5 AI Infrastructure Stocks to Buy the Dip
Five AI infrastructure names just flipped from falling knives to real bounces, and the hyperscalers gave us the all-clear to act on it
Listen to the audio version of this article (generated by AI).
President Dwight Eisenhower, known to the country as Ike, blamed the onions.
He had eaten a hamburger with raw Bermuda onion at the turn, walked off the course at Cherry Hills outside Denver with what he figured was a bad stomach, and went to bed. The chest pain woke him after midnight. The words “heart attack” emblazoned the newspaper headlines on a Saturday in September 1955, which gave the country a day and a half to sit with the news before the New York Stock Exchange opened.
The following Monday was a slaughter.
The Dow Jones Industrial Average fell 6.5%, about $14 billion of value, the worst single session since the crash of 1929. And nothing about the American economy had changed over that weekend. The steel mills were still functioning, Detroit was still churning out vehicles, the baby boom kept booming. What had changed, however, was the story investors were telling themselves about the future…
The market regained its footing within weeks. The next bear market did not arrive until the spring of 1956.
We just lived through our own version of that weekend. A single piece of writing about the situational awareness of AI models blew up the internet in late July, the AI infrastructure complex went into free fall, and some of the best growth stories in this market were cut in half or worse.
Then the “medical bulletins” (read: earnings) came in.
All four hyperscalers hiked capital expenditure guidance for 2026. All four gave bullish commentary on 2027. Amazon.com Inc. (AMZN) confirmed it on that Thursday, and the following session was the real back-up-the-truck moment.
We do not catch falling knives around here. We buy bouncing tennis balls. Right now the tennis balls are bouncing, and here are five of them we broke down on this week’s Being Exponential:
The Divergence
Before we get into the buy-the-dip names, let’s discuss the framework, because the framework is what you keep with you.
Every one of these stocks shows the same fingerprint. Forward 12-month earnings estimates go up and to the right in a clean, uninterrupted line while the share price nosedives. That divergence is fundamentally incongruent. Analysts who cover these companies are raising their numbers at the exact moment traders are dumping the shares, which tells you the selling is about sentiment rather than about the business.
The second tell is the 200-day moving average. Support that flips to resistance is a warning. A stock that keeps hitting its head on the same line in January, March, June, and July is a stock in a downtrend. The moment it knifes back through and holds, the character of the chart changes. That is the trigger. Divergence gives you the reason. The retake gives you the timing.
Palantir: The Tug-of-War Is Over
Palantir Technologies Inc. (PLTR) spent this year caught between SaaSmageddon fears and the reality of organic AI software growth. This quarter settled it emphatically.
Revenue grew 93% year over year, and more importantly grew 19% quarter over quarter. When you own a hypergrowth story, you want each three months better than the last three months. Commercial revenue rose 149% year over year and 28% sequentially, which answers the only real question anyone had about this company: how deep can Palantir push into cost-sensitive commercial buyers? Very deep, apparently. Net dollar retention of 157%. Adjusted operating margin of 62%. Free cash flow margin of 63%. A Rule of 40 score around 155%.
The knock has always been valuation. At roughly 60 times forward earnings, Palantir now trades near a two-year low multiple and within shouting distance of a five-year low, against growth I believe compounds above 50% into the end of the decade. I recommend buying this bounce, because the stock just retook the 200-day for the first time since January.
Caterpillar: The Industrial Fiber of the Buildout
Caterpillar Inc. (CAT) is the literal picks and shovels play. Prime power, backup power, the machines that move the earth a data center sits on. The buildout runs on Caterpillar iron.
Revenue rose 24% to $20.5 billion. Adjusted earnings per share jumped 73% because adjusted operating margin expanded 430 basis points to 21.9%. Free cash flow in the machinery, energy, and transportation business hit a record $5.1 billion. And the backlog climbed $9.4 billion sequentially to $72.1 billion, which means the numbers we saw this quarter are the numbers we see for the next several quarters.
Analysts model single-digit revenue growth for this year. I think that is far too low. At about 22 times forward earnings for a stable cash cow compounding earnings north of 25%, this one is priced like the buildout ends tomorrow. The chart already disagrees, because Caterpillar took a 26% drawdown, lost the 50-day and the 100-day, then V-shaped right back and reclaimed the 100-day as support.
Applied Optoelectronics and Fabrinet: the optics squeeze
AI clusters are bandwidth-hungry, and every incremental gigawatt of compute demands 800G and 1.6T transceivers, data center interconnect, silicon photonics, and eventually co-packaged optics.
Applied Optoelectronics Inc. (AAOI) is the highest-growth name I am aware of in optical networking at real scale, with triple-digit revenue growth expected this year and next and gross margins tracking from the low 30s toward the high 30s. It fell more than 65% from its July peak. It has already retaken the 200-day. That is the highest-torque expression of this theme.
Fabrinet (FN) is the quieter one. Fabrinet manufactures the hard stuff other people design, and that expertise is difficult to replicate. Call it 25% compounded revenue growth with thin gross margins that stay thin, which still delivers 30% earnings growth, at about 25 times forward estimates near a one-year low. Fabrinet reclaimed its 200-day this week and ended a run of lower highs and lower lows.
Nebius: The Wall Street 2K stock
Nebius Group N.V. (NBIS) exists because we are desperately short compute. Every hyperscaler launching more cloud capacity reinforces that, because scarcity gives every supplier of compute enormous pricing power.
Revenue is expected to leap from roughly $530 million in 2025 to $3.3 billion in 2026, then to nearly $11 billion, then toward $20 billion. Gross margin runs from about 41% this year toward 80% by the end of the decade. If you could build a stock in a video game, you would build this one. It trades near 13 times forward earnings. The stock dropped 40%, bounced clean off the 200-day, retook the 100-day, and is working on the 50-day.
The Window Is Open
This is the aggressive part of the cycle, and I say that with both hands on the wheel. The music is playing. We are in later innings, and later innings are where the big returns and the big mistakes both live.
Get the full breakdown, chart by chart, on this week’s episode of Being Exponential. Then join us later this week for our macro episode, where we tackle the situational awareness debate head-on.
In addition to the five buy-the-dip plays I mentioned today, there’s one $15 stock that could soar over the coming weeks and months. Hint: It involves Elon Musk, AI, China…
There will be a headline that spooks you. Similar stocks that drop for reasons that have nothing to do with the business underneath it. A moment where it looks, on the surface, like the story is falling apart.
That moment, if history is any guide, is exactly when the money that matters moves in.
The only question left is whether you’re positioned before that moment or after it. In fact, if you buy just one stock for the rest of the year, I urge you to make it this one.
Click here to see what you’re missing.
