Artificial intelligence has created a problem that state governments have never faced at this scale.

For decades, economic-development officials competed to attract factories, distribution centers, and corporate headquarters. Data centers were often viewed through the same lens: jobs, investment, and tax revenue.

But AI facilities are different.

A modern AI campus may require more electricity than an entire metropolitan area. It may demand dedicated power generation, new transmission lines, substantial water resources, and billions of dollars in infrastructure upgrades. Unlike traditional manufacturing, the economic footprint often looks surprisingly small relative to the physical resources consumed.

As states race to attract investment, the real policy question is no longer whether AI infrastructure should be built.

The question is whether governments are forcing developers to prove that communities, ecosystems, utility customers, and taxpayers will not be left holding a bill they regret.

The best states have begun building accountability frameworks. Others are still treating AI infrastructure like a routine real-estate project.

What Makes a Good Data Center Policy?

A sound framework should answer six questions before approval:

1. Who pays for grid upgrades?
2. Who pays for new generation?
3. How much water will be consumed?
4. What are the air-quality impacts?
5. What are the impacts on wildlife and ecosystems?
6. Can the public verify the claims being made?

The strongest states require answers before approval. Here’s the current progress report:

Oregon: A

Oregon currently offers the strongest overall model among major states.

The state’s POWER Act and subsequent utility actions created a framework specifically designed for extremely large electricity users. Rather than forcing households and small businesses to subsidize grid upgrades, Oregon requires large-load customers to bear more of the infrastructure risk themselves.

This addresses one of the most important questions in the AI era: who pays when a data center requires hundreds of millions of dollars in new electrical infrastructure?

Oregon’s approach recognizes that economic development should not become a hidden transfer from ratepayers to technology companies.

The state’s framework is not perfect. Water-use transparency could be stronger, and cumulative environmental impacts remain a challenge.

Still, Oregon currently represents the most mature policy architecture in the country.

Utah: B+

Utah’s recent executive-order framework deserves more praise than many critics have acknowledged.

Governor Spencer Cox identified nearly every major issue that should be examined before approving large AI facilities: water, Great Salt Lake impacts, air quality, wildlife, utility rates, transparency, and rural economic development.

Many states still lack any comparable statewide framework.

The weakness is enforcement.

Executive orders coordinate agencies. They do not automatically create hard legal standards, mandatory disclosure requirements, or automatic denial triggers.

Utah has correctly identified the questions. It has not yet fully institutionalized the answers.

The proposed Stratos project in Box Elder County illustrates the challenge. Reports of a 40,000-acre campus potentially supported by 7.5 to 9 gigawatts of generation have sparked intense debate over water use, air emissions, and infrastructure readiness. Utah’s framework acknowledges those concerns but has not yet demonstrated how they will be resolved.

Virginia: B

Virginia hosts the largest concentration of data centers in the United States.

That reality has forced policymakers to confront issues many states have barely begun discussing.

Recent regulatory changes require large customers to commit to substantial minimum payments for transmission, distribution, and generation capacity. This helps prevent stranded costs if projected growth fails to materialize.

Virginia deserves credit for recognizing that scale changes everything.

A data center using hundreds of megawatts cannot be treated like a normal commercial customer.

However, Virginia’s success in attracting data centers has also created visible stresses on power infrastructure, land use, and local communities. In some respects, the state is solving problems created by years of extraordinarily rapid growth.

Virginia’s framework is increasingly sophisticated, but some reforms arrived after the buildout rather than before it.

California: B

California appears headed toward a stronger framework but remains largely in transition.

Several proposals would require dedicated tariffs and infrastructure cost allocation for major data centers. Regulators increasingly recognize that AI growth could reshape electricity planning throughout the state.

California also possesses some of the nation’s strongest environmental review systems. However, the state still lacks a comprehensive, unified AI infrastructure framework.

Until California develops a more coherent statewide strategy, its approach remains promising but incomplete.

Florida: C

Florida has begun moving toward water-use permitting requirements and cost-of-service principles for large facilities. Those are encouraging developments.

The problem is that Florida’s framework remains relatively narrow compared to the broader environmental and infrastructure questions emerging nationally.

Water matters. So do air quality, ecosystem impacts, grid reliability, transmission upgrades, transparency, and public accountability.

Florida has started the conversation but has not yet built a comprehensive framework.

Ohio: C

Ohio deserves credit for recognizing that tax incentives can become extraordinarily expensive.

State leaders recently paused portions of a data-center tax incentive program after projected costs grew dramatically.

That demonstrates fiscal awareness. Tax policy alone, though, is not infrastructure policy.

Ohio’s actions address taxpayer exposure but do little to answer questions about water, electricity, emissions, or long-term community impacts.

The state has identified one risk while largely leaving others unresolved.

Michigan (Moratorium Model): D+

Michigan’s local moratorium approach reflects understandable public concern.

Communities often feel overwhelmed by proposals whose infrastructure implications remain poorly understood.

A temporary pause can create space for better planning. The problem is that moratoriums are not governance frameworks. They delay decisions rather than improving them.

Without a clear regulatory architecture waiting on the other side, moratoriums risk becoming policy placeholders rather than policy solutions.