As artificial intelligence continues to grow rapidly, its impact extends beyond software, algorithms, and the companies trying to capitalize on it. The real issue is more fundamental: electricity. Data centers are increasing at an incredible rate to handle large-scale AI tasks, and the United States is facing a new energy challenge. This challenge is putting significant pressure on an already aging power grid.
A recent report from the Information Technology and Innovation Foundation (ITIF) suggests that, despite the alarmist rhetoric, the nation can realistically meet this demand in the near and medium term. However, the report also cautions that real relief will require quick, coordinated action, not just hopeful thinking.
“Without strong intervention and planning on multiple fronts, the existing grid will face increasing pressure,” wrote Robin Gaster, research director for ITIF’s Center for Clean Energy. “The country is heading toward intense competition for access to electricity unless changes are made now.”
Gaster warns that regulators, utilities, and policymakers will soon have to balance two significant forces: a sudden surge in energy demand and political pressure to protect residential customers from rising costs. The real question isn’t whether the strain is coming, but how the country will respond.
Data Centers Under Fire: Growth vs. Grid Reality
A major debate in the energy and technology sectors is about the data centers themselves. As AI models require more computing power, the facilities that support them have become massive electricity consumers. Some critics are calling for limits on how quickly data centers can be built or even short-term bans on new grid connections.
Industry leaders believe this mindset is short-sighted and poses economic risks.
Wannie Park, founder of Pado AI, an energy management and AI orchestration company, argues that restricting data center growth is disruptive and undermines the nation’s innovation engine.
“Data centers are the backbone of the AI economy,” Park said. “They can’t just be passive loads on the grid anymore. Instead of stopping their growth, we should make them active participants in supporting and stabilizing the grid. Cutting back only stifles the innovation needed to address the power crunch.”
Park’s views are widely shared across the industry. Experts agree that AI-driven infrastructure will continue to expand, and attempts to slow this growth won’t resolve the underlying energy issues. Instead, the focus should shift to better planning, grid collaboration, and new technologies.
America’s Long-Term Grid Problem Comes Due
According to many utility analysts, the United States has spent decades underfunding its transmission and distribution systems. That delayed investment is now running into the exponential growth in demand.
Scotty Embley, an associate at the data center investment firm Hi-Tequity, points out that trying to slow down data centers will weaken the digital foundations supporting everything from banking and healthcare to transportation and national defense.
“These facilities power critical applications that cannot pause or scale down during peak demand,” Embley explained. “They’re not laundromats that can shift to evening hours. They must run continuously — 24 hours a day.”
However, he acknowledges that early collaboration with utilities is crucial. Data center operators need to plan locations and electrical infrastructure well in advance to avoid overwhelming fragile regional grids.
Why Smarter Grid Integration – Not Growth Restrictions – Is the Real Solution
While some critics push for slowing construction, many industry experts believe the U.S. should focus on integrating data centers more intelligently into the grid.
Allan Schurr, chief commercial officer at Enchanted Rock, says that a more coordinated approach can allow the industry to grow without destabilizing power systems. A key part of this is embracing on-site generation, including natural gas-powered microgrids.
Schurr notes that these systems have several benefits. They can support facilities while waiting for interconnection approval, provide flexible capacity in constrained grid areas, and offer backup power during severe weather or grid failures.
“With proper planning for the entire lifecycle of a data center – from construction to full operation — these facilities can grow while improving the resilience of the grid,” Schurr said.
Data centers usually have backup generators to ensure they can keep operating. Embley mentions that these units can help reduce load on the grid during peak times by supplying power internally.
“This balancing function is important,” he said. “It offers short-term flexibility while longer-term transmission upgrades are completed.”
Avoiding the Grid Entirely: A Radical Alternative
While many data centers rely heavily on utility power, some operators are exploring the option of running entirely off-grid.
Rick Bentley, CEO of HydroHash, a clean energy crypto-mining company, claims that avoiding the grid gives data centers more independence and eases pressure on shared infrastructure.
“When a facility is connected to the grid, utilities can cut power during times of high demand,” Bentley said. “That ensures hospitals and homes stay operational but leaves data center operators in unpredictable situations. Off-grid development saves money and provides stability.”
Though going off-grid isn’t feasible for every operator, it highlights a broader trend: companies are increasingly looking into alternative energy solutions, from microgrids to modular nuclear reactors, to guarantee reliable power.
Extracting More Capacity From the Grid We Already Have
A key point in the ITIF report is that the U.S. can significantly boost available power without waiting years for new transmission projects. The report mentions two immediate opportunities: optimizing current grid infrastructure and promoting flexible demand behaviors.
1. Using Technology to Increase Transmission Capacity
New technologies like dynamic line rating, high-temperature conductors, and better power-flow controls can enhance the capacity of existing lines. These upgrades can be implemented relatively quickly compared to the lengthy permitting needed for new infrastructure.
2. Demand-Shifting for Large Power Users
The report notes that up to 40% of data center workloads are not time-sensitive, meaning they can be scheduled during off-peak hours or distributed geographically. Many grids already perform basic demand-shifting, but not at the scale needed for today’s AI boom.
AI training workloads, batch processing, and content delivery tasks could be allocated flexibly to ease pressure during peak demand.
AI: The Cause of the Power Crunch – and a Potential Solution
One interesting argument in the report is that AI can actually help address the crisis it creates.
Brandon Daniels, CEO of Exiger, believes the U.S. should apply supply-chain-style mapping to electricity distribution. He thinks AI can help identify where power is wasted, where infrastructure is bottlenecked, and where hidden capacity is trapped behind outdated systems.
“We need visibility into the energy delivery system with the same sophistication we apply to national defense,” Daniels said.
Park supports this idea, stating that advanced AI and machine learning can help manage demand so that large consumers, including data centers, adjust their usage in real time based on grid conditions. He calls this demand-side flexibility, a strategy that can greatly reduce peak load.
The challenge lies not in the technology but in the speed. Data center deployments happen in months, while utility planning takes years.
“Regulatory processes are struggling to keep up,” Park said. “Technical requirements and reliability standards also make it hard for data centers to operate flexibly at scale.”
A Grid Designed for Yesterday, Facing the Demands of Tomorrow
Embley points out that the power crisis isn’t just about total energy needs; it’s about the gap between the way the infrastructure was designed historically and today’s computing demands.
Older servers consumed much less energy than modern AI accelerators and GPUs. Currently, individual AI racks can draw between 30 to 60 kilowatts, which is two to three times more than typical CPU-based racks used in the past.
“That shift in computing density has fundamentally changed the power needs of these facilities,” Embley said. “We’re asking a decades-old grid to handle demands it was never designed to support.”
To make matters worse, even when utilities have theoretical capacity, delays in equipment delivery have become a bottleneck. Transformers, switchgear, and high-voltage components often have 12- to 24-month wait times. These supply-chain delays, Embley notes, can stall projects more than the actual construction.
“AI demand is rising sharply each year,” he explained. “Grid upgrades take decades. That disparity – not a lack of willingness — is the main cause of the current power crunch.”
The Bottom Line: The Energy Crunch Is Real – but Solvable
The U.S. is entering a critical time where digital transformation, AI growth, and infrastructure limitations intersect. Though the demand for data centers is undeniable, experts agree that slowing or limiting their growth is not practical and may even backfire.
The way forward requires:
- Smarter grid integration
- Faster adoption of new transmission technologies
- More flexible demand-side strategies
- On-site generation and microgrids
- Greater use of AI to optimize energy distribution
If these solutions are implemented promptly, the country can support both economic growth and energy reliability. However, if policymakers hesitate, the grid will face even more pressure, and the U.S. may struggle to keep up with global competition in AI infrastructure.
For now, experts are clear: a crisis is on the horizon, but the tools to tackle it already exist. What’s needed now is the determination to act quickly.
