificate matching agents reconcile energy consumption against the operator's portfolio of granular certificates, PPAs, and on-site generation to compute the 24/7 CFE percentage and identify unmatched hours.
• Scope 3 agents track the embodied carbon of IT equipment procurement, construction materials, and other upstream sources, maintaining a running inventory of the facility's total lifecycle emissions.
• Reporting agents automatically generate compliance packages for EU EED reporting, SEC climate disclosure, and voluntary frameworks like RE100 and the 24/7 CFE Compact.

The output is a continuous, auditable, blockchain-anchored record of the data center's actual carbon performance -- not an annual estimate, not a PUE ratio, but a measured emissions profile that reflects the reality of how the facility operates and what energy it consumes, hour by hour.

From Data Center Carbon to Cloud Service Carbon

The final link in the data center carbon accounting chain connects facility-level emissions to the carbon footprint of the digital services that run inside the facility. When an enterprise customer uses a cloud computing service, the carbon footprint of that service is a function of the energy consumed by the computation, the carbon intensity of that energy, and the embodied carbon of the hardware used.

Major cloud providers have begun providing carbon footprint tools for their customers. Google Cloud Carbon Footprint, Microsoft Emissions Impact Dashboard, and AWS Customer Carbon Footprint Tool all provide some level of workload-level carbon accounting. However, these tools typically rely on annual or monthly averages and facility-level allocations that do not reflect the actual carbon intensity of a specific computation at a specific time.

For enterprises with their own climate commitments, the accuracy of these cloud carbon footprint estimates matters. A company cannot credibly claim to have reduced its Scope 3 emissions from cloud computing without reliable, granular data about the carbon intensity of the cloud services it consumes. This creates a pull-through demand for the kind of facility-level, hourly carbon intensity data that continuous monitoring agents produce.

Denominator's Approach: Continuous, Automated, Verifier-Ready

The data center sector's carbon challenge is ultimately a data infrastructure challenge. The regulations are clear. The voluntary commitments are public. The methodologies are defined. What is missing in most facilities is the operational capability to measure, calculate, and verify carbon performance continuously at the granularity that credible claims require.

Denominator's agent-native platform addresses this gap by deploying autonomous agents that integrate with the data center's existing infrastructure -- BMS, electrical metering, energy procurement platforms, IT asset management systems -- and generate continuous, verifier-ready carbon intelligence. The agents handle the complexity of hourly grid carbon factor lookups, granular certificate reconciliation, embodied carbon tracking, and multi-framework regulatory reporting without requiring manual intervention.

For colocation providers, this capability extends to tenant-level carbon reporting, enabling customers to receive verified carbon intensity data for their specific deployments within the shared facility. For hyperscale operators, the agents scale across campus-level and fleet-level deployments, providing both facility-specific and aggregated carbon performance views.

The data center industry's path to credible carbon performance runs through the same fundamental requirement that applies to oil and gas, petrochemicals, and mining: verified, continuous, product-level carbon intelligence. For data centers, the "product" is computation, and its carbon footprint must be measured with the same rigor that industrial sectors apply to physical products. The operators that build this capability now -- moving beyond annual RECs to genuine 24/7 carbon-free energy monitoring -- will set the standard for an industry that is growing faster than any other on the planet.