Scope 2 Emissions

The GHG Protocol requires dual reporting using two distinct calculation methods. The location-based method applies the average grid emission factor for the region where consumption occurs, reflecting the physical reality of the grid. The market-based method accounts for contractual instruments like Power Purchase Agreements (PPAs) and Renewable Energy Certificates (RECs), allowing organizations to report lower emissions if they have purchased clean energy. This dual approach prevents double-counting of renewable attributes while incentivizing corporate procurement.

When an enterprise runs AI workloads on AWS, Azure, or GCP, the electricity consumed by the physical servers is classified as Scope 2 for the cloud provider. For the enterprise customer, this is technically a Scope 3 Category 1 (Purchased Goods and Services) emission. However, leading cloud providers now offer granular carbon accounting tools that allow customers to treat cloud electricity as Scope 2 by reporting on a market-basis, effectively transferring the renewable energy attributes of the provider's PPAs to the customer's inventory.

PUE is the critical multiplier that converts IT equipment energy into total facility energy. A PUE of 1.0 means 100% of energy reaches the servers; a PUE of 1.5 means 50% overhead for cooling and power distribution. When calculating Scope 2 emissions for AI training, the formula is:

• Total Energy = IT Load × PUE
• Emissions = Total Energy × Grid Emission Factor Hyperscale cloud providers typically achieve PUEs of 1.1-1.2, while enterprise data centers often operate at 1.5-1.8, making cloud migration a significant Scope 2 reduction lever.

Traditional annual REC matching allows a company to claim 100% renewable energy by purchasing certificates that may not align temporally with actual consumption. 24/7 Carbon-Free Energy (CFE) is an advanced procurement paradigm requiring that every kilowatt-hour of electricity be matched with carbon-free generation on an hourly basis. For AI workloads that run continuously, this requires a portfolio of wind, solar, battery storage, and firm clean resources like geothermal or advanced nuclear to cover the full demand profile.

Using average grid emission factors can obscure the true carbon impact of flexible AI workloads. The marginal emissions rate measures the carbon intensity of the specific power plant that must ramp up or down to meet a change in demand. For carbon-aware scheduling, shifting a training job from a peak-demand hour (when a gas peaker plant is on the margin) to a low-demand hour (when wind is curtailed) can reduce effective Scope 2 emissions by 50-80% compared to the average rate, even without changing total energy consumption.

The GHG Protocol Scope 2 Guidance establishes eight quality criteria for contractual instruments used in market-based reporting. Key requirements include:

• Direct contracts: PPAs must be directly between the buyer and generator
• Delivered electricity: Instruments must be sourced from the same market as consumption
• Vintage: Generation must occur within 21 months of the reporting year
• Unique claims: No double-counting of attributes
• Retirement: Certificates must be retired on behalf of the claiming entity Failure to meet these criteria requires reverting to the location-based method.