AI stocks have taken a beating this month.

Companies with real customers, growing revenue, and billions of dollars in contracts are getting hammered right alongside companies that have little more than a good story.

To most investors, that might sound like a reason to stay away.

But in reality, this is exactly the kind of market that can create huge opportunities.

The key is learning to separate future winners from companies that may never recover.

Fortunately, there’s a simple 3-step system to help you do that.

Step 1: Follow the bottleneck

Every major growth trend eventually hits a bottleneck—where demand grows faster than supply.

In the early days of the AI boom, chips were the bottleneck. Companies were scrambling to get their hands on Nvidia’s processors, and investors who recognized the shortage early made fortunes.

But the bottleneck is shifting. All those chips need servers to house them… cooling systems to keep them from overheating… and enormous amounts of electricity to run.

That’s where the next shortages are developing.

Just look at Vertiv (VRT) and Celestica (CLS). Vertiv provides power and cooling equipment for data centers, while Celestica builds the hardware that connects and supports computing systems. Back in 2022, Vertiv traded around $9 and Celestica around $8. Both have since risen more than 20-fold.

Investors who recognized those bottlenecks early saw enormous gains.

Find what the industry desperately needs before everyone else realizes there’s a shortage.

Step 2: Find what competitors can’t easily copy

Spotting a shortage is only the beginning. The next question is: How hard is it for competitors to solve that shortage?

Investors should look for companies that already control something valuable and that can’t simply be built or bought overnight.

Take electricity. Getting a new data center connected to the power grid can take years. So a company with access to large amounts of power has a major head start over competitors still waiting for approval.

The same principle applies to companies with established customer contracts, specialized infrastructure, or the ability to bring new computing capacity online quickly. Those advantages become especially valuable when customers are racing to build.

But investors need to distinguish between companies that already have these advantages and companies that are merely promising to develop them. A “potential customer” isn’t a signed contract.

Favor companies with assets, customers, or capabilities that competitors can’t easily replicate.

Step 3: Wait for the market to throw everything away

Finding a great business doesn’t automatically make it a great investment. You also need the right price. And sometimes the best prices appear when investors panic.

That’s what’s been happening across AI stocks this month. Concerns about debt, financing costs, and whether AI spending will generate enough revenue have sent shares tumbling.

Some companies deserve that pressure… But others are getting punished simply because they operate in the same industry.

When the market sells everything indiscriminately, it can create a chance to buy stronger businesses at prices that were unavailable just weeks or months earlier.

But a falling stock isn’t automatically a bargain. Behind the decline must be a real business with customers, demand, and a reason to believe it can keep growing after the panic ends.

The key is separating a temporary stock price decline from a permanent business problem.

Look for strong businesses getting punished alongside weaker competitors.

4 beaten-down AI stocks on Frank’s radar

Based on the three steps above, here are four AI infrastructure stocks worth a closer look. Each has exposure to a critical part of the AI buildout… an advantage that could be difficult to replicate… and a stock price that’s come under pressure.

1. CoreWeave (CRWV)

CoreWeave rents access to Nvidia’s AI chips and computing infrastructure. Its customers include Microsoft, Meta, and OpenAI. It also has specialized infrastructure and relationships that a new competitor can’t instantly reproduce.

Yet the stock has fallen more than 60% from its high amid concerns about its massive debt load.

The question is whether investors have become too pessimistic about a company serving some of the biggest buyers of AI computing power.

2. Super Micro Computer (SMCI)

Super Micro builds the powerful servers that house AI chips. And demand for those servers is showing up in its sales outlook.

Last week, the company raised its quarterly revenue forecast to $18.5 billion–$19.5 billion, up from its previous estimate of $14.5 billion–$15.5 billion.

Yet shares remain nearly 40% below their high.

That disconnect between growing demand and a beaten-down share price is exactly what investors should look for during a selloff.

3. DigiPower X (DGXX)

DigiPower X is a former Bitcoin miner converting its power infrastructure into AI data centers. Its biggest advantage is access to something that can take competitors years to secure: electricity.

And it’s already attracting customers. AI chip company Cerebras has signed an agreement to take full capacity at DigiPower X’s 40-megawatt Alabama facility. The agreement is valued at about $1.1 billion over 10 years.

The company also announced a $200 million debt financing commitment, syndicated by Goldman Sachs, to help complete the project. It remains subject to closing conditions, but the proposed project-level financing could reduce the need to issue new shares and dilute existing investors. The first phase is expected to be completed in December.

The appeal lies in the combination of secured power, a contracted customer, and a clear path to bringing the site online.

4. VivoPower (VIVO)

VivoPower owns powered infrastructure in the Nordic region, where access to hydroelectric electricity provides an advantage for energy-hungry data centers.

The company recently began construction work on a 40-megawatt AI data center conversion in Mo i Rana, Norway.

It has also selected a global AI industry leader as its preferred tenant, although it hasn’t publicly named the company or announced a finalized lease.

VivoPower also has the flexibility to lease powered space rather than build out every part of a data center itself.

The bottom line

The easiest time to buy a popular stock is when everyone agrees it’s a winner. But that’s rarely when the biggest bargains appear.

A better approach is to look for an industry bottleneck, find companies with an advantage in solving it, and wait for a selloff to create a better entry price.

That’s the opportunity in AI stocks today. The market is punishing companies across the sector. But not every company deserves the same punishment.

And knowing how to tell the difference could be the key to finding the next big winner.

Want to find out how Frank identifies the most promising growth stories… before the rest of Wall Street catches on? Check out his 6 principles for finding great stocks.