Industry pressure has derailed a proposed emissions accounting standard that could have forced data center operators to match fossil fuel consumption with clean energy on a much tighter schedule. The Science Based Targets initiative (SBTi) opted not to adopt a protocol that would have required time- and location-based matching of renewable energy certificates to actual power usage. According to The Financial Times, the dropped rule could have cut CO₂ output “dozens of times faster” than current practices.

The core technical tension is straightforward: many new AI data centers—built by Amazon, Meta, and others—rely on on-site gas turbines when grid supply falls short. To avoid direct pollution penalties, operators purchase renewable energy certificates (RECs) from projects often in different regions or time zones. For example, a Texas facility running on gas at night can claim offsets from California solar generation logged during daylight hours. The Greenhouse Gas Protocol (GGP) argued this creates no credible link between consumption and generation, and proposed that certificates represent clean energy produced in the same market and within the same hourly window.

A lobbying coalition representing nearly $5 trillion in revenue—including Apple, Amazon, and GM—pushed back under the banner “May not Shall,” arguing that strict temporal and geographic matching would be too onerous and could chill clean energy investment. Google, notably, supported hourly matching, consistent with its position as the world’s largest corporate renewable buyer.

Research from Princeton’s Low-Carbon Technology Consortium and the EU shows that hourly matching could reduce emissions far faster than the current annual, region-agnostic system. The decision leaves data center operators free to continue using low-fidelity offsets, with no architectural pressure to co-locate generation or schedule compute loads to align with renewable availability.

Source: Engadget