Uncovering The Next Hyperscalar Trade In The Markets

Want to skate where the puck is going? You'll discover the emergent AI bull market that Wall Street is ignoring, allowing you to get in early.

Note: Today is a Deep Dive into a secular investing theme, along with a stock setup that’s positioned for solid upside once the market realizes the narrative and marks up the stock. The theme is available for free readers, and if you want the stock setup, then become a client of ConvexSpaces.

Where’s the Real AI Bottleneck?

NVDA is a great example of a TINA stock.

“There Is No Alternative.”

If you want to compete in AI, you’ve got to buy a boatload of H100’s. That moat has helped NVDA’s stock jam up into a $4.4T valuation.

While NVDA (currently) has a hold on the semiconductor bottleneck, there’s other spaces where demand from AI Hyperscalers can lead to some big wins.

For example— water. Data centers need water to cool off their servers, and the demand for water is already affecting areas that don’t have a lot of it.

Yet there’s one space that hasn’t seen the crunch, but we’re starting to hit key inflection points where power is going to be the biggest bottleneck AI has ever seen.

The Electrons Arms Race

Earlier this year, Elon Musk mentioned that he would love to have a 1 Terawatt (TW) AI cluster built out. It’s a lovely moonshot idea, but that’s past the total power available currently in the U.S.

A strategic failure of the US-China trade relationship is that China was able to scale up its power as they became the manufacturing hub of the world. And if they continue to build out their power generation and redirect it to AI, then they could overpass the United States. It won’t be because they have better models or more engineers, it will be because they can simply afford to run larger clusters than us.

The Trump administration is well aware of these risks, and our energy policy is starting to adapt.

Here’s a quote from an executive order:

Advanced computing infrastructure for artificial intelligence (AI) capabilities and other mission capability resources at military and national security installations and national laboratories demands reliable, high-density power sources that cannot be disrupted by external threats or grid failures.

Advanced nuclear reactors include nuclear energy systems like Generation III+ reactors, small modular reactors, microreactors, and stationary and mobile reactors that have the potential to deliver resilient, secure, and reliable power to critical defense facilities and other mission capability resources.

This EO was signed on May 23rd. Here’s OKLO, a nuclear stock, after that release:

The nuclear play has saturated the financial infosphere, but no matter how much red tape is cut, it’s still going to take time to bring on nuclear capacity.

Meanwhile, the Hyperscalars aren’t waiting… but how are they getting over the power bottlenecks?

Location, Location, Location

A few days ago a PR came out about Meta receiving $29B in financing to build a data center. It won’t be located in Northern Virginia or Silicon Valley. It’s going to be in the middle-of-nowhere, Louisiana.

They aren’t building there for the tax breaks. It’s for cheap energy access.

Back in 2022 I went out to Laffite, Louisiana to look at a crypto mining setup. They leased out an orphan oil well, slapped a turbine on it, and was mining BTC on a satellite link.

The Hyperscalars are using the same playbook. They’re addressing the power bottleneck like how High Frequency Traders would colocate their servers next to the exchanges. If the cost to transport energy is cheaper, then they can chip away at that bottleneck.

That puts natural gas in play, but we’re going to head west to the other opportunity.

Arizona. It’s an absolute boomtown.

Phoenix already has 100 data centers, with the possibility to ratchet up as AI comes into town. Tucson just shot down a data center project dubbed “Project Blue,” a 600MW site that could have built out a total capacity of 1.3GW.

You’ve also got TSMC moving into Arizona with a $100B fab buildout. AMD is going to leverage those plants to ramp up domestic chip production. Even Intel, yes Intel, has announced a $20B project to expand their fabs in Chandler, AZ.

A moment ago I mentioned how water was a bottleneck. Why are all of these tech giants building in the middle of the desert?

Because the true bottleneck is power, and the fastest way to bring capacity online is solar.

Solar is the New “TINA” Trade

A summary of the AI energy landscape:

Nuclear’s going to take time.
So will coal.
Yes, coal still exists. Go look at China’s buildout of Coal-Fired plants.
Nat gas will work on the margins.
Maybe some oil.
Hydro is interesting as we have the spare capacity.
Wind is a filthy power source. This statement shouldn’t be shocking in 2025.

That leaves the SOLBAT trade. Solar generation, utility and local scale, coupled with batteries.

There Is No Alternative.

Here’s a weekly chart of TAN:

Solar’s been in a 4 year bear market. Completely justified! There was an overbuild of solar production, then interest rates skyrocketed so financing new projects got tougher. Input costs went up, cancellations hit, and then the world realized that they couldn’t afford to subsidize the industry anymore.

Honestly, good riddance. It got so bad that they were cutting down virgin forests in Scotland to put in solar farms. Scotland. The veritable land of sunshine.

The bearish solar narrative reached a peak late last year, with a few big bankruptcies and capitulative price action.

But with AI power demand, solar can stand on its own, regardless of subsidies. Instead of being paid with inefficient grant money, capital allocators will build out solar in the Sunbelt instead of Scotland.

Don’t Buy The Shovels, Buy the Shovelers

When I want to play a sector that’s been killed, I want to buy convexity. This usually is in the Services of the industry, names that are incredibly leveraged against the success of the big players in the industry..

For example, if you think deepwater oil is coming back in vogue, then you want to make a play in Vallartis (VAL) and Noble (NE).

Or if you think biotech is going to have a moment, maybe you buy Rapid Micro Biosystems (RPID) back at $1.

In fact, the biggest winners in the [AI | Data Center] trade have been services plays. Vertiv (VRT) has been an absolute monster as a Data Center Service stock:

I’ve found a Solar Service Stock that was absolutely killed. We’re talking down 90%. Left for dead.

Yet it’s shown relative strength against the sector, holding its lows while TAN had one more push to the downside.

This is the Convexity I’m looking for. It’s a smallcap stock that can easily double as it drafts from the tailwinds of the New Solar Revolution.

For Convex Spaces clients, I’ll be sharing the stock pick below. If you’d like to get my full deep dive on this name, then become a client today.