Powering AI: New Framework Aims to Break Data Center Gridlock

HOUSTON, TX – March 23, 2026 – A major bottleneck threatening to stall the artificial intelligence revolution is not a shortage of microchips or software, but a lack of power. To address this critical challenge, the Electric Power Research Institute (EPRI) today launched Flex MOSAIC, a landmark framework designed to slash the 'time to power' for the massive data centers that fuel the digital world.

Announced at the CERAWeek energy conference, the voluntary framework establishes a standardized, common language to define and classify the energy flexibility of large electric loads. Developed through EPRI's DCFlex initiative with over 65 utilities, tech giants, and regulators, Flex MOSAIC aims to build trust between grid operators and data center developers, potentially cutting years off grid connection timelines.

“As demand from AI and data centers grows at unprecedented speed, flexibility is becoming the third leg of the speed-to-power stool, alongside generation and transmission,” said EPRI President and CEO Arshad Mansoor. “This framework allows everyone — utilities, regulators, and large‑load developers — to have common language about flexibility and to trust what that language means. That shared understanding is essential to moving faster while maintaining reliability.”

The Multi-Year Wait for Power

The urgency for a solution like Flex MOSAIC is rooted in a growing crisis facing both the tech industry and the nation's power grid. The explosive growth of AI has triggered an unprecedented demand for electricity, and utilities are struggling to keep pace. The result is a massive logjam in interconnection queues, where data center projects can languish for years waiting for a connection.

Across the United States, the average wait time for large projects to connect to the grid has ballooned from under two years a decade ago to over four years today. In high-demand areas like Northern Virginia, a global data center hub, wait times can stretch to seven years or more. Some developers have reported receiving utility quotes for connection timelines as long as a decade.

This gridlock has tangible consequences. Tech giants like Amazon and Meta have publicly announced pauses on billion-dollar data center construction projects to reassess the feasibility of securing adequate power. The delays are so significant that the hardware and infrastructure planned for a new facility can become outdated before the lights are even turned on. This bottleneck is now widely considered the single biggest obstacle to data center expansion.

The grid itself, largely built for a different era, was not designed for the concentrated, continuous, and massive power draws of modern AI facilities. Utilities, using planning tools not built for this surge, are seeing load forecasts skyrocket, with some receiving more connection requests from data centers than their entire system can currently support.

Creating a Common Language for the Grid

Flex MOSAIC tackles this problem not by building new power lines, but by creating a new form of currency: a shared, credible definition of flexibility. The framework provides a uniform way to classify a data center's ability to adjust its power consumption based on four key characteristics: the magnitude of power reduction, the timing or speed of its response, the duration of the adjustment, and the frequency with which it can offer this flexibility.

By organizing these traits into a clear set of classes, a utility in Texas and a hyperscaler in California can use the exact same terminology to negotiate a connection agreement. This common language gives grid operators the confidence to integrate these huge loads without compromising reliability. Instead of planning for a data center's worst-case, maximum-power-draw scenario 24/7, they can now credibly account for its ability to ramp down during periods of peak system demand.

This shift is critical for accelerating deployment. “Flexibility is critical for fast access to power for data centers,” said Vladimir Troy, vice president of AI Infrastructure at NVIDIA. “By clearly defining flexibility, the Flex MOSAIC framework gives all parties the confidence needed to accelerate deployment and meet the growing needs of AI.”

Support from grid reliability organizations underscores the framework's importance. “As large, flexible loads play a growing role in the power system, having clear, technically grounded definitions of flexibility is critical for reliability,” stated North American Electric Reliability Corporation (NERC) President Jim Robb.

Unlocking Growth and Greening the Grid

The implications of unlocking this gridlock extend far beyond the tech industry. Faster data center deployment is a direct enabler of economic growth, powering innovations in science, medicine, and business. On a local level, data centers are significant economic engines, with a single facility contributing millions in tax revenue and creating both construction and high-tech operational jobs.

More importantly, a more flexible grid is a greener grid. Data centers are voracious energy consumers, projected to account for up to 17% of total U.S. electricity demand by 2030, up from about 4% today. Meeting this demand solely with new generation could require building an additional 25 to 50 gigawatts of fossil fuel power plants.

Load flexibility offers a powerful alternative. By enabling data centers to shift their consumption, utilities can better integrate intermittent renewable energy sources like wind and solar. A data center could, for instance, ramp up its computing tasks when solar power is abundant and cheap, and scale back during evening peak hours. Research suggests that if half of all data center demand were flexible, it could eliminate the need for nearly half of the new natural gas capacity that would otherwise be required.

State regulators, tasked with protecting consumers, are watching closely. “As demand from data centers accelerates, state regulators are focused on ensuring customers are not burdened by the costs of serving new, large loads, as well as maintaining grid reliability,” said Ann Rendahl, President of the National Association of Regulatory Utility Commissioners (NARUC). A standardized framework like Flex MOSAIC provides a tool to help evaluate these complex trade-offs.

A Piece of a Larger Puzzle

Flex MOSAIC arrives as part of a broader, industry-wide effort to modernize the grid and its connection processes. It complements other solutions, such as federal reforms aimed at streamlining interconnection queues and the increasing trend of data centers building on-site power generation with batteries, solar, and other resources.

While those solutions focus on the process of connecting or the source of the power, Flex MOSAIC provides the foundational language to define the behavior of the load itself. This standardization allows the flexibility offered by on-site batteries or dynamic workload scheduling to be quantified and integrated into system-wide planning.

The initiative launches with a powerful coalition of initial participants, including tech giants Google and Meta, and dozens of major utilities and grid operators such as Southern Company, Exelon, California ISO, and Midcontinent Independent System Operator (MISO). This broad collaboration between traditionally siloed industries signals a shared recognition that the future of digital infrastructure is inextricably linked to the evolution of the energy grid.