The artificial intelligence boom may sound like a story about software, algorithms, and the cloud—something ethereal that we cannot touch. The reality is much more concrete. Behind it all are power plants, transmission lines, substations, cooling systems, and an enormous amount of electricity.
The challenge is that this physical system will have to grow alongside a technology that can scale very quickly. Powering the AI Boom, a new Goldwater Institute report by Domenico Ferraro, PhD, Associate Professor at Arizona State University, argues that the United States is capable of meeting this challenge, given a market-oriented policy framework.
“The policy question is not whether to accommodate the AI boom,” Ferraro writes, “but how to get the next increment of power generation, transmission, and local grid capacity operational on a timeline consistent with demand growth and at a manageable cost.”
The more AI grows, the more pressure it will place on the United States’ physical infrastructure. Our nation is moving into a period when demand for electricity could rise much faster than it has in recent decades. New data centers will be built across the country, and some of those facilities will consume hundreds of megawatts of power. That is not the electricity demand of a typical office park. It is the demand of a major industrial project.
The United States begins from a position of strength, with a diverse mix of energy sources. Natural gas remains the largest source of electricity, nuclear power provides a stable foundation, and wind and solar have expanded rapidly over the past decade. But a strong starting point is not the same thing as being ready for a rapid surge in demand. As the report puts it, “The real issue is timing.”
A large data center can often be planned and built within two or three years. The infrastructure needed to serve it may take much longer. Utility-scale solar projects and simple-cycle natural gas turbines can sometimes move relatively quickly. But larger transmission upgrades, geothermal projects, combined-cycle natural gas plants, hydropower facilities, and nuclear plants can take years to permit, finance, and complete.
That mismatch is where the real pressure will first emerge.
The United States may have enough electricity in the aggregate while still struggling to deliver enough power in the places where data centers are actually built. If several major projects arrive in the same region at once, the local utility may need a new substation, additional transmission capacity, and more dependable generation. Ferraro explains that “the technologies that need electricity can often arrive faster than the infrastructure needed to serve them.” A national debate about artificial intelligence can quickly become a local debate about whether the grid can keep up and who should pay for the upgrades.
America has adapted to major increases in electricity demand before. Industrial growth, suburban expansion, the spread of air conditioning, and earlier technological revolutions all forced the power system to expand. None of those transitions happened without difficulty. But the country did not respond by trying to prevent households from buying appliances or businesses from opening factories. It invested in the infrastructure needed to support economic growth.
The same approach is needed today.
In the near term, utilities and regulators should focus on the projects that can move fastest: targeted substation improvements, targeted transmission upgrades, faster interconnection reviews, solar paired with storage, and additional natural gas generation where appropriate. Over the medium term, the country will need larger investments in transmission, geothermal energy, and dependable sources of power. Over the longer term, nuclear energy—including small modular reactors—could become an increasingly important part of the answer.
But building more infrastructure will require clearer rules.
Energy projects require large upfront investments and years of planning. Utilities, developers, and investors need to know how long approvals will take, who will pay for necessary upgrades, and whether regulators will allow reasonable costs to be recovered. Uncertainty makes projects more expensive and delays investment precisely when speed matters most.
Large data center developers also have a role to play. If a new campus requires expensive local upgrades, the company should make meaningful commitments before ordinary ratepayers are asked to shoulder the risk. Long-term contracts, phased construction plans, and cost-sharing agreements can protect existing customers while giving utilities the confidence to build ahead of demand.
Innovation can help ease the strain as well. More efficient chips, better cooling systems, improved server designs, and smarter energy management can reduce the amount of electricity needed for each unit of computing power. Some data center operations may also be shifted to hours when the grid is under less pressure.
The AI boom is not a reason for alarm. It is a reason to build. The United States already serves nearly four trillion kilowatt-hours in annual electricity sales from a diverse power system, and that base is built to grow. What it cannot do automatically is expand fast enough, in the right places, at the right time. America has the resources. The question is whether policymakers will remove the barriers fast enough to use them.
Read Powering the AI Boom here and follow Goldwater’s research on data centers here.
