Energy

The grid will fall short of AI’s power needs. These 6 resources can close the gap.

This audio is auto-generated. Please let us know if you have feedback.

Simon Irish is CEO of Terrestrial Energy, a developer of small modular nuclear reactors.

Most planning for artificial intelligence’s electricity demand starts with a forecast and asks whether it is right.

I would start somewhere else: Assume the grid will fall short, then ask how to close the gap — and how fast.

Simon Irish

 

The evidence supports that starting point. McKinsey & Company estimates the U.S. will need 30 GW to 55 GW of additional capacity by 2030 as data center demand grows and aging plants retire. PJM Interconnection, the largest U.S. grid operator, has already seen the impacts of this trend. Its latest capacity auction fell about 6.6 GW short of its reliability requirement and cleared at the price cap.

Forecasts can be wrong, and requested load can be speculative or duplicative. But the PJM result is a market outcome, not a forecast. The useful question is which tools can close the gap, by when and at what cost.

6 ways to close the gap, and where each runs out

The first four options below address serving data centers through the grid. The final two consider generation at or near the data center, with an optional grid connection. These categories overlap; on-site generation is a delivery strategy, not a generation technology, and can use multiple technologies.

Flexible demand. Duke University’s Nicholas Institute estimates the grid could absorb 76 GW of new load if that load accepted curtailment averaging 0.25% of annual consumption, in events of about 1.7 hours. It is the fastest option because it needs no new generation. It makes better use of capacity already on the grid.

But that capacity is finite, and beyond it, further load growth requires new supply. There is also a practical hurdle; to date, many data centers have shown little willingness to curtail AI workloads, because the value of computing far exceeds the cost of power.

Existing nuclear plants. Increasing output at operating plants through uprates and restarting closed plants can add reliable electricity supply, while extending operating licenses keeps existing plants running. A U.S. Department of Energy initiative targets 5 GW of added capacity at existing nuclear plants by 2029, and DOE has closed a $1.9 billion loan to restart Iowa’s Duane Arnold plant, returning 615 MW to service.

The limit is opportunity: Few operating plants remain candidates for significant uprates, and restarts are underway at nearly every plant where one is practical.

Keep up with the story. Subscribe to the Utility Dive free daily newsletter

Natural gas. Gas offers firm power and a familiar path, but gas turbine supply has become the constraint. GE Vernova’s gas turbine backlog and slot reservations reached 116 GW at midyear 2026, and the company has said a heavy-duty turbine ordered today will not arrive until 2031. Buyers also take on fuel-price exposure and decades of emissions.

Wind, solar and storage. These resources can be deployed quickly and at scale. Their limit is that data centers require dependable power around the clock, so matching that requirement takes substantial overbuilding, energy storage and transmission infrastructure to deliver the same capacity as a firm resource.

On-site generation. Building a power plant at a data center can bring electricity online sooner by avoiding some grid upgrades and connection delays, where permitted.

However, it does not speed up equipment deliveries. An on-site gas turbine plant faces the same turbine shortages as one supplying the grid. Reliable operation also requires sufficient on-site backup capability or access to grid power during outages and maintenance. If a project relies on the grid, developers need clear terms defining that access and its cost.

Advanced nuclear energy. Advanced reactors offer firm, clean power and can be built as a fleet. They can serve data centers directly, with an optional grid connection. They will not close the gap this decade, and their real test is cost at fleet scale.

Guiding principles

No single option is fast, firm and large enough on its own. Three principles follow.

Match each solution to its clock. 

Flexible demand, uprates at existing nuclear plants, and new wind, solar and storage can help meet near-term electricity needs.

Natural gas and on-site generation can help fill the gap while longer-term resources are developed, but only if equipment deliveries, permits and grid interconnection align with the data center’s schedule.

Advanced nuclear energy can contribute in the 2030s and beyond.

Utilities must plan now for both natural gas and advanced nuclear energy. DOE’s long-term modeling indicates 550 to 770 GW of additional firm, clean generating capacity could be needed by 2050, with advanced nuclear energy providing about 200 GW.

Infrastructure built to meet an immediate shortfall should also have a useful role later. A gas plant could provide firming capacity or operate during periods of peak demand as longer-term supply becomes available.

Make flexibility the price of speed. Utilities can offer earlier interconnection to customers accepting curtailment under defined, metered and enforceable terms. For data center operators, that is a choice between some power sooner and firm power later. Customers that can curtail load may get power sooner; those that cannot must wait for firm supply. This converts flexibility from an assumption into a contract and rewards developers whose flexibility is real.

Secure the long-lead items for firm, clean power now. Buyers who reserved gas turbine production slots early secured earlier deliveries; those ordering now may have to wait until 2031.

Firm, clean power has its own long-lead items, including fuel supply and equipment manufacturing capacity. Utilities and data center operators can sign reservation or offtake agreements that give suppliers the demand signal to build capacity before it is needed.

These principles also address the forecast debate. Some investments — flexibility, uprates and transmission — remain valuable across demand scenarios. Others are options worth buying, such as reservations for firm supply in the 2030s and beyond.

And some are bets to avoid: irreversible commitments sized to a demand number that may not arrive. If forecasts prove too high, the cost is the price paid to preserve an option. If they prove right and no one bought the option, the cost is the kind of shortfall PJM has already seen.

Planning for a shortfall is not pessimism. Of all the inputs AI needs, reliable power may prove the scarcest. The race to scale AI is a race to deliver power. Utilities, regulators and developers need tools that arrive on time, hold their value if forecasts change, and keep electricity reliable and affordable for existing customers.

via Utility Dive https://ift.tt/6mYDd4n

Categories: Energy