The rapid rise of AI is transforming data centers from quiet background utilities into headline energy consumers. As workloads grow, so does scrutiny of their power use. Yet, as this panel of leaders from Amazon Web Services, Westinghouse, Synopsys, and renewable developers emphasized, data centers are less an energy problem than a catalyst for long-overdue modernization of the grid.
For decades, U.S. electricity demand was flat, aided by efficiency gains and demand-side management. Grid investment lagged. Now, AI-driven data centers are finally exposing the cost of that neglect—and often being blamed for it. In reality, they are rapidly becoming anchor customers that help fund grid upgrades and accelerate the transition to cleaner, more resilient infrastructure.
New analysis of Amazon’s data centers, for example, shows that in several U.S. states the average facility generates millions of dollars in surplus for utility partners, money that can be reinvested in grid modernization. As AI demand grows, those surplus benefits are expected to increase, not shrink.
Creating sustainable AI-ready infrastructure is as much a regulatory challenge as a technological one. The panelists converged on a clear message: permitting reform and market redesign are now critical enablers of both clean energy and digital infrastructure.
At the federal level, bipartisan support exists for permitting reform, but legislation remains stalled. Without streamlined processes, clean generation, transmission, and next-generation nuclear all move too slowly to keep up with AI demand. The same problem appears in interconnection queues at regional transmission operators (RTOs), which were designed for incremental, broad-based load growth—not for large, concentrated, always-on data center clusters.
States are beginning to innovate around these constraints. Ohio’s “shot clock” legislation limits the duration of permitting processes, and Pennsylvania is digitizing approvals to increase transparency and accountability. Similar approaches could dramatically reduce timelines across the country.
For executives, the implication is clear: policy engagement is now a core component of infrastructure strategy. Organizations that help shape pragmatic frameworks—rather than simply react to them—will secure earlier access to capacity and more favorable conditions.
Pittsburgh illustrates what an “AI-ready, sustainable” region can look like when legacy assets, engineering talent, and public–private partnerships align. Once defined by heavy industry, the region has quietly become an epicenter for AI, robotics, and advanced energy.
Westinghouse’s long-standing presence has seeded a deep ecosystem in nuclear engineering and power systems. Institutions like Carnegie Mellon, the University of Pittsburgh, and Penn State supply a continuous pipeline of talent and research. Synopsys (via its ANSYS heritage) brings decades of simulation expertise—from chip-level cooling to full-facility optimization—directly into the design of next-generation data centers and their supporting infrastructure.
Equally important is the collaborative fabric: regional councils and industry networks that convene utilities, hyperscalers, manufacturers, and startups. These networks are becoming “force multipliers,” translating innovations from domains such as EV battery storage or locomotive power systems into data center-ready solutions.
For other regions, the lesson is not to copy Pittsburgh’s assets but to deliberately map and mobilize their own—ports, transmission corridors, manufacturing clusters, or research institutions—as foundations for sustainable digital infrastructure.
A truly sustainable data center will not be powered by any single technology. The emerging model is a “tango” between nuclear, renewables, and advanced storage, coordinated by intelligent grid and facility controls.
Large-scale reactors such as Westinghouse’s AP1000 are now operating in multiple locations worldwide, providing carbon-free baseload power. The next decade is likely to see the first commercial wave of small modular reactors (SMRs) co-located near data center clusters, especially where nuclear-friendly communities and existing plants already exist. At the same time, solar, wind, and storage continue to expand, and long-duration energy storage (LDES), such as 100-hour iron-air batteries, promises to replace diesel backup with cleaner, more resilient options.
Behind-the-meter generation will play a role, but not universally. Many operators still require grid backup, triggering regulatory debates about network charges and cost allocation. The likely path forward is a portfolio of options—front-of-the-meter, behind-the-meter, and hybrid arrangements—supported by clearer rules rather than one-size-fits-all mandates.
Twenty years ago, data centers behaved like “toasters”—relatively simple, predictable electrical loads. Today, they resemble supercomputers: dense, thermally complex, and highly dynamic, especially with GPU-driven AI workloads. That shift makes traditional, spreadsheet-based planning untenable.
Panelists stressed the growing importance of “physically accurate digital twins” that model everything from chip-level liquid cooling to full-site thermal management and regional grid behavior. These tools enable stakeholders to simulate different designs, locations, weather profiles, and technology refresh cycles before committing capital.
Digital twins are not just for operators. Utilities, regulators, vendors, and policymakers can use shared reference architectures and models—such as those under development at the U.S. Department of Energy—to align assumptions and accelerate approvals. When everyone can see how a proposed cluster will interact with the grid, with local climate, and with future chip generations, decisions become faster and less contentious.
In parallel, AI will increasingly manage operations in real time—optimizing power flows, predicting failures, and dynamically adjusting cooling and compute allocation. But those capabilities only reach their potential when the underlying processes and infrastructure are digitized first.
Perhaps the most overlooked dimension of sustainable AI infrastructure is its impact on communities. Many of the regions now attracting data center and energy investments are places whose traditional industries—coal, textiles, heavy manufacturing—have declined, eroding jobs and tax bases.
Developers are targeting legacy coal plants and other industrial sites because they already have land, water, and grid interconnections. Repowering these sites—potentially with nuclear, gas, renewables, and storage—alongside data centers can rebuild local tax bases and create diversified employment, from electricians to AI data scientists.
Hyperscalers like Amazon report similar dynamics in rural counties across multiple states, where data center and energy investments are bringing new jobs and public revenues. Yet misinformation persists, particularly around water usage and environmental impacts. In reality, many modern facilities use less water annually than a typical restaurant, thanks to efficient evaporative cooling and careful siting.
The panel’s closing message was unambiguous: the future of AI infrastructure will be connected, autonomous, and intelligent—but it must also be inclusive. Sustainable data centers are not just about electrons and emissions. They are about rebuilding places, modernizing systems, and aligning public and private interests around a shared, AI-enabled future.