U.S. policy debate over artificial intelligence is moving from model safety and chip leadership toward a more immediate question: who will pay to connect power-hungry data centers to an already constrained electricity grid. Proposed federal legislation and state-level approval limits increasingly place electricity costs, interconnection queues, and local permitting at the center of AI infrastructure planning.
That shift could reshape the economics of building large computing campuses. The largest technology companies have committed tens of billions of dollars to servers, buildings, and networking equipment, yet access to electricity is becoming a separate bottleneck. A project can have financing, hardware orders, and customer demand, but still face delays if a utility cannot provide generation or transmission capacity.
Federal proposals target ratepayer exposure
Democratic Senator Cory Booker and other lawmakers introduced the Power for the People Act in January 2026. The proposal would require large data centers to cover costs associated with grid expansion needed to serve them, aiming to prevent households and smaller businesses from absorbing infrastructure expenses through electricity rates.
The issue has also attracted bipartisan support. The Republican-led House passed the Ratepayer Protection Act in September 2026 by a 417-3 vote, with support from many Democrats. The bill later failed to clear a Senate procedural vote, falling 57-43 short of the 60 votes needed to advance.
The bills differ in their legislative paths and details, but both reflect mounting concern that a handful of large computing projects could force utilities to build expensive transmission lines, substations, and generation capacity. Utilities generally recover approved capital spending from customers over long periods, placing the allocation of those costs at the center of regulatory disputes.
The Department of Energy's 2024 data-center study illustrates why the issue has reached Washington. The department estimated that U.S. data centers consumed about 176 terawatt-hours of electricity in 2023, equal to 4.4% of national consumption. It projected that demand could rise to between 325 and 580 terawatt-hours by 2028, representing 6.7% to 12% of U.S. electricity use.
Those estimates include conventional cloud facilities as well as AI-focused computing sites. Training and operating advanced AI models requires dense clusters of specialized chips, which can draw far more power per facility than older enterprise data centers.
States are adding another layer of risk
State governments are becoming equally consequential for project developers. Following the November 3, 2026 midterm elections, reporting said Texas governor-elect Gina Hinojosa, a Democrat, pledged to pause approvals for new data-center projects until consumer and community protections are clarified.
Republican Governor Greg Abbott had already taken steps to halt some new grid interconnections and subject data-center projects to greater scrutiny. Texas has become a focal point because its ERCOT power system has attracted energy-intensive industries, including data centers, cryptocurrency mining, and large industrial facilities, while also facing tight reserve conditions during periods of extreme weather.
By September 2026, at least five states had Republican candidates proposing or tightening limits on data-center development, according to election-cycle reporting and campaign proposals. The politics vary by state, but the practical questions are similar: whether developers must bring dedicated generation, how much they should pay for transmission upgrades, whether projects can be curtailed during grid stress, and how local communities can challenge large facilities.
A moratorium or delayed interconnection review does not necessarily cancel a proposed campus. It can nevertheless alter a project's financial profile. Developers often arrange financing and equipment contracts around expected energization dates. A delayed power connection can postpone revenue, increase borrowing costs, and complicate agreements with tenants seeking computing capacity.
AI spending surged before the grid became the constraint
The new emphasis on power follows an extraordinary expansion in AI-related infrastructure spending. The Nasdaq Composite rose 43.42% in 2023 and another 28.64% in 2024, according to Nasdaq market data, recovering from a 33.10% decline in 2022. The two-year gain of roughly 84.5% coincided with the rapid commercial rollout of generative AI services.
NVIDIA's filings show the scale of demand for the processors used in AI data centers. The company reported data-center revenue of $15.005 billion in fiscal 2023, $47.525 billion in fiscal 2024, and $115.186 billion in fiscal 2025, which ended January 26, 2025. That represents roughly a 7.7-fold increase over two fiscal years.
The companies building AI services have also raised capital spending. Meta reported $39.23 billion in 2024 capital expenditures, including principal payments on finance leases, in its annual filing. Microsoft reported $44.48 billion in fiscal 2024 cash outlays to acquire property and equipment.
Much of that market and infrastructure expansion occurred while Democrats controlled the White House. President Joe Biden signed the Chips and Science Act in August 2022, providing $52.7 billion for semiconductors and related programs, including about $39 billion in manufacturing incentives and roughly $11 billion for research and development.
Biden also issued Executive Order 14110 in October 2023, covering AI safety, risk management, competition, talent, research, and government use. The administration convened cloud providers, AI companies, power firms, and data-center operators in September 2024 to discuss grid buildout, energy, permitting, and workforce needs. Executive Order 14141, signed in January 2025, directed the Defense and Energy departments to identify suitable federal land for advanced AI data centers.
Existing power rights gain strategic value
Grid constraints could place a premium on sites with established interconnection agreements, permits, substations, and contracted power supplies. That dynamic may extend beyond AI developers to certain cryptocurrency-mining facilities, many of which secured access to power infrastructure before the present rush for high-density computing capacity.
The Cambridge Centre for Alternative Finance estimated in a July 2026 update that global digital-currency operations consumed about 190 terawatt-hours of electricity. Some established mining operators hold long-term power agreements or operate near generation assets, though available capacity, contract terms, and local regulations differ sharply across facilities.
Where contracts allow it, owners of existing energy-intensive sites could seek to lease unused electrical capacity or convert portions of their infrastructure for high-performance computing. The approach would face technical and commercial limits: AI workloads require different cooling systems, networking equipment, reliability standards, and customer arrangements than Bitcoin mining.
Distributed computing networks are another area drawing attention from market participants. These networks coordinate computing resources from independently owned machines rather than relying entirely on centralized campuses. CryptoRank estimated in July 2026 that tokens associated with such networks had a combined market value of $3.46 billion. Their ability to serve demanding AI workloads remains uneven, particularly for applications requiring low latency, verified hardware, or large clusters operating continuously.
The emerging policy environment gives project developers a stronger incentive to secure electricity early, negotiate transparent cost-sharing arrangements, and consider sites with proven grid access. For AI companies, the next expansion race is increasingly likely to be decided as much by utility commissions and transmission planners as by chip supply.
As AI reshapes infrastructure, explore how AI copy trading applies intelligent automation to crypto market strategies.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
