Is the US power grid equipped for data center demands?

Is the US Power Grid Prepared to Energize Data Centres?

As the US positions itself for a critical role in the AI race, the establishment of a robust network of data centres is paramount. However, experts warn that **the nation’s ageing power grids and persistent supply chain constraints** could significantly throttle this deployment.

Utilities across the country are inundated with requests for additional power capacity, highlighting an unprecedented shift in demand. Last year, Sempra revealed that its Texas utility, Oncor Electric, received requests totaling approximately **119 gigawatts** from data centres seeking to connect to the grid.

**Chris Seiple**, vice-chair of energy transition and power and renewables at Wood Mackenzie, articulates a concerning reality: “The pace of data centre expansion in the next five years will hinge on our ability to augment energy supply to the grid.”

This issue isn’t limited to the US. A recent report by the International Energy Agency highlights that grid limitations could delay around **20% of planned global data centre capacity** by 2030.

What adds urgency to this problem is the dramatic upturn in energy demand in the US, reversing two decades of stagnation. Though data centres are expected to account for a smaller share of future electricity needs, their impact on grid stability is profound, especially as AI-related facilities typically consume significantly more energy than their traditional counterparts.

The Gigantic Energy Demand Dilemma

To put things in perspective, consider that **a single data centre can have a capacity requirement** comparable to that of a major city, like San Francisco. **Jim Robb**, CEO of the North American Electric Reliability Corporation, points out, “Building sufficient power generation and delivery infrastructure to meet this demand poses formidable challenges.”

The pressure is amplified by the surge in demand stemming from the electrification of transportation and the reshoring of manufacturing. However, AI technologies require a unique set of solutions; significantly increased energy resources and upgraded infrastructure are imperative to support hyperscale data centres.

While many data centres are committed to leveraging renewable energy, this is not enough to satisfy the growing power requirements. Thus, there’s a pressing need for more **natural gas-fired power plants** alongside renewable solutions. Seiple stresses the importance of constructing new power plants that not only meet rising demand but also adapt to evolving capacity accreditation standards while replacing retiring plants.

Infrastructure Challenges Loom

However, the task of building new combined cycle power plants is daunting. Costs have surged, now averaging over **$2,000 per kilowatt**, doubling since before the pandemic. Furthermore, integrating renewables necessitates updating transmission lines to effectively channel energy into the grid.

Robb also notes that **multi-state transmission projects** can take 15 to 17 years to develop. “We need to figure out how to reduce that timeline to three to five years,” he urges.

Adding to the complexity, supply chain constraints are affecting essential components of grid infrastructure, such as transformers and gas turbines. “If you didn’t order your transformer three years ago, sourcing one now is going to be a **daunting challenge**,” warns Seiple.

Trump’s recent tariff policies may exacerbate these difficulties, potentially inflating costs for crucial infrastructure. **Karen Wayland**, CEO of GridWise Alliance, states, “Tariffs will escalate grid project costs and intensify supply chain constraints further.”

These uncertainties could complicate future investment decisions, as utilities grapple with the reality that there’s never been so much uncertainty surrounding customer needs. “Utilities are left questioning whether all ten data centre projects seeking connection will actually materialize,” comments Seiple.

Future Outlook: Demand vs. Reality

Some experts suggest that demand projections for data centres might be overly optimistic. A recent report from Energy Intelligence cites **Nvidia**, which has dramatically increased chip efficiency, making power demands much more manageable. Specific advancements have made some of their processes **45,000 times more energy efficient** over the past eight years.

Michael Collins, Director of Energy Transition Research at Energy Intelligence, notes, “The trajectory of technology innovation typically includes efficiency improvements.”