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

Pay more. Get less.

It’s quite the odd sales pitch. But during the pandemic, car dealers could apparently make it work.

In our latest episode of Being Exponential with Luke Lango, we’re joined by Money Flows’ Jason Bodner who recalled shopping for a car during the chip shortage. You’d pay a premium for less features. Even a powered tailgate was too much to ask.

The automobile industry could build the big, expensive parts. But without enough of the little electronic ones, buyers had to settle.

There’s an investing lesson tucked inside that inconvenience somewhere: A component doesn’t have to be glamorous to become valuable. It has to be difficult to replace when everybody needs it.

Which brings us to artificial intelligence.

Investors spend a great deal of time debating which AI model will be smartest, which assistant will win our loyalty, and whether the whole business has become a bubble.

All fair questions. Meanwhile, someone has to supply the memory for those systems and the optical connections that transmit information between chips.

That’s where our conversation got interesting.

I invited Jason and fellow Money Flows co-founder Lucas Downey to defend the AI bull market while I tried to poke holes in it.

You heard that right… I took the AI bear case.

Semiconductors are cyclical. Customers can overspend. A fat backlog today doesn’t guarantee fat profits tomorrow. And investors who mistake peak earnings for permanently higher earnings can discover that a “cheap” stock was actually pretty expensive.

Their response centered on rising earnings estimates, supply constraints, and signs of institutional money returning to parts of the AI infrastructure trade.

Lucas singled out photonics — technology that uses light to transmit information — as a potential next major opportunity.

But the price chart alone won’t tell you which is happening. A stock can climb sharply and still become cheaper relative to its earnings, if those earnings grow faster. It can also fall sharply without becoming a bargain.

So before you decide the AI boom has already made its money, consider the lesson from that car lot.

The next opportunity may lie in an essential part that everyone needs.

In this episode, we’ll examine where those shortages are emerging, what the money flows suggest, and how to separate businesses delivering profits from stocks selling promises:

Is the Semiconductor Boom Nearing Its Peak?

We’ve seen semiconductor shortages before.

Demand surges. Manufacturers expand capacity. Customers stockpile chips because they’re afraid of running out. Eventually, supply catches up, orders slow, and yesterday’s shortage becomes tomorrow’s inventory problem.

That’s why a low price-to-earnings ratio can be misleading in this industry. A stock trading at six times earnings looks inexpensive… until those earnings get cut in half.

I put that argument directly to Lucas and Jason. Several years into the AI infrastructure buildout, why should investors assume this cycle has more room to run?

Their answer was that the evidence they follow hasn’t signaled an earnings peak.

Lucas pointed to rising earnings expectations among memory and storage suppliers. Jason emphasized the multiyear demand outlook described by manufacturers and their customers.

Neither observation abolishes the semiconductor cycle. But both challenge the assumption that its reversal is imminent.

The question is whether the conditions that end a boom are actually developing.

Are customers cutting orders? Are earnings estimates falling? Is new capacity overwhelming demand?

Those are more useful questions than simply asking how long stocks have been going up.

In the discussion, I noted that the expected earnings peak for Micron (MU) had kept moving further into the future as estimates changed. A valuation built around an approaching collapse deserves another look if the business keeps pushing that collapse further away.

It also deserves continued scrutiny, as forecasts can change in either direction.

The Spending Has to Earn Its Keep

Now, there’s a second objection: What if the companies buying all this infrastructure eventually decide they aren’t earning enough from it?

That’s the harder question.

A company can announce an enormous spending plan. It can sign contracts, reserve capacity, and promise shareholders that AI will transform its business. Eventually, the economics have to justify the spending.

I asked whether enterprise demand could persist long enough to support the buildout.

Jason, ever the practical one, described himself as a surfer of money flows, watching the conditions in front of him. His investment process combines institutional trading activity with business fundamentals: sales, earnings, margins, and financial strength.

That framework doesn’t settle the ultimate return on every AI investment. It does help distinguish a possible future problem from deterioration already showing up in the data.

For investors, that means keeping two ideas in mind: The long-term spending debate remains unresolved, while suppliers can still have substantial opportunities serving demand that exists today.

Is This the Next AI Bottleneck?

That brings us to one of the most interesting parts of the conversation: photonics.

Buying more powerful processors is only part of building a more capable AI system. Those processors also need to exchange enormous amounts of information.

Imagine expanding a warehouse until it can process twice as many orders while leaving its loading docks unchanged. At some point, moving goods becomes the constraint.

AI infrastructure faces its own version of that problem.

Photonics uses light to transmit information. As computing systems grow, optical connections become increasingly important to moving data efficiently between their components.

Lucas identified this as a potential next major investment theme. We discussed Marvell (MRVL) and Corning (GLW) as two companies with exposure to that buildout.

Marvell supplies semiconductor technology used in data infrastructure, including optical connectivity. Corning brings fiber and optical communications expertise.

They occupy different parts of the system. The investment question is how much growing demand translates into revenue, profits, and durable competitive advantages for each.

A compelling technology story is the beginning of that analysis.

Shareholders still need the business to capture value, and the purchase price to leave room for a worthwhile return.

Which raises another question I pressed during the episode: If the opportunity is so attractive, why had several optical stocks spent months going sideways?

Lucas and Jason pointed to the distinction between business performance and trading pressure.

They discussed forced selling, portfolio adjustments, interest rates, and uncertainty as possible explanations for weakness. Their money-flow analysis was aimed at identifying whether institutions were continuing to sell or beginning to return.

But selling can occur for reasons unrelated to a company’s operating outlook. A leveraged investor facing a cash demand may sell a good business simply because it can be sold.

But investors shouldn’t turn that possibility into a blanket excuse for every falling stock. Sometimes a price decline is warning you about a business problem. The work is to compare the selling with what’s happening to orders, earnings expectations, and competitive positioning.

How the AI Winners Could Change

There’s also a broader point here that gets lost when investors treat “AI stocks” as one trade.

The companies benefiting from this buildout can change.

While critical components remain scarce, suppliers may enjoy stronger pricing power. If those components become cheaper, the companies buying them could benefit from lower costs.

That shift would create winners and losers within the AI economy. It wouldn’t automatically end the opportunity to invest in it.

Eventually, the benefits could reach further.

We discussed whether businesses outside technology could improve their margins as they learn to apply AI. Technology companies have an obvious head start. Retailers and other consumer businesses may take longer to turn experimentation into measurable results.

That potential broadening is worth watching. It still has to show up in the numbers.

For now, my takeaway from trying to play the bear is this: The bullish case is strongest when it rests on specific businesses delivering specific results.

Jason warned against paying extravagant prices for companies that have barely begun to monetize their promises. Lucas emphasized opportunities investors may overlook inside the broader AI theme.

I agree with both.

You don’t need to assume every spending commitment will pay off. You do need to distinguish an essential supplier with growing profits from a company whose main asset is an exciting presentation.

Keep watching earnings revisions. Watch whether shortages persist as capacity expands. Look for evidence that institutional buying is supported by improving fundamentals.

And remain willing to change your mind when those conditions change.

The missing tailgate offers a useful starting point. Follow the shortage to the supplier, then follow the supplier’s economics.

That’s where the story becomes an investment case.

Watch the full Being Exponential conversation for our debate on memory, photonics, institutional money flows, and the risks that could change the outlook.