Data centers aren’t the problem — it’s really about how we build them

New data center construction has become a controversial issue in many communities across the United States. They are electricity and water hogs , pushing utility costs higher for residents and businesses. Some areas have held referendums to keep them away . Despite such headwinds, more data centers are not only needed, they are essential national infrastructure to ensure that the AI revolution will continue.

Data centers are at the foundation of the digital economy , housing the computing power needed to train and operate increasingly sophisticated artificial intelligence systems. Preserving American leadership in AI requires growing this computing infrastructure. To reduce tensions around data centers, we must also rethink how to build them and position them to be better partners with communities where they are sited.

The problems surrounding data centers are well documented and very real. Data centers consume massive amounts of electricity . This power demand strains local grids and can potentially put upward pressure on electricity prices . The same issue applies to water, which many data centers use for cooling their computer equipment. No one wants to pay more for essential resources because of a shortage that they are not responsible for and perhaps do not directly benefit from.

But these concerns do not capture the full picture of the situation.

The U.S. houses more than 4,700 data centers . That is more than the total number in Canada, the United Kingdom, France, the Netherlands, Italy, Germany, Japan and China combined. This concentration of computing infrastructure may seem more than enough to some, but such computing power does not just contribute to America’s AI leadership — it is a precondition for it. Training state-of-the art AI models requires massive computing capacity that takes years to build.

Many of these data centers are critical for driving progress on AI, which is impacting more of us each day and continuing to expand its footprint of influence and application. Limiting their growth because local communities view them as undesirable may be shortsighted in pursuit of global AI leadership.

What’s more, the energy challenges created by data centers are solvable.

As of 2023, data centers consumed around 3 percent of electricity produced in the U.S. This is expected to climb to 9 percent by 2030, a substantial increase that catches people’s attention. But national statistics do not reflect what host communities fear and feel. A single data center can draw as much power as tens of thousands of homes , and it is that local, incremental demand, not the national framing, that shows up on a resident’s utility bill.

It is this disconnect between local and national views that explains why structural fixes are needed.

Satisfying this growing electricity demand will require every opportunity available for power generation. Closing or discouraging wind and solar is the wrong move. Having data center developers co-invest in utility-scale renewables and dedicated microgrids is a reasonable expectation. At the same time, local environmental emissions should not be ignored if natural gas is used to generate electricity. State utility commissions and local land development permit authorities must require and enforce these conditions before construction, rather than relying on goodwill afterward.

The same logic applies to water, which deserves equal attention. Developers should be required to adopt closed-loop or reclaimed-water cooling systems whenever possible, and to provide funding for local water system upgrades. In the interim, impact fees , tied to a facility’s actual energy consumption and water usage, can be assessed, with the funds used to reduce local resident utility costs, lower real estate taxes, or contribute to water system improvements.

Independent of these financial issues, some communities want greater input into how land in their area is used. This concern will not fully disappear even with a well-structured deal; land-use changes are difficult to predict and even harder to reverse.

Residents are justified to be skeptical of commitments that outlast the officials who negotiated them. Community benefit agreements created prior to construction would alleviate local resident concerns. Traffic, noise, local hiring and how a site can grow into the surrounding landscape should all be on the table prior to approval as negotiation points between city councils and developers. Without such preemptive discussions, relationships can easily become adversarial rather than fostering a stakeholder mindset.

Of course, community benefits like tax revenue and jobs created should not be ignored. The Meta data center being built in northeast Louisiana is an example worth noting. It is paying for its power consumption, with no negative impact on residents, and contributing to local tax revenue in meaningful ways.

Forcing data centers to address the energy and water demands they create would give developers and operators an incentive to become better partners with their host communities, while giving communities the added incentive to attract them locally. If done with care, communities will be more apt to welcome data centers. Indeed, the choice ahead is not whether to build data centers. It is whether we choose to build them right.

Sheldon H. Jacobson, Ph.D., is a professor of Computer Science at the University of Illinois Urbana-Champaign. He applies his expertise in data-driven risk-based decision-making to evaluate and inform public policy.