The electricity needed to power America’s artificial intelligence boom is now colliding with state-level climate policy, and climate policy is losing. In North Carolina, lawmakers have stripped out a 2030 carbon reduction deadline that applied to Duke Energy, the state’s dominant utility. In Virginia, regulators have created an entirely new rate category for data center customers drawing 25 megawatts or more from Dominion Energy. Both moves, finalized in spring 2025, reveal a widening tension between the long-term promise of decarbonization and the short-term scramble to keep power flowing to an industry that never sleeps.
North Carolina scraps its 2030 carbon target
North Carolina’s General Assembly passed Senate Bill 266, removing the state’s interim requirement that electric utilities cut carbon dioxide emissions by 70 percent from 2005 levels by 2030. The enrolled session law text confirms the change: the statutory language directing regulators to take “all reasonable steps” toward that benchmark has been deleted. The state’s 2050 carbon neutrality goal remains on the books, but without the interim milestone that was supposed to force measurable progress along the way.
The bill did not become law quietly. Governor Josh Stein vetoed the measure, warning that it would weaken clean energy commitments and slow the state’s transition away from fossil fuels, as the Associated Press reported. Republican supermajorities in both chambers then overrode the veto, a step the AP documented in its coverage of the legislative session. Supporters called the override a pragmatic response to new load forecasts, while opponents said gutting the interim target leaves the 2050 goal as little more than an aspiration with no enforceable checkpoints.
Duke Energy, which serves roughly 3.4 million customers in North Carolina, is the utility most directly affected. Without the 2030 mandate, the company faces less statutory pressure to retire coal plants or speed up renewable energy procurement on the original schedule. In regulatory proceedings, Duke can now argue for a slower pace of plant closures without running afoul of state law, provided it maintains a stated trajectory toward midcentury carbon neutrality. For ratepayers, the tradeoff is uncertain. Slower fossil fuel retirements could hold down near-term electricity bills by deferring expensive capital investments in replacement generation. But they would also lock in higher emissions for years, making the eventual push to carbon neutrality steeper and potentially far more costly.
Lawmakers repeatedly cited data centers and AI-related infrastructure as key drivers of rising electricity demand, framing the rollback as necessary to keep the grid reliable during a period of rapid industrial growth. Yet Senate Bill 266 itself does not mention data centers by name, and the legislative record made available so far does not include detailed modeling that ties specific load forecasts to the decision to eliminate the 2030 benchmark. That gap matters. It means the political narrative about reliability and economic development moved faster than the publicly available technical analysis supporting it.
Virginia creates a data center rate class
Virginia took a different approach to the same underlying pressure. During its biennial review of Dominion Energy Virginia, the State Corporation Commission issued an order establishing a new GS-5 tariff for customers demanding 25 megawatts or more. The threshold is designed to capture large data centers and other extremely power-intensive operations, separating them from the general commercial rate pool.
The logic is straightforward: prevent smaller commercial and residential customers from subsidizing the massive grid upgrades that concentrated data center loads require. Northern Virginia’s Loudoun County corridor already hosts one of the world’s densest concentrations of server farms, according to commercial real estate analyses from firms like CBRE. Dominion’s own long-term resource plan describes rapidly growing power demand across its service territory, and state energy data compiled through Virginia’s publicly accessible data portal illustrate the sharp rise in large-load consumption reshaping utility planning in the region.
For everyday electricity customers in Virginia, the rate class split is about who pays for new power plants, transmission lines, and grid reinforcements. If the costs of serving data centers are effectively ring-fenced within GS-5, residential bills should be more insulated from the infrastructure buildout needed to power multi-megawatt campuses. If cost separation proves incomplete in practice, ratepayers could still absorb part of the tab through general rate increases or systemwide transmission charges that spread costs across all customer classes.
The order also reflects a balancing act between economic development and fairness. Data centers bring significant capital investment and local tax revenue, and state and local officials have courted them aggressively for years. But their round-the-clock electricity consumption can strain existing infrastructure, prompting new gas-fired plants, transmission corridors, or accelerated upgrades that would not be necessary without such concentrated demand. By carving out a dedicated rate class, the commission is signaling that the primary beneficiaries of those investments should shoulder a proportionally larger share of the costs.
How much electricity data centers actually use
Neither North Carolina nor Virginia is acting in a vacuum. Across the country, utilities are revising load forecasts sharply upward, and data centers sit at the center of that revision. According to the U.S. Department of Energy, data centers consumed an estimated 4.4 percent of total U.S. electricity in 2023, a figure the department projects could rise to roughly 6 to 12 percent by 2028 depending on the pace of AI deployment and efficiency gains. The Electric Power Research Institute has published broadly consistent estimates, noting that a single large hyperscale data center campus can draw several hundred megawatts of continuous power, comparable to the load of a small city. Grid operators like PJM Interconnection, which manages the wholesale electricity market covering Virginia and parts of the Mid-Atlantic, have reported that data center interconnection requests now dominate their planning queues.
What complicates the narrative is that many of the largest data center operators, including Google, Microsoft, and Amazon, have made their own clean energy commitments. These companies have signed billions of dollars’ worth of power purchase agreements for wind and solar generation and have invested in next-generation nuclear technology. Whether those corporate procurement efforts will offset the emissions from the new fossil fuel capacity that utilities say they need to meet data center demand is an open and contested question. The answer depends on how quickly renewables and storage can be built, how firm those corporate commitments prove under financial pressure, and whether grid planners count corporate PPAs toward overall system decarbonization or treat them as separate accounting exercises.
Where the evidence is firm and where it is not
Several core facts are well established. North Carolina’s General Assembly did finalize legislation eliminating the 2030 interim carbon reduction requirement; the enrolled session law text confirms both the original 70 percent target and the continuing 2050 neutrality goal. The Virginia State Corporation Commission did create the GS-5 rate class for customers with loads at or above 25 megawatts. Dominion Energy Virginia did publish a long-term resource plan that explicitly addresses rising demand from data centers and other large users. And contemporaneous reporting documented the North Carolina legislative fight, including the veto and override, in real time.
The connection between data center growth and the weakening of emissions targets is strongly implied by the timing and stated rationale of lawmakers and regulators. But no utility filing reviewed for this article contains a direct, quantified link between projected data center load and the specific decision to abandon the 2030 goal. The causal chain is plausible and widely cited in legislative debate: data centers and AI clusters require vast amounts of electricity, utilities warn of looming capacity shortfalls, and policymakers respond by loosening near-term climate constraints or restructuring rates. The specific modeling that would show how much of the new demand comes from these facilities, and how that demand affects emissions trajectories under different policy scenarios, has not been fully released to the public.
Similarly, the Virginia commission’s order clearly identifies very large customers as a distinct cost driver, but it does not break down how much of Dominion’s projected capacity expansion is attributable to data centers versus population growth, transportation electrification, or other industrial loads. The GS-5 class is best understood as a financial and regulatory tool for cost allocation, not a comprehensive accounting of where rising demand originates.
Open questions for ratepayers and grid planning through mid-2026
The most consequential unknowns are practical, not procedural. In North Carolina, the question between now and mid-2026 is whether Duke Energy will use the removal of the 2030 target to slow coal retirements and delay clean energy investments, or whether market forces and federal policy will push the utility toward decarbonization regardless of what state law requires. Duke has not publicly released executive statements in the reviewed materials that explicitly tie data center demand to the rollback; the rationale has come primarily from lawmakers citing reliability and economic competitiveness.
In Virginia, the open question is how Dominion will actually meet the new demand. Whether additional load from data centers is served primarily by gas-fired plants, renewables, nuclear units, or some combination will determine the real carbon impact. No official projection from the Virginia SCC or Dominion reviewed for this article quantifies how the GS-5 rate class will affect the utility’s overall emissions trajectory. The structural change in rate design is clear; its environmental outcome is not.
Forecasts for AI-driven electricity demand can shift quickly as technologies evolve, efficiency improves, or market conditions change. But the policy choices being made now in Raleigh and Richmond are not easily reversed. Scrapping an interim emissions target or restructuring utility rates creates new baselines that future regulators and legislators will inherit. Whether these moves ultimately align with long-term climate commitments or entrench a more carbon-intensive path that will be harder and costlier to unwind is a question that will play out over the next decade of power plant construction, transmission buildout, and rate cases across the Southeast and Mid-Atlantic.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
