Robert Rapier has spent years building Utility Forecaster around one idea: instead of betting on which company wins the AI race, own the essential-service companies that every AI winner has to pay. Today’s article gives you three examples of what that looks like in practice. See the full portfolio he’s built around this thesis →

Two weeks ago, I wrote about the AI boom entering a “show me” phase. Investors are becoming less willing to reward enormous capital spending simply because the letters AI are attached to it. Eventually, all that spending has to produce an economic return.

But there is another side to the story that I think may offer a more durable investment opportunity.

Every AI data center ultimately needs electricity. A lot of it. And no matter how sophisticated the chips inside the building may be, that electricity still has to be generated, transmitted, transformed to the appropriate voltage, distributed, and managed through physical equipment.

That is where America is running into a bottleneck.

U.S. electricity consumption reached another record in 2025, and the Energy Information Administration expects new records in both 2026 and 2027. Data centers are one of the biggest sources of new commercial demand. At the same time, some of the equipment required to expand the grid is becoming increasingly difficult to obtain. High-voltage transformer lead times have stretched into multiple years, with some exceeding 160 weeks. Utilities and developers are ordering equipment years before they expect to need it.

That creates an interesting investment proposition. Instead of trying to determine which AI model ultimately wins, we can look at companies supplying the infrastructure that virtually all of them will require.

Here are three I think deserve a closer look.

GE Vernova: Generating and Moving the Power

GE Vernova (NYSE: GEV) may offer the broadest exposure of the three to the electricity buildout.

GE Vernova was spun out of General Electric in 2024 and operates across power generation and electrification. Among other things, it produces some of the large natural gas turbines that utilities and data-center developers are scrambling to secure, as well as grid equipment used to transmit and manage electricity.

The numbers show just how strong demand has become. In the second quarter, GE Vernova recorded $24.2 billion of orders, up 88% organically from a year earlier. Its backlog reached $176 billion. The company’s gas-power equipment backlog and reservation agreements increased to 116 gigawatts, and management now expects at least 125 GW under contract by the end of the year.

The data-center connection is becoming increasingly direct. GE Vernova reported more than $5 billion of data-center orders during the first half of 2026, more than double what it recorded during all of 2025. To meet demand, the company is expanding annual gas-turbine production capacity toward 24 GW in 2028 and 30 GW by 2030.

This is essentially the other end of the AI trade. Nvidia may sell the chips, but somebody has to produce the electricity that runs them.

There is certainly risk here. GE Vernova shares have already benefited enormously from enthusiasm surrounding electricity demand, and at more than $1,000 a share, investors shouldn’t confuse a strong business outlook with a guarantee of a strong return from any purchase price. The company also has to execute on an enormous backlog while expanding manufacturing capacity.

Still, few companies are positioned as directly at the intersection of rising electricity demand, gas generation, and grid expansion.

Quanta Services: Somebody Has to Build It

The second company approaches the opportunity from a different direction.

Quanta Services (NYSE: PWR) doesn’t primarily manufacture generating equipment. It builds and maintains the infrastructure that connects everything together.

Quanta designs, constructs, installs, and services electric transmission and distribution systems, substations, pipelines, generating infrastructure, and increasingly the infrastructure associated with large data centers and other major power users.

That may turn out to be one of the most important positions in the entire electricity buildout.

You can announce a new power plant. You can manufacture a transformer. You can approve a new transmission line. But ultimately somebody has to put crews in the field and build these projects. Skilled labor and execution capacity are themselves becoming constraints.

Quanta’s latest results illustrate the demand. Second-quarter revenue jumped 41% from a year ago to $9.56 billion, while adjusted earnings reached $4.24 per share. Total backlog climbed to a record $53.4 billion, including nearly $43.8 billion in its electric segment. Management responded by significantly raising its 2026 financial expectations.

That’s not merely an AI story. Quanta benefits from aging-grid replacement, transmission expansion, renewable-energy interconnections, natural gas infrastructure, utility reliability spending, industrial reshoring, and data-center construction. AI is accelerating a trend that was already underway.

That diversification is one reason I find Quanta particularly interesting. Even if today’s most aggressive data-center demand forecasts prove too high, the United States still has an old grid that requires enormous investment.

Again, valuation matters. Quanta shares have risen substantially as investors have discovered this theme. The stock closed Friday at $685.78, although that remains about 13% below its May high. I wouldn’t chase it blindly. But this is exactly the sort of company I would want on a watchlist when the market provides a better entry point.

Hubbell: The Less Glamorous Part of the Grid

My third choice may be the least familiar to many investors, but that is part of its appeal.

Hubbell (NYSE: HUBB) makes the electrical components that allow power systems to work. Its Utility Solutions business produces equipment used to transmit and distribute electricity, while its Electrical Solutions business provides components used to connect, protect, wire, and manage power in buildings and industrial facilities.

These aren’t products that generate a lot of excitement at cocktail parties. Insulators, connectors, wiring systems, and substation components don’t have the appeal of the latest AI chip.

But you can’t build an electric grid without them.

Hubbell’s second-quarter sales rose 15% to $1.71 billion, including 10% organic growth. Electrical Solutions sales jumped 25%, while Utility Solutions increased 10%. Management specifically cited grid modernization, rising electricity loads, and data-center investment as drivers of the quarter. Adjusted earnings increased 12% to $5.52 per share.

The company also raised its full-year adjusted earnings outlook to $20.25 to $20.55 per share and now expects organic sales growth of 9% to 11%.

Hubbell does face risks. Raw-material inflation and tariffs have pressured some margins, and increased capacity investment costs money before it produces returns. But the company’s products sit squarely in one of the areas where America needs to spend heavily simply to connect all the generating capacity and new loads now being proposed.

The Picks-and-Shovels Side of AI

I wouldn’t interpret any of these three names as “buy them tomorrow regardless of price.” All have already attracted investor attention, and the AI infrastructure trade has become crowded in places. I may sound like a broken record, but you can always sell a put. All three currently offer extraordinary return parameters for getting paid while you wait for a cheaper entry point.

In any case, I think the underlying investment thesis has more staying power than simply betting on the next AI application.

America’s electricity problem increasingly isn’t a shortage of potential energy resources. We have natural gas, nuclear plants, solar, wind, and enormous amounts of capital willing to finance new generation.

The problem is turning those resources into electricity and delivering it where it is needed on schedule.

That requires turbines. Transformers. Transmission lines. Substations. Electrical equipment. Skilled workers. And companies capable of putting all those pieces together.

GE Vernova, Quanta Services, and Hubbell occupy three different positions along that chain. I wouldn’t necessarily buy all three today, but all three pass my first test for this theme: They sell something the power buildout cannot proceed without.

And when looking for investment opportunities, supplying something that has become a bottleneck is often a very good place to start.

The three companies I described today represent something I think about constantly in Utility Forecaster: you don’t have to determine which AI model wins to participate in the AI boom. You just have to own the infrastructure that every winner has to pay for — the turbines, the transmission lines, the electrical components that cannot be skipped regardless of what happens in the model race. I’ve built two portfolios around that thesis, and the supply-side setup looks stronger to me today than when I started covering these companies. If you’d like to see what I’m currently holding — and which companies I think are the best buys right now — you can join Utility Forecaster for $49 for the first year →