AI Stock Outlook: Memory and Data Bottlenecks May Shape the Next Semiconductor Winners
Sep 29, 2026•5 min read• views
## The Next AI Stock Winners May Be Hiding in the Bottlenecks
The AI buildout depends on more than powerful chips. Memory and the connections that move data between processors can become valuable when demand outruns supply. That’s the investment lesson in a shortage: an unglamorous component matters when customers can’t easily do without it.
During the pandemic chip shortage, car buyers sometimes paid more for vehicles with fewer features. Money Flows co-founder Jason Bodner recalled that even a powered tailgate could be out of reach. Automakers could build the large, expensive parts; missing electronics held back the finished product.
That example frames the debate in the latest episode of *Being Exponential with Luke Lango*. I asked Bodner and fellow Money Flows co-founder Lucas Downey to make the case for AI stocks, while I tested the bearish view: semiconductor demand is cyclical, customers can overspend, and today’s backlog doesn’t guarantee tomorrow’s profits. A low price-to-earnings ratio can be deceptive if earnings are near a peak.
Their counterargument rested on rising earnings expectations among memory and storage suppliers, persistent supply constraints, and signs that institutional investors may be returning to parts of the AI infrastructure trade. Downey also pointed to photonics, which uses light to transmit data, as a possible next opportunity.
The price chart alone won’t settle the question. A stock can rise while becoming cheaper relative to its earnings if profits grow faster; a falling share price can still leave investors with an expensive business. The more useful test is whether demand, earnings estimates, and supply conditions support the story.
That’s where we’ll focus: which AI infrastructure bottlenecks may create openings, what money flows can tell us, and whether companies are turning demand into profits rather than promises.
## Is the Semiconductor Boom Nearing Its Peak?
The semiconductor cycle gives investors good reason to be cautious. Demand spikes, manufacturers add capacity, and customers stockpile chips. When supply catches up, orders can cool and shortages can turn into excess inventory. A stock trading at six times earnings may look cheap until those earnings are cut in half.
I put that concern to Downey and Bodner: several years into the AI infrastructure buildout, what suggests the cycle still has room to run? They said the evidence they track hasn’t yet pointed to an earnings peak. Downey cited rising earnings expectations for memory and storage suppliers; Bodner pointed to the multiyear demand outlook described by manufacturers and customers.
Neither signal makes a downturn impossible. Forecasts can change, and semiconductor cycles have a way of doing so. But they do challenge the assumption that a reversal is imminent. The key questions are whether customers are cutting orders, earnings estimates are falling, or new capacity is starting to outstrip demand.
Micron’s expected earnings peak, I noted, has repeatedly moved further into the future as estimates changed. That makes a valuation built around an imminent collapse worth revisiting—while keeping a close eye on whether the forecasts hold.
## Spending Has to Earn Its Keep
The harder question is whether companies buying AI infrastructure will eventually earn enough from it to justify the bill. Big spending plans, contracts, and reserved capacity show commitment. They don’t prove that the investment will pay off.
I asked whether enterprise demand could last long enough to support the buildout. Bodner described his approach as surfing money flows: tracking institutional trading activity alongside sales, earnings, margins, and financial strength. That combination can help investors see whether a concern is already showing up in business results, though it can’t resolve the long-term returns on every AI investment.
For investors, the distinction matters. The debate over eventual returns remains open, while suppliers may still be serving substantial demand today.
## Is Photonics the Next AI Bottleneck?
More capable processors are only part of the equation. They also have to exchange vast amounts of data. Think of a warehouse that doubles its order-processing capacity without expanding its loading docks: sooner or later, getting goods in and out becomes the constraint.
AI systems face a similar challenge. Photonics uses light to transmit information, and optical connections help move data between components. Downey identified the field as a possible next investment theme. The conversation covered Marvell (MRVL), which supplies semiconductor technology for data infrastructure, including optical connectivity, and Corning (GLW), known for fiber and optical communications.
They serve different parts of the system. The investment case depends on whether rising demand becomes revenue and profit, and whether either company can defend its position. A promising technology is a starting point; shareholders still need the business to capture value at a price that leaves room for a return.
That raises a fair question: if optical stocks look attractive, why have some traded sideways for months? Downey and Bodner discussed forced selling, portfolio adjustments, interest rates, and uncertainty as possible sources of pressure. Their money-flow analysis looks for signs that institutions are still selling or beginning to return.
Sometimes investors sell a strong business because they need cash. But that explanation can’t excuse every decline. Compare the selling with orders, earnings expectations, and competitive position to judge whether the market is reacting to trading pressure or deteriorating fundamentals.
## How the AI Winners Could Change
“AI stocks” isn’t one trade. The companies benefiting from the buildout can change as shortages ease and costs shift. Scarce suppliers may gain pricing power; when components become more available and cheaper, companies buying them may benefit instead.
The opportunity could spread beyond technology, too. Retailers and other consumer businesses may eventually improve margins by applying AI, but turning experiments into measurable results takes time. The potential is there; the evidence still has to appear in the numbers.
My takeaway from arguing the bear case is straightforward: the strongest bullish arguments point to specific businesses delivering specific results. Bodner cautioned against paying extravagant prices for companies that have barely begun to monetize their promises. Downey highlighted opportunities investors may overlook within the broader AI theme. Both points deserve attention.
Track earnings revisions, watch whether shortages persist as capacity expands, and look for institutional buying backed by improving fundamentals. Be ready to change your view when those signals change. The missing tailgate is a useful starting point: follow the shortage to the supplier, then examine the supplier’s economics.
Watch the full Being Exponential conversation for the debate on memory, photonics, institutional money flows, and the risks that could change the outlook.
About the analyst
Luke Lango is editor of Hypergrowth Investing. His work focuses on identifying early investment opportunities in fast-growing industries and the broader trends shaping them. You can learn more about Luke.