Global data-center (DC) IT capacity is projected to roughly double, from about 112 GW in 2024 to 224 GW by 2030, with electricity use rising from 854 TWh to nearly 1,900 TWh. Much of this buildout will need to comply with the GHG Protocol’s ongoing Scope 2 revisions, which shift toward hourly matching of supply and consumption rather than annual accounting. As grids decarbonize at different speeds, DC operators face a choice: build their own firm clean-power stack or buy it from a utility.
That choice increasingly depends on where a data center can get a grid connection. Ireland's Dublin stopped granting new connections between 2021 and 2025, and New York paused new hyperscale permits in 2026 after tariff disputes over rising bills. In 2022, AWS and Microsoft reportedly shifted planned capacity to London, Frankfurt, and Madrid. Markets with land, renewable resources, and a favorable policy ecosystem are natural landing spots for that relocated demand. India has begun positioning itself as a destination of choice.
Self-procurement is faster: the DC operator builds its own renewable energy and storage with full control over siting, technology, and timeline. But going alone necessarily means overbuilding. Each operator has to size its backup for the worst-case (highest) demand scenario, factoring in reliability requirements. Utilities serving many customers may not see coincident peaks with data centers. One shared storage pool could cover everyone, even with a smaller installed base, compared with each party building for its own needs. Self-procurement erodes this saving, competes for the same scarce copper and transformers everyone else is buying, increases the landed cost of energy, and tends to work only for the largest, best-funded operators.
A pooled utility tariff solves both problems: it spreads the cost across many customers and is open to buyers who could not self-procure at scale. But for that to work, data centers need their own rate class, priced for exactly what they use. A utility serving growing demand keeps contracting for new generation and storage at a falling marginal cost, allowing it to hold the tariff flat over a long horizon. One Indian state is exploring letting its utilities offer a round-the-clock green tariff, verified block by block.
Two other conditions matter just as much. The tariff needs a long-term minimum-take commitment, so if the data center’s demand doesn’t materialize, other customers aren’t left paying for capacity built to serve it. And its costs need to be settled against a cost study that the utility actually publishes. Usage should be metered hourly or sub-hourly, matching supply to demand in real time rather than netted annually. That way, a “round-the-clock clean” claim can be checked against real numbers instead of taken on trust.
A well-designed tariff lets a utility attract data-center investment and demand without making other consumers pay for it. A badly designed one does neither: it either drives the data centers elsewhere or leaves everyone else footing the bill.
Prateek Aggarwal is a Senior Programme Lead, and Karthik Ganesan is a Fellow and Director, Strategic Partnerships, at the Council on Energy, Environment and Water (CEEW). Views are personal.
Guest Authors
