When AI Meets Industrial Electrification: Are We Headed for a Power Squeeze?

As artificial intelligence drives an unprecedented surge in electricity demand, what happens to the economics of industrial electrification? For years, “electrify everything” has been a rallying cry. But what if the grid itself becomes the constraint? Ginger Rothrock considers the options, and the implications for those invested in industrial infrastructure.

For the better part of a decade, one phrase has anchored industrial decarbonization strategies: electrify everything. Replace gas-fired boilers with electric ones; swap combustion for induction; use heat pumps where steam once ruled.

The logic is straightforward: If the grid gets cleaner, and if electricity remains affordable, then electrification becomes one of the most powerful levers available to reduce industrial emissions. But the world has obviously changed. A second electrification wave is underway, this time driven not by decarbonization policy, but by artificial intelligence, and the two trends may be on a collision course.

The AI Load Is Not Incremental

Data centers are no longer background infrastructure; they are becoming one of the dominant forces shaping electricity demand.

In 2024, U.S. data centers consumed roughly 183 terawatt-hours of electricity, that’s about 4% of national demand. Projections by the IEA and others suggest that figure could more than double by 2030. Globally, demand could approach the equivalent of Japan’s entire electricity consumption within the decade.

AI is the accelerant for this. Training large models and running inference at scale is extraordinarily power-intensive. Estimates suggest that by 2035, AI-optimized data centers in the U.S. could require over 120 gigawatts of power, close to 15% of current national generation capacity.

Individual campuses are now being planned at gigawatt scale and utilities are seeing connection requests that rival those of entire cities. For decades, electricity demand in North America was flat, but thanks to AI, that era appears to be over.

Can the Grid Keep Up?

The good news is that in theory, yes, the grid can keep up. There are nearly 100 gigawatts of new gas-fired power plants planned across the United States, and hundreds more gigawatts of wind and solar are in development. Former coal plant interconnection points are being repurposed, and hyperscalers are building “energy parks” that combine renewables, storage, and on-site generation. I’ve spoken with several founders who are leading the charge on novel geothermal and SMR projects. The energy generation innovation cycle is as strong as ever!

In practice, though, it is complicated. Grid interconnection queues are backed up; transformer shortages are slowing projects (hint – great innovation opportunities). In certain regions, data center growth has contributed to substantial increases in capacity market prices. In parts of Virginia—home to the world’s largest data center cluster—electricity demand from data centers already represents more than a quarter of statewide consumption.

The real question may not be whether supply can eventually catch up, but at what cost. Wholesale prices in key markets are forecast to rise in the mid-2020s, according to the Energy Information Administration. Retail rates are already increasing in several data-center-heavy states and utilities are exploring new rate classes to prevent industrial and residential customers from subsidizing hyperscaler expansion.

All said, if electricity becomes structurally more expensive, what happens to the economics of industrial electrification?

Where Electrification Stands Today

Industrial electrification is real, but uneven. Low- and medium-temperature processes (those under ~200°C) are increasingly viable candidates for electrification. Think electric boilers, industrial heat pumps, mechanical vapor recompression systems, induction heating, and electric arc furnaces, which are either already commercially mature or are rapidly scaling. Many food and beverage facilities, general manufacturers, and certain chemical processes can already electrify substantial portions of their heat demand.

However, high-temperature, continuous processes—cement, primary steel, glass, some petrochemicals—remain more challenging. Electric plasma and next-generation solutions do exist at pilot scale, but the economics and reliability of these remain open questions.

Even today, the economics of electrifying certain industrial processes are finely balanced. For a straightforward one-to-one fuel switch—such as replacing a natural gas boiler with an electric resistance boiler—electricity typically needs to cost around 7¢ per kilowatt-hour or less to compete on pure operating cost. But in many parts of North America, industrial power prices sit only marginally below, or even slightly above, that threshold.

Electrification becomes more compelling when efficiency gains are involved. For example, a high-performance industrial heat pump can deliver three units of heat for every one unit of electricity consumed. In effect, that multiplies the value of each kilowatt-hour and shifts the cost comparison in electricity’s favor.

Incentives, tax credits, and carbon pricing can also tilt the equation. But absent those advantages, many electrification projects are economically viable only within a narrow band of power prices. If electricity rises meaningfully from today’s levels, that band tightens quickly, which leads to the uncomfortable question: If electricity prices rise meaningfully over the next decade, will electrification momentum slow?

Is There an “Electric Ceiling”?

If power prices increase 20–30% in high-demand regions, which is certainly plausible under certain projections, then many borderline electrification projects could stall. Companies are unlikely to replace functioning gas equipment with electric alternatives if the result is higher operating costs and new exposure to grid volatility.

We have seen in other markets what happens when electricity prices spike unexpectedly: Industrial users revert to fuel alternatives, delay upgrades, or self-generate.

None of this invalidates electrification as a decarbonization pathway, but it does suggest that electrification may not be the only (or even always the optimal) path forward.

So What Should We Be Watching?

At HG Ventures, we are watching this closely. It would be a mistake to frame this as “electrification versus AI.” Rather, we are asking:

• Will energy efficiency technologies that reduce total electricity demand become even more valuable?
• Do flexible systems, such as hybrid configurations, demand response and thermal storage, gain importance in a volatile grid environment?
• Will industrial heat storage emerge as a buffer between low-cost renewable generation and high-temperature process needs?
• Could decentralized or modular generation provide resilience for energy-intensive facilities?
• Might clean fuels, including synthetic fuels produced from captured carbon, re-enter the conversation in a different way than expected?

In other words, if electricity becomes more precious, technologies that increase flexibility, efficiency, and resilience could become disproportionately attractive. Rather than a retreat from electrification, it may be a refinement.

A Decade of Grid Expansion Ahead

One thing is clear: The coming decade will test North America’s ability to expand generation and transmission at a pace not seen in generations.

If we succeed—adding renewables intelligently, deploying storage, modernizing grid infrastructure, and maintaining affordability—then industrial electrification can continue expanding.

And if we fall short, well, industry will adapt, as it always does. That adaptation may include hybrid systems, on-site generation, alternative fuels, or entirely new process innovations.

The intersection of AI-driven demand and industrial decarbonization is not a binary choice, it is a systems problem, and systems problems rarely reward single-track solutions.

The grid of 2035 will not look like the grid of 2020. The question is: will our industrial strategies evolve fast enough with it?