Nvidia Has a Heat Problem. These Stocks Get Paid to Solve It.

Listen to the audio version of this article (generated by AI).

Editor’s Note: The best technology investments I’ve ever seen have one thing in common: they weren’t obvious at the time.

Joe Austin has spent 40 years developing a nose for exactly those opportunities — the infrastructure companies, the component makers, the “picks and shovels” plays that make a megatrend physically possible and get overlooked while everyone chases the headline names.

Today he makes the case that AI has a whole layer of those companies hiding in plain sight. He opens with a NASCAR story that I bet you’ll still be thinking about by the time you get to his investment thesis.

He and Marc Chaikin shared the full picture earlier this week during a special free broadcast. Click here to watch the replay.

Here’s Joe…

Denny Hamlin was ready to clinch his first NASCAR championship. But it all fell apart because of a piece of tape.

Hamlin was considered one of the best drivers of his generation, but the title had always eluded him.

In 2019, he was NASCAR’s comeback kid. He’d already won six times that season — including the iconic Daytona 500. The championship all came down to a cool November night at the Homestead-Miami Speedway.

The day started well. Qualifying had been canceled due to bad weather, which gave Hamlin the pole position based on his points standing. In the first two stages, he finished fifth. By the third stage, things were looking up — he was running second, posting quicker lap times than his teammate Kyle Busch, who led the race.

Then, with 58 laps to go, crew chief Chris Gabehart made a fateful call. He brought Hamlin in for an early pit stop and had the crew slap a big piece of black tape on the front grille.

The idea was sound, in theory. In NASCAR, crews routinely tape grilles to gain a competitive edge. Normally, air enters the grille and bounces around the engine, creating drag. Tape restricts that airflow, forcing it to flow smoothly over the car instead. More speed, more downforce, better grip.

But tape also restricts cooling.

NASCAR engines run at around 290 degrees Fahrenheit — about 90 degrees hotter than a typical road car. The margin for error is razor-thin. In Hamlin’s case, the tape backfired immediately. Temperature gauges maxed out. Steam started spewing from the engine. Engine failure seemed imminent.

Gabehart had to call Hamlin back to the pit after just 12 laps. He finished 10th — dead last among the four championship contenders. Kyle Busch won both the race and the title.

Excessive heat isn’t just a NASCAR problem. Data centers running AI chips face the exact same dilemma. 

And for investors, it represents one of the most overlooked opportunities in the entire AI boom.

The Physics Problem That Makes ‘Behind the Scenes’ Companies Essential 

AI chips need massive amounts of power to train models and run computations. More power means more heat. And if you can’t cool the chips fast enough, performance crashes, or the hardware fails entirely.

Inside a data center are long rows of computer racks — tall cabinets stacked with servers. A typical AI data center contains hundreds or even thousands of them. Nvidia Corp.’s (NVDA) next-generation Vera Rubin chip uses 120 to 130 kilowatts per rack. That’s the annual electricity consumption of about 100 U.S. homes — per rack.

Bigger versions of the Rubin chip will use five times that much power.

That creates an unavoidable physics problem. More power means more heat, in a nearly one-to-one relationship. And delivering this much electricity requires completely rethinking how power gets moved through a building.

Think of it like water through a hose. You can deliver the same volume using high pressure through a small hose or low pressure through a massive one. Electrical power works the same way — high voltage with low current, or low voltage with high current. High current dangerously overheats cables. Traditional power systems can’t handle it.

Engineers solved this by raising the voltage. The industry has shifted to 800-volt power systems, which deliver the same power with far less current and far less heat.

But operating at 800 volts requires power chips made from entirely different materials. Materials that only a handful of companies know how to produce.

This is the bigger point about investing in a megatrend like AI.

The AI technology itself gets all the media attention. Nvidia, Microsoft Corp. (MSFT), Palantir Technologies Inc. (PLTR): Everyone knows those names. 

But an entirely different set of companies — the ones making the components, materials, and systems that let AI physically function — are just as essential. And far less picked over.

No matter which big-name AI company is building the next data center, the “behind the scenes” businesses making it all run are going to get paid. The question is whether you own any of them.