States Can Leverage Tax Incentives to Make Data Centers Better Grid Citizens

There’s a deal to be struck in statehouses across the country to ensure local communities and the American electricity grid benefit from the buildout of data centers, but not enough people are paying attention.

Hyperscalers are in the middle of the largest infrastructure investment cycle in the history of the technology industry. Between 2024 and 2028, energy use from data centers is expected to triple. They need power fast — and are willing to pay for it.

This explosive load growth puts hyperscalers on a collision course with America’s sclerotic electricity grid. Much of the transmission and distribution infrastructure in the ground today was built more than half a century ago. The grid is old, congested, and currently incapable of delivering power to the industries that need it — and nowhere near capable of powering whole sectors of the economy, like transportation and industry, within the next decade. Our planning, permitting, and payment mechanisms for supporting a robust and resilient grid have failed and are in urgent need of reform.

Meanwhile, public support for data center development is basically nonexistent. At least 11 states have proposed legislation to restrict or ban data center development since 2025, and last week Maine became the first state to ban construction outright. Several jurisdictions that granted tax abatements years ago are now moving to revoke or dramatically reduce them, responding to constituent anger over electricity prices and a broad sense that the deal struck with industry has not delivered what was promised. Voters’ principal concerns, according to a March 2026 survey from Searchlight, are data centers driving up consumer utility rates and grid capacity.

To be fair to the states that created these programs, the rationale for data center tax abatements was understandable. Data centers bring real economic benefits: meaningful tax revenue (even at reduced rates), the kind of anchor investment that tends to attract additional development, and construction jobs. In rural and post-industrial communities particularly, a hyperscaler campus could be genuinely transformative. The states that moved earliest and most aggressively, like Virginia, built dominant market positions that proved self-reinforcing, as proximity to existing infrastructure and a trained workforce made subsequent investment easier to attract. And at the time these programs were designed, the grid implications were manageable: A data center in 2010 drew a fraction of the power that an AI-optimized facility draws today.

Besides local impacts, there is also a broader argument that building AI infrastructure in the United States — rather than ceding that development to other countries — has genuine economic and national security benefits. The compute capacity being built in American data centers is increasingly foundational to defense applications, intelligence capabilities, and the kind of frontier AI research that will shape global economic competitiveness for decades. A policy framework that makes the United States a hostile environment for data center investment could pose consequences that extend well beyond electricity bills.

The problem is not that states created these programs — it’s that circumstances have changed, and policy has not kept up.

The goal should not be to ban data center development in the U.S., but to steer that development in societally desirable directions. What is missing is a policy framework that channels public pressure into durable structural reform, rather than ad hoc retaliation that drives investment to the next jurisdiction without the conditions that ensure the industry is as aligned with the public good — and the public grid — as possible.

The data center buildout is among the largest private infrastructure investments in American history, and the public has not been told what it gets in return. State legislation conditioning tax abatements on real grid commitments is the fastest way to define that deal in public, on the record, with real enforcement. The current political moment is ripe for action.

Voters are angry, states are moving to ban data centers outright, and the industry knows the status quo is not sustainable. Because most facilities cannot meet the full standard today, the framework is tiered, with immediate requirements that are achievable everywhere and a graduation pathway that rewards facilities for advancing toward full grid citizenship as infrastructure and policy catch up. While conditionality does not by itself build the grid, it does generate an evidentiary record and political proof of concept that federal and RTO reform actually requires, and makes hyperscalers active stakeholders in fixing it.

What success looks like

The data center buildout has revealed the frailties of America’s existing grid, exposing issues policymakers have ignored for decades. Our grid is old, underfunded, and structurally incapable of powering the economy we have today, let alone the one we are building and need to sustain to maintain global economic competitiveness. If we let this moment of rapid demand growth and shifting political terrain pass us by without fixing these problems, we will face these same problems for the next 30 years, and will have handed infrastructure investment to an industry with no obligations, locked in decades of fossil generation, and left ratepayers to pay for infrastructure that serves the world’s largest technology companies.

We are already seeing signs of this playing out in states across the country. Transmission investment driven by large load growth has the potential to be socialized across the rate base, raising residential bills. In the absence of firm grid connection on a timeline that matches their build schedules and economic incentives, hyperscalers are defaulting to behind-the-meter gas, building dedicated facilities that could instead be grid-connected, flexible assets. That choice removes gigawatts of potential dispatchable load from grid operator visibility at exactly the moment they need it most, while threatening to lock in fossil generation for decades. The public response to this has been to ban data centers rather than steer their development in constructive directions. That outcome is bad for ratepayers, industry, and the climate.

But there is a window of opportunity to change direction. If we act now, the data center buildout has the potential to function as a demand signal that accelerates transmission investment rather than straining a static grid. If hyperscalers show up as forecast, they will have given transmission operators the information they need to plan proactively. Data center load participating in demand response at scale can contribute to more reliable systems, especially during stress events. Long-term offtake agreements from hyperscalers for new geothermal, nuclear, and long-duration storage could drive cost declines and deployment in technologies the U.S. and the world need. And the connection between large load growth and residential rate increases will not just have been severed, it will have been reversed by verified cost causation and a growing customer base.

There is also an important political opportunity that we are poised to capitalize on to secure the policy reforms that are necessary to modernize America’s grid. For the past decade, the coalition for serious grid reform has been limited and relatively weak. It has not been sufficient to drive the federal and RTO reforms that modernizing the grid requires. The data center buildout changes the composition of that coalition in fundamental ways. If we do this right, we will convert hyperscalers from grid customers into active stakeholders in fixing the grid.

This outcome is achievable, but the window for action is narrow. State conditionality is a fast and available lever to establish terms on verified load commitments, cost-causation rules, and grid operator visibility into large flexible loads, among other things. It is not sufficient on its own, and federal and RTO reforms are critical. But we can get to work today, and state tax abatement conditionality could be one important tool.

The abatement “problem”?

According to the National Conference of State Legislatures, at least 38 states offer dedicated tax incentives for data centers, ranging from sales and use tax exemptions to property tax abatements. According to a 2025 analysis, at least 10 states are spending more than $100 million per year on data center tax incentives. Tax treatment is complicated across states, in part because data centers are defined differently in different jurisdictions. In general, the major kinds of tax benefits offered to data centers fall into a few categories:

Sales tax incentives: Examples include everything from equipment essential to the data center, like computer servers and software, to the fences that surround a facility. Sometimes backup generators are exempt as well.
Electricity exemptions: 14 states exempt the electricity used by data centers in their regular operations. Minnesota recently amended its program to remove the exemption for electricity usages and created a new fee for large-scale data centers based on energy use.
Property tax incentives: 11 states offer data centers a statewide property tax incentive.

Data centers do bring real benefits to local communities, U.S. competitiveness, and national security. They can provide communities with property tax revenue (even at reduced rates), anchor investment that can attract additional development, and generate jobs. Data centers in Loudoun County, Virginia, pay enough in property taxes to fund the county’s entire general operations budget.

Source: National Conference of State Legislatures

Conditionality is not a new idea

Many states already attach conditions to data center abatements. Thirty-one states require data centers to make a minimum capital investment to qualify, ranging from $2 million to $250 million. Twenty-three states require job creation commitments, ranging from 5-100 new positions. Nine states incorporate geographic factors, offering lower investment thresholds for facilities built in opportunity zones or distressed communities.

States know how to write conditions into abatement programs. The problem is that today’s conditions don’t meet the moment and won’t align data center development with the grid investment the country needs.

The Idea: Tax Incentives Conditional on Being Good for the Grid

State legislation would condition the award or renewal of data center tax abatements on the recipient demonstrating “grid citizenship”: a defined set of behaviors that ensure large load customers contribute to, rather than extract from, the grid infrastructure they depend on and that the country needs.

Importantly, state conditionality does not directly accelerate RTO transmission planning timelines, resolve regional cost-allocation disputes at FERC, or manufacture the engineering and supply chain capacity that grid buildout requires. Those are federal and regional problems that require federal and regional policy solutions. That said, state conditionality can establish the accountability terms and build the data infrastructure that large reform requires. It can also help correct the financial incentive that causes regulated utilities to prefer socializing large load infrastructure costs across the rate base, rather than allocating them to the facilities that cause them. This is an issue that individual deal-by-deal negotiations cannot fix, because utilities will consistently negotiate against cost causation when their shareholder returns depend on rate base growth. The conditionality idea proposed here is meant to capitalize on a near-term political opportunity and make the federal and RTO work more tractable by creating the evidentiary record and political precedent that larger reform requires.

Specifically, a qualifying data center would need to meet the following requirements:

Take-or-pay load commitments. Certify that projected electricity demand forecasts are accurate within a defined tolerance, and accept financial liability for infrastructure built to serve demand that never materializes. This addresses a fundamental problem for grid planning: overstating demand. When a facility materializes at a smaller scale than proposed or does not materialize at all, the transmission and/or generation infrastructure built to serve it can become stranded and is paid for by ratepayers. According to Lawrence Berkeley National Laboratory, only 13% of capacity that entered the interconnection queues between 2000 and 2019 reached commercial operation by the end of 2024. Take-or-pay is a mechanism to address this problem, where a facility that commits to a defined demand level accepts financial liability for infrastructure to build that commitment.
Full cost causation. Agree that transmission and distribution infrastructure built specifically or primarily to serve the facility will be cost-allocated to the facility, not socialized across the general ratepayer base. This can be very complicated in practice: There’s often no clear line between infrastructure that serves a specific facility and infrastructure that benefits the broader grid. The cost-causation principle does not require drawing a falsely precise line, but it does require that a facility bear the costs it primarily and proximately causes, with shared infrastructure allocated on a pro rata beneficiary basis rather than spread across all ratepayers regardless of benefit. The reason this needs to be legislated is that it preempts deal-by-deal negotiations with data centers, where utilities will have an incentive to secure socialized cost recovery over cost causation.
Verified demand flexibility. Demonstrate the technical capability to curtail or shift a defined percentage of load during grid stress events, and commit to participating in utility or grid operator demand response programs. Flexibility commitments should be verified, with third-party metering, real-time telemetry to the grid operator, and actual dispatch tests before the commitment is certified, and there should be fines for failure to comply.
Clean firm generation support. Either contract for new, additional clean and firm generation capacity, contribute to a clean firm generation fund, or demonstrate behind-the-meter generation that meets the grid visibility requirements below. A facility that generates its own power behind the meter can only claim credit under this condition if it does the following: pays full stranded cost insurance on infrastructure built to serve it, provides real-time telemetry to the grid operator, participates in demand response programs under Condition 3, and commits to converting to full grid connection when interconnection service is available. A facility that uses behind-the-meter generation to avoid grid operator visibility does not qualify under any condition.
Transparency and reporting. File annual verified load forecasts with the state utility regulator, including actual versus projected consumption data, enabling utilities and regulators to do honest, long-term transmission planning. The idea here is that mandatory verified reporting at the state level translates into better regional planning information that effective RTO planning requires and individual negotiations are failing to produce. The verified load forecasts generated by state conditionality reporting are data that federal regulators need to standardize non-firm access products and that RTOs need to implement Order 1920’s proactive planning requirements.

Crucially, states that adopt grid citizenship conditions will generate clear and publicly legible terms that hyperscalers will either accept or refuse. These terms are currently opaque. A state that enacts conditionality legislation creates a public record usable in regulatory proceedings, rate cases, and federal policy development. It could also help address the current race-to-the-bottom dynamics, because unconditional abatement is the implicit baseline.

A handful of states are working on adjacent issues. New Jersey’s AB 5462 requires data centers to guarantee they will use 85% of contracted power over 10 years before infrastructure is built to serve them. Virginia’s HB 897, which did not pass in 2025, proposed conditioning tax exemptions on demand response participation. Texas SB 6 allows utilities to disconnect large loads during emergencies and pays large users to curtail during peak stress. West Virginia’s HB 2014 allows data centers to build their own microgrids but requires them to cover full infrastructure costs and contribute to a grid stabilization fund.

What Is Achievable Today

The five conditions outlined above are the right conditions in principle. However, whether a given state can require all of them immediately depends on the regulatory, market, and physical infrastructure that exists in that jurisdiction. States should enact what they can and build toward the full standard as rapidly as possible.

Most conditions are achievable now, everywhere, with no grace period: verified load forecasting with financial penalties for significant overstatement, take-or-pay liability for infrastructure already committed, annual transparency reporting, and the procurement or contribution floor for clean and firm generation support. These conditions require a data center with a commitment and a state regulatory framework to receive and publish filings — any state with a functioning utility commission can implement them today.

The goal of this framework is to define what good grid citizenship looks like and what states should require, especially if those states are going to offer public subsidies to data centers.

What Must Change to Enable Full Grid Citizenship

It’s worth interrogating the most significant barriers to full grid citizenship today, and identifying the policy solutions required to enable full compliance:

Permitting and the Federal Power Act are constraining the pace of transmission buildout. Transmission lines that cross federal lands, jurisdictional waters, or multiple state boundaries require environmental review processes that routinely take 7-10 years. In addition, the Federal Power Act was not designed for the speed or scale of infrastructure investment the grid now requires, and FERC’s siting authority for interstate transmission is limited. Congress must modernize permitting and the Federal Power Act to address these issues.
Transmission planning is reactive, not proactive. Regional transmission organizations plan in response to interconnection requests rather than in anticipation of load growth. FERC’s Order 1920 requires RTOs to conduct long-term regional transmission planning that accounts for anticipated future conditions, rather than responding to interconnection requests as they arrive. RTOs need to implement Order 1920 fully and without delay, incorporating the verified load forecasts that state conditionality reporting requirements can generate. SPP’s recently approved consolidated transmission planning process is one useful model, and other regions should adopt similar approaches.
The interconnection queue is broken. Transmission cost-causation and interconnection agreements depend on a facility actually getting a firm interconnection agreement on a reasonable timeline. This would be partially addressed by reforming permitting processes, as described above. Interconnection queue reform is critical to enable full grid citizenship.
Clean firm generation does not exist at sufficient scale. Procurement from new, additional clean and firm generation resources is critical for reducing local pollution and addressing climate change. Firm clean resources, like geothermal, advanced nuclear, and long-duration storage, are nascent. A hyperscaler that wants to comply with a clean firm procurement requirement today faces a tough market, where the product may not exist at scale or may take too long to implement. Federal investment in clean firm generation is crucial to enable this condition, and hyperscaler long-term offtake commitments can help pull this investment forward.

Not Yes or No, but HOW

The abatement-conditionality proposal here is about the terms on which public subsidies are extended, not an argument for or against data centers, per se. Sponsors of this kind of legislation should work to steer data center investment toward public benefit. The concern that developers will simply move to jurisdictions with laxer rules is real, but it is mitigated by two factors: Hyperscalers do need to be near existing fiber infrastructure, power transmission corridors, and labor markets, and the framework’s effectiveness scales with interstate coordination. The more states that adopt similar frameworks, the less the race-to-the-bottom dynamic applies. If enough states adopt frameworks that share the core conditions of good grid citizenship, the framework could function as a national standard that helps data center developers and the public distinguish between good and less good grid citizens.

Why Legislation?

The alternative to legislation is the status quo: individual states or utilities negotiating grid citizenship terms directly with hyperscalers on a deal-by-deal basis. Some of this is already happening: Virginia’s large load tariff proceedings, PJM interconnection negotiations, and bespoke power purchase agreements all involve some version of cost and behavior negotiation between utilities and large customers. Still, the existing deal-by-deal approach is inadequate, for several reasons:

Negotiating leverage is asymmetric and temporary. A hyperscaler announcing a new campus has maximum leverage at the moment of site selection, before infrastructure has been built. Once the facility is operational and the utility has committed capital to serve it, that leverage is largely gone. Legislation would take effect upstream of these decisions, so that every future facility enters on the same terms rather than on whatever terms the utility could extract in a one-time negotiation under time pressure.
Individual negotiations produce inconsistent outcomes. Deal-by-deal negotiation results in a patchwork whereby there are some facilities with take-or-pay commitments, some without; some with flexibility obligations, some without; some paying cost-causation allocations; some benefiting from socialized cost recovery. That inconsistency is not just inequitable, it creates perverse incentives. In contrast, legislation sets a floor that applies uniformly, preventing the race to the weakest terms.
Utilities face structural incentives not to push hard. Regulated utilities earn returns on rate-base investment. When a data center drives transmission and distribution buildout, the utility profits, in that the infrastructure cost is passed to ratepayers, but the return on that capital accrues to shareholders. This means utilities may actually prefer socialized cost recovery over cost-causation arrangements, because the former produces more rate-base investment. Legislation removes this conflict by setting cost-causation rules that the utility cannot negotiate away.
Negotiations are not publicly accountable. When a utility and a hyperscaler reach a bespoke interconnection arrangement, the terms are often confidential or embedded in complex tariff proceedings that are opaque to most stakeholders. Ratepayers who will bear the cost consequences of a bad deal have no meaningful visibility or recourse. Legislation creates a public framework with defined terms, verified compliance, and clawback mechanisms, all of which are subject to democratic accountability in a way that private negotiations are not.
The grid planning problem requires population-level data. Utilities and grid operators make transmission investment decisions based on long-run load forecasts. When those forecasts are built on unverified hyperscaler demand projections, which have historically been significantly overstated, the result is stranded infrastructure, rate spikes, and misallocated capital. No individual negotiation can solve this because the planning problem is systemic: It requires accurate, verified, standardized load forecasts across all large customers simultaneously. Only a legislative framework with uniform reporting requirements can generate the data quality that honest grid planning requires.

Conclusion

The abatement-conditionality idea proposed here represents a near-term and tractable solution to the question of how to maximize data center demand growth for public benefit. That said, the state framework proposed here is only as effective as the policy environment it operates in. Most facilities today cannot meet the full grid citizenship standard today because the policies and infrastructure required to meet it do not exist. Wishing for this to change while state bans proliferate is not a solution. While state conditionality cannot fix these things on its own, it can establish, for the first time, a public definition of what the country should get in return for the largest private infrastructure investment in American history, and it can make the undeniable case for fixing the policy barriers standing in the way by generating the public record, data, and political proof of concept that federal and RTO reform requires.

The federal and RTO reforms required to make full grid citizenship real are possible and they cannot wait. FERC must finalize interconnection reform and mandate proactive regional planning, including defending and implementing Order 1920. RTOs need to incorporate verified large load forecasts into forward-looking planning and investment decisions, perhaps following the recent SPP model. Demand response markets need to be designed and deployed at scale. Federal investment in clean firm power must close the gap between what the market will finance today and what the grid needs to deliver power reliably while limiting pollution. Congress must modernize the Federal Power Act and our permitting regimes to enable the grid buildout required to power our country and meet our national security objectives. This can and should happen immediately — it is largely a question of political pressure and prioritization, and now is the moment. State conditionality is a place for that work to start, but it is not where it ends.