Turning Data Centers Into Grid Assets: Insights from Oak Ridge

Key Highlights & Insights:

• Rising Load: Data centers’ rapid growth — driven by AI and cloud services—is straining electric grids and outpacing traditional planning.

• Opportunity for Flexibility: Oak Ridge National Laboratory’s MEGA-DC project models how data centers can evolve from simple power consumers to grid-orchestrated assets that shift demand, absorb excess renewables, and support reliability.

• Four Integration Levels: Off-grid → Grid-connected → Grid-cognizant (predictable load) → Grid-orchestrated (active, two-way coordination).

• Regulatory Lag: Permitting, interconnection, and market rules (even with FERC Order 2222) struggle to keep pace with data center expansion.

• Potential Benefits: With real-time coordination, data centers could lower peak stress, reduce reliance on peaker plants, and enhance renewable utilization — turning a major challenge into a stabilizing force for the grid.

Following decades of steady 1% electric load growth per year, the U.S. electric grid has been forced into action — faced with the sudden reality of the seemingly insatiable energy appetite of data centers across the nation. A traditionally risk-averse, slow-moving industry, electricity providers had considered 1 MW to be a larger-than-usual strain on the grid. Driven primarily by solar and wind generation farms as well as electric vehicles, the Federal Energy Regulation Commission (FERC) issued Order 2222 — to create more distributed energy resource (DER)-based supply-side liquidity — recognizing the need for prosumer participation in decentralized energy generation, alleviation of cost pressures on maintenance and aging T&D networks.

Generative artificial intelligence (GenAI) and commercial construction companies are currently investing billions in the rapid buildout of data centers. With few incentives to align with the regulatory frameworks of the North American Electric Reliability Corporation (NERC) and FERC, many are bypassing the traditional role of electric utilities in generation, transmission, and distribution. As a result, some data centers are pursuing behind-the-meter generation, while others are planning to go fully off-grid in pursuit of energy self-reliance.

Most data centers still rely heavily on the grid, placing significant strain on utility operations, existing infrastructure, and ultimately on energy affordability for customers outside the data center industry. At the same time, states, counties, and utilities have a regulatory window of opportunity to reframe technology, research, and development to address the rapid load growth driven by AI. Yet regulatory bodies such as public utility commissions remain limited in their ability to match the influence of national utility monopolies in electricity, natural gas, and water. At the core of these challenges are the risks to consumer-level energy affordability and reliability.

In recognition of this complex challenge, Oak Ridge National Laboratory (ORNL) is spearheading the Modeling Energy Growth Associated with Data Centers (MEGA-DC) project. Demonstrating an emerging core capability and providing resilience as a service, ORNL is moving at the speed of the industry, collaborating with states, utilities and sponsors to win the AI wars with global competitors in what is like the Manhattan Project of this era.

Data Center Landscape

Because data center designs vary widely, the MEGA-DC project models facilities of all types, sizes, and locations. These range from small Bitcoin and edge centers to traditional enterprise sites, collocated GenAI training hubs, and large hyperscalers. Each build differs in its purpose, components, energy and cooling sources, and computing applications, among other factors.

Facing challenges tied to electricity, water, cooling, storage, broadband, and fiber, the data center industry is racing to boost efficiency and reduce dependence on utility infrastructure. Some enterprise facilities are refreshing systems and adopting newer technologies to improve performance and cut interdependencies. GenAI start-up data centers, focused on training and scaling AI agents for cloud infrastructure, are exploring collocation for energy, fiber, and broadband to strengthen security and redundancy. Large hyperscalers are also pursuing collocation, with an emphasis on circularity—testing innovations such as on-chip cooling to improve both power utilization effectiveness (PUE) and water utilization effectiveness (WUE).

Despite rapid innovation and a growing push for circularity, the industry still faces barriers from state energy offices, regulated utilities, adjacent industries, and public stakeholders. Across the country, concerns have been raised about land use, real estate availability and costs, supply chain pressures, and the environmental and noise impacts of data center siting and operations.

To ease the strain of rapid data center growth and its indirect impacts on the grid, ORNL is developing an integrated systems approach to help states, utilities, and other stakeholders transform data centers into grid-orchestrated assets. Through its MEGA-DC initiative, ORNL works with an advisory board and partners to shape scenario-based decision-making frameworks tailored to the needs of states and localities. In examining factors such as economic and market trends, regulatory constraints, energy affordability, and infrastructure resilience, four major scenarios have emerged across the data center landscape:

1. Off-grid data centers
2. Grid-connected data centers
3. Grid-cognizant data centers
4. Grid-orchestrated data centers.

Each of these scenarios will be modeled in MEGA-DC’s digital twin environment for every state and its counties.

Off-Grid Data Centers

To win the global AI race, competing U.S. innovators must focus on training agents faster on larger data libraries. The desire for speed and intensity of this development, however, greatly outpaces the electric grid industry. But, because limitations exist in areas such as energy price, water availability, real estate availability and cost, broadband and fiber connectivity as well as workforce availability, data centers are gaining momentum in behind-the-meter circularity that are, in turn, at odds with regular energy customers, policymakers and the electrical utilities themselves.

This move by private industry toward off-grid data centers, as seen recently with MARA Holdings’ acquisition of a 114-MW Bitcoin data center and subsequent off-grid transition, is an evolution that may alleviate the electric affordability concerns voiced by regulators.

While the industry’s innovation drives ahead, with the goal of sloughing off the reliance on utilities and other stakeholders for resource provision, market-related conflicts arise for some players. A deeper story emerges when the off-grid data center challenges the constructs and desires of investor-owned utilities (IOUs). Still, for municipalities, member-driven cooperatives and utility districts, unique and granular challenges exist — while service area and community impacts endure across resources, land use and affordability for a data center’s neighbors.

Grid-Connected Data Centers

In MEGA-DC, the grid-connected data center is a paradigm in which the facility itself is an opaque load to the grid — with the existing infrastructure serving as a primary or secondary energy source. Utilities see this as the traditional model where the data center may be subject to power-purchasing agreements to reduce their energy costs. Even with load-shifting or load-shedding clauses, these agreements still define the grid-straining data center as a single utility customer.

MEGA-DC considers this data center type as being closely aligned with the electric utility and, therefore, its regulatory construct. Yet, a multitude of concerns and challenges are being forced on the grid and its T&D infrastructure because of data center interconnectedness. States and utilities now are faced with a balancing act between providing for the data center’s energy needs and cascading impacts across service areas, counties and localities.

States such as Ohio, Washington and Indiana have begun to experiment with mitigating effects of data center strain on grid infrastructure and energy consumption, such as consumer energy affordability. For example, to offset the costs of infrastructure improvements, data centers may be subject to things like minimum energy purchase commitments requiring larger payments for their large load. However, regulatory overhaul’s time-consuming nature cannot outpace the industry. Furthermore, strategic innovation beyond the electrical grid is needed — as high-consumption data centers place strain on other resources, such as water and land. And, while the technology industry’s rapid advancements are reducing data center energy footprints, this progress has been just as quickly offset by the seemingly exponential proliferation of data centers across the U.S.

Grid-Cognizant Data Centers

MEGA-DC defines the grid-cognizant data centers as facilities with more predictable energy demands on the grid in performance of AI-based tasks. Even with this predictability, the challenges posed by a grid-cognizant data center become critical for those that are either larger in size or even in number. In a state with high data center concentration, conversations shift to a grid-wide network of hurdles to overcome, such as state energy reserves, interconnect queue wait times and transmission-side congestion analysis.

The transmission industry, a focal point of this scenario, is investigating grid-enhancing technologies like advanced conductors, dynamic line rating, grid inertia, series and shunt capacitance devices. Funding for new transmission lines exists, yet the permitting, development and deployment processes create barriers across state lines and regulatory environments, while impacting large customer populations and stakeholders. At a higher level, oversight from NERC, FERC and the interplay between the balancing authorities (that is, PJM Interconnection, Midcontinent Independent System Operator and Electric Reliability Council of Texas) add another layer of time-consuming coordination, consensus building and decision-making. Up against the GenAI players whose expediency is unmatched, T&D utilities are falling increasingly behind.

ORNL’s current intellectual property and research on the grid-cognizant data center class is prepared for pilot demonstration, where the upstream benefits of load cognizance to the electric utility can be translated into load-forecasting improvements at a distribution level.

Grid-Orchestrated Data Centers

Where traditional data centers operate as passive energy consumers with fixed, opaque load, having limited grid interaction, MEGA-DC’s ultimate vision is to empower states, utilities and customers by enabling the grid-orchestrated data center. Featuring real-time, two-way communication, this data center class would be actively managed as a grid asset, with a flexible, transparent load.

From an energy and load transparency perspective, there is industry precedence in game-changing innovation, such as DERs like solar and wind energy, batteries and vehicle-to-grid electric vehicles. Yet, to date, as data center players are 10-to-100 times bigger than their DER counterparts, with energy needs and consumption profiles to match, there is critical need for rapid research, development and deployment of a stabilizing force for the grid.

By leveraging MEGA-DC and managing large data center loads more flexibly, these facilities could play a vital role in supporting the grid—shifting load, reducing demand during peak times, and helping maintain balance. This stabilizing force within the GenAI industry could also drive market growth, supported by surplus energy generated through a network of collocated facilities across the U.S. Such an oversupply has the potential to reshape energy market dynamics, lower prices, and create new opportunities for improved energy affordability for utility customers.

Collaboration Welcomed

ORNL is accelerating U.S. competitiveness, leveraging existing and emerging core capabilities at the GenAI innovation frontier. To transform data centers into integrated grid assets and ensure reliable grid performance, the MEGA-DC team is engaged in a holistic approach for siting and leveraging of data centers — analyzing their AI-load profiles and optimizing their load flexibility and planning for transmission capacity — while maintaining resource availability, power provision and affordability for the customer.

Under the guidance and oversight of the MEGA-DC advisory board, ORNL is collaborating with states and industry to develop this multidisciplinary tool for electric grid resilience. As it progresses in its journey, ORNL welcomes new stakeholders and partners to provide feedback, ask questions and join in this exciting collaboration.