Incentives

Key Takeaways

• Virginia waives $928M annually in data center sales taxes
• Incentive types: sales tax exemption, property tax abatement, infrastructure investment, rate discounts
• "Race to the bottom" creates prisoner's dilemma among states
• Infrastructure capacity often matters more than incentive generosity

The Competition for Capital

In 2023, Virginia waived $928 million in data center sales taxes. That's nearly a billion dollars in annual forgone revenue—more than the state spends on community colleges, about half of the entire state parks budget.

Why would Virginia deliberately forgo this revenue? Because data centers have become the most sought-after economic development prize in America, and states are willing to sacrifice enormous sums to attract them.

The logic seems straightforward: offer tax breaks to attract billion-dollar projects, which create jobs and boost the tax base. But the reality is more complex—and troubling. When every state offers similar incentives, does anyone actually gain? And who bears the cost?

This is the story of how incentive competition creates a prisoner's dilemma that states can't escape, even when they know the game is rigged against them.

Types of Incentives

1. Sales Tax Exemption

The most common incentive exempts data center equipment and electricity purchases from sales tax. For a $7 billion facility, this can mean $350-500 million in tax savings over the initial buildout, plus ongoing savings on electricity consumption.

Virginia pioneered this approach with its Data Center Exemption in 2010, requiring minimum capital investment of $150 million and creation of 50 jobs. The thresholds have crept upward over time ($250 million in some cases), but the core structure remains: exempt server and networking equipment, cooling systems, backup generators, and electricity from sales tax.

At least 35 states now offer similar exemptions. Some require job creation commitments. Others require matching local tax abatements. But the baseline is now universal: data center equipment isn't subject to sales tax.

2. Property Tax Abatement

Property taxes are typically levied by counties and municipalities, not states, but many states have authorized local governments to offer abatements for data centers.

PILOT agreements (Payment In Lieu Of Taxes) are the standard mechanism. Instead of paying full property tax, the developer makes fixed annual payments—typically 15-30% of what full taxes would be—for 10-20 years.

For a $7 billion facility, full property taxes at 2% of assessed value would be $140 million annually. A PILOT at 20% of full value means $28 million annually—saving the operator $112 million per year.

Critics point out that local governments are giving up enormous revenue. Proponents argue that 20% of something is better than 100% of nothing—if the incentive is necessary to win the project. This is where the prisoner's dilemma begins.

3. Infrastructure Investment

Some states go beyond tax breaks to fund infrastructure directly. This can include:

• Electrical substations and transmission upgrades ($50-200M per facility)
• Road construction and improvements ($10-50M)
• Water and sewer extensions ($5-20M)
• Fiber optic connectivity ($1-10M)

In Ohio, the state offered to fund a new substation for a Google data center—a $120 million public investment for a $600 million private facility. The logic: infrastructure benefits the region long-term, enabling future development beyond this single project.

But this raises the cost socialization question: should electric ratepayers across the state fund infrastructure primarily benefiting a single private company? The utility recovers costs through rate increases, meaning every household and business pays slightly more for electricity to support the data center's connection.

4. Electricity Rate Discounts

Some utilities offer special tariffs for large industrial loads, providing 20-40% discounts compared to standard commercial rates. These "economic development rates" are designed to attract and retain major employers.

A 500 MW facility consuming 4.4 billion kWh annually at $0.08/kWh pays $352 million for electricity. A 30% discount saves $105 million annually. Over 10 years, that's $1+ billion in reduced costs.

Who pays the difference? Other ratepayers. Utilities are regulated to recover costs plus a reasonable return. If one customer class pays below cost, another must pay above cost to maintain the utility's revenue requirement.

5. Workforce Development Programs

States often throw in training programs, hiring assistance, and customized community college curricula to support data center operations. These typically cost $2-5 million—small compared to other incentives, but politically valuable for showing concern about local employment.

The Race to the Bottom

Here's the prisoner's dilemma: If State A offers incentives and State B doesn't, State A wins the project. If both offer incentives, the project goes to whichever offers more (or has better infrastructure). If neither offers incentives, the project goes wherever infrastructure is best.

The dominant strategy is always to offer incentives—even though collectively, states would be better off if none offered them.

Virginia's $928 million in annual tax waivers started as a competitive advantage in 2010. By 2015, North Carolina, Ohio, and Texas had matched it. By 2020, most states competing for data centers had similar programs. Virginia's advantage evaporated, but they can't unilaterally eliminate the incentive without losing future projects.

This is a textbook race to the bottom. Each state, acting rationally in its own interest, creates a collective outcome where billions in public revenue are transferred to private companies without meaningfully affecting location decisions.

And it's accelerating. Michigan offered 11 additional months of sales tax exemption for brownfield versus greenfield sites—trying to differentiate within the incentive structure. Other states are proposing 100% property tax abatement for 20 years. The bids keep escalating.

Cost Socialization

The true cost of data center incentives extends beyond forgone tax revenue. Much of it is socialized—paid by people who gain no direct benefit.

Grid Upgrades and Ratepayer Impact

When a utility builds a $150 million substation to serve a gigawatt data center, who pays for it? The standard regulatory model has the utility finance construction, then recover costs through rates charged to all customers.

In theory, the data center pays its allocated share based on cost-of-service methodology. In practice, cost allocation is complex and often favors large customers:

• Data centers argue they use power efficiently (high load factor, low peak contribution)
• They negotiate special contracts with favorable terms
• Infrastructure upgrades are often treated as "regional system improvements" with costs spread broadly

The result: residential and small business ratepayers subsidize infrastructure primarily benefiting large industrial users. In Northern Virginia, residential rates have increased partly to fund the transmission buildout supporting Data Center Alley.

School Funding Impacts

Property tax abatements directly affect school district revenue in most states, where schools are funded primarily through local property taxes. A PILOT agreement at 25% of full value means the school district receives 75% less revenue than it would otherwise.

Some jurisdictions carve out schools from abatements, requiring full payment of the school portion of property tax. Others negotiate direct payments or require developers to fund school construction. But many simply accept reduced school revenue as the cost of landing the project.

Infrastructure Burden on Rural Communities

A 2-3 year data center construction period brings hundreds of workers, heavy equipment, and continuous truck traffic to rural areas with limited infrastructure. Roads designed for agricultural use get pounded by concrete trucks. Water systems built for 5,000 residents must suddenly serve construction crews plus permanent facility demands.

Developers sometimes agree to road improvements or water system upgrades, but often these are inadequate for the actual impact. The township or county must then repair damage and expand infrastructure—costs that fall on existing residents.

Job Claims vs. Reality

Data center incentive announcements always emphasize job creation: "500 new jobs!" But examine the details, and the picture changes.

A gigawatt facility typically creates:

• Construction jobs: 1,500-3,000 (temporary, 2-3 years)
• Permanent operations jobs: 50-200

For context, that's one permanent job per $35-140 million in investment. Compare to:

• Manufacturing plant: one job per $300K-800K in investment
• Distribution center: one job per $1-2M in investment
• Corporate headquarters: one job per $200K-500K in investment

Data centers are extraordinarily capital-intensive and minimally labor-intensive. That's not a criticism—it's the nature of automated infrastructure. But it means the jobs rationale for massive incentives is weak.

Moreover, many operations jobs require specialized skills that local workers may not have. Facilities often bring in experienced technicians from other regions, meaning the local employment impact is even smaller than the headline numbers suggest.

And because job creation commitments are typically the minimum required to qualify for incentives, they're often set deliberately low. Virginia requires 50 jobs for a $250 million investment—a threshold so low it's virtually guaranteed to be met.

Do Incentives Actually Matter?

Here's the uncomfortable question: do incentives actually affect location decisions, or do they just transfer wealth to projects that would have been built anyway?

The evidence is mixed:

Infrastructure Matters More

Site selection is overwhelmingly driven by grid capacity, transmission access, and interconnection timelines. A site with available power but minimal incentives beats a site with generous incentives but a 7-year interconnection queue.

This is why Texas and Northern Virginia—two regions with very different incentive structures but both with strong grid infrastructure—continue to attract projects despite different policy approaches.

Incentives Enable Marginal Projects

For projects on the margin—where multiple sites have adequate infrastructure—incentives can be decisive. The difference between a 15-year and 20-year property tax abatement might determine which of three finalist sites wins.

But "marginal" is key. Incentives rarely create projects that otherwise wouldn't exist. They shift projects from State B to State A, generating zero-sum competition rather than net new investment.

Timing and Cash Flow Effects

Even if incentives don't change total returns, they affect cash flow timing. Sales tax exemption reduces upfront capital requirements. Property tax abatement improves early-year operating margins. For highly-leveraged projects, these cash flow effects matter.

So incentives may accelerate projects that would eventually happen anyway, or allow slightly more aggressive debt financing. They're not irrelevant—but they're also not the primary driver of investment decisions.

Can States Escape the Trap?

If incentive competition is a prisoner's dilemma harming all participants, can states coordinate to escape it?

In theory, states could agree to limit incentives—either through interstate compacts or federal legislation. This is how the European Union handles state aid, prohibiting subsidies that distort competition across member states.

But in practice, such coordination faces enormous obstacles:

• States have strong incentive to cheat on any agreement
• Defining "equivalent" incentives across different tax structures is nearly impossible
• Federal legislation would face constitutional challenges under state sovereignty
• Economic development agencies are politically judged on projects landed, creating institutional pressure to offer whatever it takes

The most realistic path is transparency and public debate. When citizens understand that $928 million in annual tax waivers means higher taxes elsewhere or reduced services, political pressure can constrain the race to the bottom.

Some jurisdictions are experimenting with "but-for" requirements: incentives can only be offered if the project demonstrably would not occur without them. But proving the counterfactual is difficult, and developers have every incentive to claim the incentive was decisive.

The uncomfortable reality is that states are probably stuck in this equilibrium. As long as data centers are viewed as prestigious economic development wins, states will compete by offering ever- larger incentive packages—even if collectively, they'd be better off spending that money on education, infrastructure, or simply lower taxes for everyone.