The infrastructure powering artificial intelligence is facing a critical bottleneck: building data centers takes months, but building the power plants and transmission lines to run them takes years. This mismatch is creating a crisis that will reshape how electricity is distributed across America, potentially driving up energy costs by 30 to 40 percent by 2030 and forcing difficult choices about who gets power during peak demand.

On July 28, the power flickered in homes across 13 states from Washington DC to Chicago when a data center disconnected from PJM, a massive regional power grid serving 67 million customers. The incident exposed a larger problem: it's physically and economically impossible to build power infrastructure fast enough to meet the demands of AI data centers, and when problems occur, everyone pays the price.

Why Is Building Power Infrastructure So Slow?

A skilled data center builder can complete a warehouse full of graphics processing units (GPUs) in under a year. But the power plants and transmission lines needed to run that facility could take a decade or longer to construct. Transmission lines, the long conductive cables that move energy across the grid, face particularly severe delays.

Suzanne Glatz, an energy consultant and co-author of a Johns Hopkins report on data centers' impact on the power grid, explained the complexity involved. "Planning a new greenfield transmission line is likely to take 7 to 8 years from the date it is approved, but it can take 10 years or longer," she stated. Every transmission line faces unique challenges including geography, voltage requirements, community impacts, environmental reviews, and interconnection points with existing infrastructure.

The bottlenecks extend beyond transmission lines. Large power transformers, devices that distribute energy from power plants to homes, can take 1 to 2 years to order and deliver. The power industry was already struggling to build enough transformers before the data center boom. Tariffs have made them more expensive, and supply chain disruptions have created major logistical challenges.

What Are Companies Doing to Bypass the Grid?

Facing these delays, tech companies and investors are pursuing unconventional solutions. Some are attempting to power data centers with on-site generation rather than relying on the public grid. Kevin O'Leary, the Shark Tank investor, wants to build a data center in Box Canyon, Utah that would consume nine gigawatts of power, more than double Utah's current electricity consumption. The plan relies on natural gas turbines pulled from the nearby Ruby Pipeline, which could raise the state's carbon emissions by 64 percent.

Elon Musk's xAI operates an enormous Colossus 2 data center in Tennessee powered by 59 natural gas turbines. While this approach avoids taxing the local grid, it comes at a significant environmental cost. Some turbines operate without public permits and are destroying air quality for nearby residents. On July 27, the Trump administration's Environmental Protection Agency announced that power plants exclusively serving data centers won't be subject to the Clean Air Act, a decision that may face legal challenges.

Other companies are exploring more exotic solutions. FTAI, a company that refurbishes jet engines for commercial aircraft, is now marketing repurposed jet engines as a power solution for data centers. "The accelerating demand from AI hyperscalers has created an urgent need for immediate power solutions," the company's chief operating officer David Moreno stated.

How Are Tech Giants Planning to Solve This?

Meta, Amazon, and Google are all betting on nuclear power to solve their energy crisis. Microsoft wants to bring Three Mile Island back online, though the timeline keeps getting pushed back. Many tech companies are hoping for breakthroughs in small modular reactors, a technology that startups have promised is "five years away" for the past decade.

The scale of investment required is staggering. Nvidia is reportedly providing a $250 billion financing guarantee to help OpenAI secure a 10-gigawatt data center campus in southern Ohio, developed by SoftBank. The full project cost is projected to exceed $500 billion, with an estimated $350 billion earmarked specifically for acquiring Nvidia's advanced AI chips. The first phase, an 800-megawatt segment, is scheduled for completion by 2028.

This project is also tied to geopolitical considerations. A US-Japan trade deal includes a substantial $33 billion investment from Japan in a natural gas plant designed to supply the necessary energy for the Ohio campus, highlighting how the race for AI dominance is also becoming a race for energy resources.

What Happens When the Grid Runs Short?

Neil Chatterjee, who chaired the Federal Energy Regulatory Commission (FERC) under President Donald Trump, warned that the consequences of power shortages could be dire for ordinary Americans. "If the grid operator is confronted with the choice of sending power to a 24/7 data center to support AI or to a residential household for air conditioning on the hottest day of summer, it's going to that data center," he explained.

Analysts at Bloomberg project that data centers will consume close to one-fifth of all US electricity by 2035, roughly a fourfold jump from today's share. This surge in demand is already visible in regions with high data center concentrations. Officials in Henrico County, Virginia, which hosts 37 data centers, recently told government employees that electricity bills for government and school facilities will increase by 25 percent.

During extreme weather events, the problem becomes acute. A punishing heatwave across the US this summer forced grid operators to order data centers onto backup generators to free up power for homes. Chatterjee pointed to Winter Storm Uri in Texas as another example of how unprepared America's grid is for shocks, when solar, wind, gas, and even a nuclear plant all failed simultaneously in the cold.

Steps to Understand the Data Center Energy Crisis

• Understand the timeline mismatch: Data centers can be built in under a year, but transmission lines take 7 to 10 years to plan and construct, creating a fundamental infrastructure gap that cannot be quickly closed.
• Recognize the cost implications: Electricity bills could rise 30 to 40 percent by 2030 as data centers compete with homes for grid power, with the highest costs concentrated in regions hosting multiple facilities.
• Know the priority system: During grid stress, power will be directed to 24/7 data centers before residential customers, meaning your air conditioning could be cut during peak demand while computing facilities remain online.
• Track alternative power solutions: Companies are pursuing on-site generation using natural gas turbines, repurposed jet engines, and nuclear power, each with environmental, safety, or timeline tradeoffs that remain unresolved.

Chatterjee acknowledged that the tech industry has mishandled its public messaging, focusing on efficiency and job creation rather than explaining why the infrastructure matters. "The nuance doesn't fit on a bumper sticker," he noted, pointing to AI's potential role in the race against China and its promise to accelerate breakthroughs in cancer research.

Chatterjee

Public opposition to data centers is growing. A Reuters/Ipsos poll in June found that just one in three Americans back the rapid pace of construction, while more than half said they would oppose a data center being built in their own community. Data center opponents staged more than 140 protests across 42 states last month in what organizers called the first coordinated national effort to channel anger at the AI infrastructure boom.

Suzanne Glatz warned that the situation could deteriorate significantly. "In the coming years, PJM is showing a shortfall that could lead to outages for customers," she stated. One proposed solution is to subject data centers that don't arrange for new generation to curtailment first before other customers, meaning they would rely on less efficient emergency backup generation. "Either way, based on the load growth projections and the supply commitments, there is a thinning margin of reserve supply which will increase the risk of not enough generation in coming years, the risk being highest on the more extreme days when demand is highest," she added.

Suzanne Glatz

As extreme weather becomes more frequent and dangerous due to climate change, the stakes of this infrastructure crisis continue to rise. The decisions made today about how to power AI will shape electricity costs, grid reliability, and energy policy for decades to come.