Under the Radar: The AI Trade Is Becoming an Energy Trade

There is an old investing rule that has served me pretty well over the years.

When everybody is rushing toward the same gold mine, spend some time looking at the people selling the picks, shovels, and dynamite.

Artificial intelligence is the biggest gold rush in the market right now. Everybody wants to talk about NVIDIA, GPUs, hyperscalers, frontier models, and which software company is going to make half the white-collar workforce obsolete.

That is all interesting, and some of it will probably prove enormously profitable.

But I think there is another part of the story that is getting more interesting by the month.

AI needs a staggering amount of electricity.

Neel Somani lays the case out very well in his recent work, Power 2026: Electricity Pricing in the Age of AI. The basic idea is simple enough. We can build all the data centers we want. We can order all the GPUs NVIDIA can manufacture. None of it matters very much if we cannot find enough reliable electricity to run the machines.

That is where the opportunity starts.

Wall Street has already discovered the obvious beneficiaries. The nuclear names have had their moment in the sun. The large electrical-equipment companies are hardly secrets. Utilities in major data-center markets are getting plenty of attention.

I am more interested in the companies sitting a few rows behind the parade.

That brings us to Transocean (RIG), Eos Energy Enterprises (EOSE), and SOLV Energy (MWH).

They are three very different businesses, and they are certainly not three versions of the same trade. What they do have in common is exposure to a world that suddenly needs a lot more energy infrastructure than almost anyone expected five years ago.

We need more electricity.

We need more natural gas.

We need more storage.

We need more transmission.

We need more construction.

We may also need more oil than the consensus has been willing to admit.

The important part is that we need all of this sooner rather than later.

Somani points out that data centers already consume a meaningful share of U.S. electricity, and demand is still climbing rapidly. You do not have to accept the most aggressive forecasts to see where the problem is headed.

Building another data center is not especially difficult compared with building the energy system required to serve it.

Power plants take years to permit and construct. Transmission lines can take even longer. Turbines, transformers, and other electrical equipment have developed long lead times. Natural gas pipelines have their own regulatory headaches.

Then there is the grid itself.

Electricity is a very local business. You cannot build a giant generating facility in the middle of nowhere and magically send all that power to Northern Virginia, Texas, or wherever the next data-center cluster happens to emerge.

Transmission matters. Interconnection matters. Fuel supply matters. Time of day matters.

Somani also makes an important point about how power prices are set.

The last megawatt needed to satisfy demand often determines the market price.

That can create some very interesting economics.

Suppose cheap nuclear, solar, and efficient natural gas plants are handling most of the demand in a region. Then several large data centers come online and the system suddenly has to fire up a more expensive gas plant.

That expensive plant can become the marginal generator that sets the price for everybody.

The lower-cost producer did not suddenly become more expensive to operate. It may simply get paid more.

That is one reason a relatively modest increase in electricity demand can have a much larger impact on power prices and generator profits.

It is also why I think investors need to stop looking at AI as nothing more than a semiconductor and software story.

The AI trade is becoming an energy trade.

Natural gas is part of it.

Nuclear is part of it.

Solar is part of it.

Battery storage is part of it.

Transmission is part of it.

Oil is still part of it.

The great energy transition was never going to happen in a neat, straight line, and AI is making that increasingly obvious. The world is discovering that if we want more electrification, more manufacturing, and more computing power, we are going to need an awful lot of molecules and electrons along the way.

Transocean (RIG)

RIG is probably the least obvious AI stock you are going to hear about this week.

I like that.

Transocean does not build data centers. It does not make transformers. It does not manufacture batteries.

It owns and operates some of the most sophisticated offshore drilling rigs in the world.

At first glance, that sounds like it has nothing to do with AI.

I do not think that is the right way to look at it.

The broader energy-demand story matters. If global electricity use continues climbing because of data centers, reshoring, electrification, and ordinary economic growth, the world is going to need more natural gas and continued oil production.

That means upstream investment has to continue.

Offshore production matters in that environment because large offshore fields can produce enormous volumes for years once they are developed.

The interesting part is that offshore drilling capacity is not something the industry can recreate overnight.

The last decade was brutal for offshore drillers. Companies went bankrupt. Rigs were scrapped. New construction slowed dramatically.

Capital discipline was not optional. The industry learned the hard way what happens when too many rigs chase too few projects.

That painful history is one reason the current setup is interesting.

The supply of high-end rigs is much tighter than it once was.

At the same time, offshore development activity has improved and dayrates have recovered sharply from the bottom of the cycle.

Transocean has been adding substantial backlog, and that is exactly what I want to see in a cyclical business.

The balance sheet is still the part of the story that will make conservative investors uncomfortable.

That does not bother me.

I do not want anybody confusing RIG with a sleepy utility stock.

This is a leveraged cyclical company. If offshore activity weakens badly, the debt matters. If dayrates remain strong and utilization improves, that same leverage can make the upside much more interesting.

That is the attraction.

I would not buy RIG because another data center opens next Tuesday.

I would look at RIG because the explosion in global energy demand makes it increasingly difficult to argue that the world can simply stop investing in major oil and gas resources.

If energy demand continues surprising to the upside, offshore drilling is not going away.

Transocean owns scarce equipment that producers need to develop those resources.

That makes RIG a strange but very real second-order beneficiary of the power boom.

Eos Energy Enterprises (EOSE)

Eos Energy Enterprises is a much more direct play.

One of the biggest misconceptions in the electricity discussion is that we can solve everything by adding enormous amounts of solar and wind.

We cannot.

Those technologies are important, but electricity has to be available when people actually want to use it.

Solar can produce huge amounts of cheap electricity in the middle of the day.

That does not help much at midnight.

Wind is even less interested in your preferred schedule.

Data centers, however, want power all day and all night.

That is where storage comes in.

Eos is developing zinc-based battery systems aimed primarily at long-duration energy storage.

Most investors hear the word battery and immediately think lithium-ion.

Lithium-ion is a fantastic technology and will remain important.

Long-duration grid storage is a different problem.

Eos is trying to solve that problem with zinc.

The appeal is pretty easy to understand. Zinc is abundant, the chemistry is designed for stationary storage, and the company is targeting applications where power may need to be stored for many hours rather than simply shifted from one part of the afternoon to another.

This is not a clean, polished growth story yet.

Far from it.

Eos has spent years trying to move from an interesting technology into a real manufacturing business.

That process has been expensive and occasionally ugly.

Revenue is growing rapidly. Backlog is substantial. The commercial pipeline is enormous.

Profitability is another matter.

The company is still losing money, and gross margins remain ugly.

That is why the stock belongs in an Under the Radar discussion rather than a conservative income portfolio.

EOSE is a commercialization bet.

The bullish case is that production volume keeps climbing, manufacturing efficiency improves, and today’s backlog turns into a profitable business.

The bearish case is that the company keeps burning cash while trying to prove the economics.

That is the fight.

What makes the opportunity especially interesting is that the market for long-duration storage may be arriving at exactly the right time.

As electricity demand rises and more intermittent generation gets added to the grid, utilities and developers need ways to move power from periods of oversupply into periods when the grid is tight.

That need becomes even greater when customers are data centers that cannot tolerate extended interruptions.

Eos does not need to replace lithium-ion batteries.

It does not need to dominate the entire storage market.

It only needs to win a meaningful share of a very large and rapidly growing niche.

If the company proves that it can manufacture at scale and improve margins, the upside could be substantial.

That makes EOSE one of the more interesting speculative names tied directly to the power shortage.

SOLV Energy (MWH)

Then we have SOLV Energy.

MWH may actually be the cleanest way of the three to think about the infrastructure boom.

Why?

Because somebody has to build all this stuff.

Investors love arguing about which technology will win.

Solar versus nuclear.

Natural gas versus renewables.

Lithium-ion versus long-duration storage.

Meanwhile, the people pouring concrete, installing electrical equipment, building substations, and maintaining power projects are getting paid regardless of which argument wins on social media.

SOLV Energy is part of that group.

The company develops, builds, operates, and maintains large-scale solar and battery projects, and it has been expanding into broader electrical infrastructure.

That is the picks-and-shovels side of the power boom.

The recent numbers are already telling us something.

Revenue has been growing quickly.

Backlog has been expanding.

The company has a large base of projects under operations and maintenance contracts.

That last point is important because recurring service work helps make the business a little less dependent on winning the next construction job.

SOLV has also been moving beyond traditional solar work into utility substations and related electrical services.

I think that is exactly the right direction.

The electricity shortage is not simply about generating more power.

We need substations.

We need grid upgrades.

We need battery installations.

We need interconnections.

We need electrical contractors.

We need crews capable of putting all the pieces together.

SOLV is positioning itself right in the middle of that spending.

There are no magic margins here.

Construction is still construction.

Projects can get delayed. Costs can move against you. Customers can change timelines. Margins can fluctuate.

I am fine with that.

What matters to me is that the demand backdrop looks unusually strong.

If we really are entering a decade of heavy spending on electricity infrastructure, companies capable of actually building and maintaining these assets are going to have a lot of work.

The AI Boom Has a Physical Bottleneck

This is the part of the AI story I think investors are still underestimating.

Forget for a moment which model is better than which other model.

Forget benchmark scores.

Forget the latest chatbot feature that everyone will stop talking about in three weeks.

Ask one very simple question.

Where is all the electricity going to come from?

Once you start asking that question, you end up in some interesting places.

You end up looking at natural gas.

You end up looking at offshore drilling.

You end up looking at storage.

You end up looking at transmission and substations.

You end up looking at electrical construction.

Eventually, you end up looking at stocks that nobody would have dreamed of calling AI plays a few years ago.

That is where things get interesting.

Transocean gives us leveraged exposure to the offshore energy cycle at a time when rising global energy demand is making continued oil and gas investment necessary.

Eos Energy gives us speculative exposure to long-duration storage, which could become increasingly valuable as demand rises and the grid becomes more dependent on intermittent generation.

SOLV Energy gives us exposure to the people actually building the physical infrastructure needed to make all of this work.

None of these stocks looks anything like NVIDIA.

That is exactly the point.

The obvious AI names have already been discovered.

The next opportunities may be farther down the food chain.

Every data center needs electricity.

Every new gigawatt of electricity needs fuel, generation, storage, transmission, and construction.

We are not just building an artificial intelligence industry.

We are being forced to rebuild a meaningful part of the energy system to support it.

That process may last a lot longer than the current enthusiasm surrounding any single AI model.

When a multitrillion-dollar technology boom runs headfirst into a physical shortage, I want to spend some time looking at the companies that can help fix the shortage.

RIG, EOSE, and MWH are three very interesting places to start.