The GRID Act: A Bipartisan Reckoning
On February 12, 2026, Senators Richard Blumenthal (D-CT) and Josh Hawley (R-MO) introduced the Guaranteeing Rate Insulation from Data Centers Act—better known as the GRID Act (S. 3852). In a political environment where agreement across party lines has become vanishingly rare, the bipartisan nature of the bill signals just how serious the problem has become.
The legislation targets any new data center with a power demand of 20 megawatts or more—a threshold that captures essentially every hyperscale facility operated by Amazon, Microsoft, Google, and Meta. Under the bill’s core provision, covered entities would be prohibited from drawing power from the shared electrical grid. Instead, they would be required to source all electricity—including backup power—from captive power plants, on-site generation, or entirely separate off-grid systems.
The penalty for non-compliance is steep: a minimum civil fine of $1 million per day of violation. Existing data centers would receive a ten-year transition window, during which they could maintain grid connections only by obtaining and renewing a “Zero Rate Effect Certificate”—a designation confirming their operations are not shifting costs onto ordinary utility customers.
Why This Bill Exists: The $1 Million-Per-Day Problem
The GRID Act didn’t emerge in a vacuum. It is a direct response to a growing crisis that has been quietly simmering for years and boiled over in 2025 and 2026: the exponential growth of AI infrastructure is overwhelming the electrical grid, and ordinary Americans are paying for it.
Data centers now consume between 3% and 4% of all U.S. electricity—a figure that the Department of Energy projects could reach 9% or higher by 2030. The problem isn’t just total consumption; it’s the speed and concentration of demand growth. A single large AI training cluster can draw as much power as a small city. When hyperscalers negotiate direct interconnection agreements with utilities, they often lock in priority access to grid capacity that was originally built and paid for by residential and commercial ratepayers.
The consequences are already playing out in communities across the country. In a striking example reported in May 2026, nearly 50,000 residents near Lake Tahoe faced potential power disruptions after a utility redirected transmission lines originally serving the region toward nearby data center facilities. Ohio suspended major tax incentives for data centers after projected exemption costs surged unexpectedly, with residents subsequently pushing a ballot initiative that could ban hyperscale data centers from the state entirely.
Parallel legislation is also moving through Congress. Senator Chris Van Hollen’s Power for the People Act would require data centers to pay the full cost of any grid capacity expansions their demand triggers, preventing cost-shifting to residential and small business customers. Senator Tom Cotton’s DATA Act takes a different approach, creating a new regulatory category called “consumer-regulated electric utilities” for fully off-grid power systems serving data centers—essentially building a legal framework to accelerate the transition the GRID Act mandates.
The Infrastructure Crisis: Half the Pipeline at Risk
Even setting aside the legislative threat, the data center construction boom is running into hard physical limits. Of the approximately 16 gigawatts of new data center capacity announced across the United States for 2026, analysts estimate that only around 5 GW is actually under construction and on track to meet its timeline. The rest—roughly 30% to 50% of the entire pipeline—faces delays of one to three years, or outright cancellation.
The bottleneck is not money. Capital is abundant. The constraint is equipment and grid interconnection. High-voltage transformers, which are critical for stepping down utility power to data-center-usable voltages, now carry lead times of 36 to 48 months in many cases—up from 12 to 18 months just three years ago. Switchgear assemblies, battery storage systems, and other critical electrical components face similar backlogs.
The supply crunch is partly a consequence of trade policy. Many of the components for these systems rely on Chinese manufacturing inputs, and ongoing tariff disputes and export control policies have introduced friction into supply chains that were already stretched thin. The irony is sharp: the same geopolitical tensions driving U.S. investment in domestic AI infrastructure are choking the supply chains needed to build it.
Grid interconnection queues compound the problem further. PJM Interconnection, the largest grid operator in the United States serving over 65 million people across 13 states, projects it will be six gigawatts short of its reliability requirements by 2027. Hyperscalers requesting new interconnection points are waiting four to six years in some regions simply to get a study completed—let alone approved and built out.
Community Opposition Goes Mainstream
If the supply chain and grid constraints represent the technical ceiling on data center growth, community opposition has emerged as an increasingly potent political floor. A 2025 analysis found that at least 25 data center projects were cancelled due to local pushback—a fourfold increase from the year prior. Maine became the first state to pass a statewide moratorium on large data center construction.
The complaints from residents are consistent: industrial-scale power consumption, water usage for cooling, visual blight, noise from backup generators, and the sense that local infrastructure is being commandeered for the benefit of trillion-dollar corporations that pay relatively little in local taxes.
Data center cancellations due to community opposition quadrupled between 2024 and 2025, and that momentum is continuing into 2026. In Ohio, where the legislature’s suspension of data center tax breaks made national headlines, a grassroots ballot campaign is actively seeking to give voters a direct say in whether the state can host hyperscale facilities at all.
What Comes Next
The GRID Act faces significant lobbying headwinds. The hyperscale cloud providers have enormous political influence, and their arguments—that mandating off-grid power will increase costs, slow AI development, and make the U.S. less competitive against China—will find receptive ears in both chambers. The bill’s bipartisan sponsors, however, are betting that constituent anger about utility bills and grid reliability will prove stronger than industry lobbying in a midterm election environment.
For the data center industry itself, the direction of travel is becoming clear regardless of whether the GRID Act passes in its current form. The era of assuming unlimited cheap grid access is ending. Major hyperscalers are already accelerating investments in dedicated nuclear power purchase agreements, on-site gas generation, and co-located renewable energy projects—less because of regulatory pressure and more because they’ve seen the interconnection queues and concluded that waiting for the grid isn’t a viable option.
The American electrical grid was built over a century to serve homes, factories, and offices. Retrofitting it—or circumventing it entirely—to support the power demands of a new AI industrial revolution is a challenge that will define infrastructure policy for the next decade. The GRID Act is only the opening move.
