Over the next five years, data centers in the ERCOT region, which covers most of Texas, will experience significant growth. This dynamic is changing how a network designed for millions of dispersed consumers plans its expansion, increasing the weight of large customers. The terms of electrical grid planning are becoming the terrain of growing negotiations between regulators, utilities, and hyperscalers.
The essentials
- According to the IEA, data centers will account for roughly 50% of electricity demand growth in the United States through 2030.
- In the ERCOT region (Texas), data center load is expected to grow significantly between 2025 and 2030.
- Large technology customers are negotiating special terms for network access, modifying traditional interconnection processes.
- This concentration of purchasing power raises a governance question: the growing involvement of private actors in grid investment creates tensions with the capacity for public planning.
Texas as a laboratory for a global shift
ERCOT is a special case in the American electrical landscape. The Texas network operates largely outside federal FERC regulation and operates in a liberalized electricity market where prices fluctuate in real time based on supply and demand. This architecture has attracted hyperscalers precisely because it offers them flexibility: direct market access, the ability to negotiate bilateral contracts, and an abundance of cheap land with space to build.
The ERCOT network is recording rapid growth in data center-related load. ERCOT counts 232,546 MW of large load requests for 2030, of which roughly 72% are classified as data centers, reflecting the scale of the phenomenon without establishing a centralized forecast by a single source.
What makes this progression different from previous cycles of industrial electrification is its concentration. AI data centers consume massive amounts of electricity, but they are owned by five or six companies. When a single investment decision—building a 500 MW campus—can shift a regional network’s demand curve by several percentage points, the power dynamic with grid planners fundamentally changes.
The private buyer dictates terms to the network
Traditional electrical planning rests on a logic of pooling. Thousands of consumers, millions of households, diverse industries: utilities and regulators aggregate these demands, model their growth, and invest accordingly in lines, transformers, and power plants. Demand is diffuse, predictable at a statistical scale, and no individual has negotiating power.
Large technology customers are not ordinary consumers. When one of them plans a data center campus, it represents a load equivalent to a medium-sized city. This scale gives them increased weight: large customers go through a specific, regulated interconnection process that can involve transmission operators and ERCOT, according to NERC standards and established protocols.
Co-location arrangements may aim to speed up access, but they are not established as a legal way around interconnection procedures. In the United States, the interconnection queue is notorious for its delays. Large technology buyers have the resources to finance or co-finance infrastructure and offer long-term load guarantees, which gives them a competitive advantage in grid access. This phenomenon of concentrated purchasing power creates asymmetries in access to interconnection procedures.
S&P Global, in its Grid Edge Outlook 2026, points to the tensions created by the growing involvement of large private customers in grid investment. Utilities find themselves in a paradoxical position: they need these large customers to justify infrastructure investments, but these same customers possess significant resources that change the terms of negotiation.
Three concrete consequences for the grid
The first consequence is potential pressure on prices for other users. Negotiated rates or special interconnection arrangements may create a risk of cost transfer to other customers, but regulators seek to avoid this through tariff mechanisms. An effective increase in load for other users is not established as a general fact.
The second consequence concerns grid reliability. AI data centers have very specific consumption profiles: massive and relatively stable loads, interrupted by peaks during model training phases. ERCOT, which came close to rolling blackouts during the February 2021 winter storm, must integrate these new load curves into its operational management. The flexibility that hyperscalers promise—temporarily cutting their consumption during peaks—remains more theoretical than proven at large scale for now.
The third consequence is geographic. Data centers concentrate in specific corridors: the Dallas-Austin-San Antonio triangle in Texas, Northern Virginia in the United States, similar zones in Canada. These concentrations create local bottlenecks on networks designed to distribute energy more evenly. Grid reinforcement investments follow this geography, to the detriment of other regions.
The State’s capacity to steer the energy transition in question
The technical trajectory intersects here with a fundamental governance question. The energy transition requires planning: deciding which energy sources to develop as a priority, where to build transmission lines, which uses to favor. This planning rests on the capacity of states and regulators to define collective priorities and impose them on market actors.
But the rise of hyperscalers as structuring players in the grid creates tension with this capacity. When a significant share of electricity growth is driven by a limited number of private companies that participate in negotiating their access conditions, collective planning risks fragmenting. Priorities become those of the most powerful buyers.
The scenario that worries grid planners goes like this. Hyperscalers, pressed by their carbon neutrality commitments, have massively invested in Power Purchase Agreements for wind turbines and solar panels. These investments are real and significant. But they follow a corporate portfolio logic—achieving carbon neutrality on paper at the company level—rather than a network logic. A solar power plant built to supply a Google data center in Texas does not necessarily solve the problem of Texas grid reliability during winter peak hours.
Decarbonizing the grid and decarbonizing the company are two distinct objectives, which may or may not align.
The alternative scenario, more favorable, hinges on the financial power of the hyperscalers themselves. Some large technology customers have resources to participate in co-financing grid infrastructure. Some are already participating in co-financing grid infrastructure or storage projects that benefit the overall system. If regulators succeed in conditioning bilateral agreements on counterparties of general interest—shared grid reinforcement, additional storage, support for civil nuclear deployment—the power of hyperscalers can become a lever for the transition rather than an obstacle.
The IEA forecasts in its Electricity 2026 report strong demand growth linked notably to data centers and calls for increased investment in grid and decarbonized capacity. The question is whether regulatory frameworks allow this demand to be harnessed in service of collective objectives.
The regulatory levers still available
The window for regulatory action exists, but it is closing as private actors consolidate their positions. Several avenues are being explored in the United States and Canada.
The first is strengthening requirements for counterparties in interconnection agreements. Some states are considering conditioning priority grid access to firm commitments on consumption flexibility, co-financing of shared infrastructure, or quotas of verifiable decarbonized energy in real time, not on averaged annual bases.
The second avenue concerns the transparency of demand forecasts. ERCOT has improved its information-gathering mechanisms with large consumers, but announcements of data center projects often remain opaque until construction begins. Better visibility would allow for less reactive grid planning and better coordination with renewable energy deployment.
The third avenue is the emergence of a more structured flexibility market. If data centers can modulate their consumption, defer certain training loads to off-peak hours, temporarily shut down non-critical systems during peaks, this flexibility has considerable economic value for grid stability. Well-designed market mechanisms could remunerate this flexibility and align hyperscaler incentives with the needs of the electrical system.
These three avenues require one thing: that regulators have the expertise and institutional capacity to negotiate on equal terms with companies whose legal and technical teams are among the best equipped in the world. It is a public competence challenge as much as a regulatory one. On this front, the question resembles other governance challenges posed by large digital platforms—an asymmetry of expertise and resources that states are seeking to bridge with uneven results.
Nuclear as an equilibrium variable
One fact changes the equation in a potentially favorable direction. Hyperscalers themselves have begun financing civil nuclear development. Microsoft concluded a deal to buy power from the Three Mile Island plant, reopened for this purpose. Google and Amazon have announced investments in small modular reactors (SMRs). These commitments are still at the stage of industrial bet: SMRs are not deployed at large commercial scale, and construction timelines remain uncertain.
But the logic is sound. Data centers need reliable 24/7 electrical supply from the grid, backup resources, and a generation portfolio that can include variable resources. Nuclear provides firm 24/7 electricity and can complement a renewable mix to meet this need. If hyperscalers become anchored customers of civil nuclear, they provide the long-term demand guarantee these projects desperately need to secure financing.
It is one of the signals to watch to assess whether the rise of hyperscalers in electrical grid planning is a threat to the transition or, in certain configurations, a resource to accelerate it. The answer will depend less on companies’ intentions than on the quality of the frameworks regulators manage to build in the period ahead.
The extent to which states are prepared to reclaim this planning power, and the instruments they intend to deploy, remains to be defined.
Sources
- RBC Capital Markets, Global Power Demand Reshapes Infrastructure Investment for 2026, February 2026, https://www.rbccm.com/en/insights/2026/02/global-power-demand-reshapes-infrastructure-investment-for-2026
- S&P Global, Grid Edge Outlook 2026, no link (proprietary report)
- ERCOT, Long-Term Load Forecasts and Interconnection Queue Data, https://www.ercot.com/gridinfo/load/load_hist
- International Energy Agency (IEA), Electricity 2026, no link (report available on iea.org)
