Key Points
- AI infrastructure stocks are selling off even as suppliers report sold-out capacity, rising guidance and multi-year backlogs, according to TradeSmith's Lucas Downey.
- Lumentum, Dell Technologies and Everpure each show falling share prices alongside rising earnings estimates, compressing forward valuations and creating a widening gap between price and fundamentals.
- Downey views the decline as driven by investor positioning and seasonal weakness rather than broken fundamentals, and says upcoming earnings revisions will determine when the sell-off ends.
- Special Report: 100% overnight gains?
The artificial intelligence (AI) trade finally cracked. After months of resilience while utilities, real estate and consumer discretionary names took a beating, the market's favorite growth story is now the one getting sold, and investors sitting on big gains are rushing to lock them in.
The fear has a familiar shape. Earlier this year, software stocks were written off as AI roadkill, and the slide dragged on for roughly 60 days. Many of those same names have since reclaimed all-time highs, because the earnings never broke the way the headlines said they would.
That's the tension now. Prices are falling across AI infrastructure while the companies building it keep raising guidance, reporting sold-out capacity and pointing to backlogs that stretch for years. When price and fundamentals move in opposite directions, valuations compress, and that gap is where opportunity tends to surface.
AI Sell-Off Looks Driven by Positioning, Not Broken Fundamentals
The selling looks more like positioning than a verdict on the businesses, per Lucas Downey, editor at TradeSmith. Many investors came into the slide underweight in the interest-rate-sensitive stocks that look most attractive as rates come down, while sitting on large embedded gains in AI. Selling the winners is the fastest way to rebalance, while some heavily leveraged investors may have no choice but to sell.
Seasonality adds friction. Late September has historically been weak, and midterm election years tend to run volatile as investors handicap the fight for Congress. Downey sees signs of capitulation and puts a potential bottom in the next couple of weeks, with earnings season the likely catalyst. By his read, further downside remains on the table.
The deeper worry, though, is whether hyperscalers can keep paying for the build-out as data center costs climb. Downey points to what suppliers are reporting. Taiwan Semiconductor Manufacturing (NYSE: TSM) and ASML Holding (NASDAQ: ASML) have described demand as strong and, in many cases, accelerating, and committed contracts stretch out for years.
Optical Components Sell Out Into 2029 as Earnings Estimates Rise
Lumentum Holdings (NASDAQ: LITE) makes the lasers and optical components that move enormous volumes of data at high speed inside AI data centers. As demand runs into the limits of materials and manufacturing capacity, those parts become a bottleneck.
CEO Michael Hurlston has said the company is sold out of optical components for AI servers through early 2029. Sales over the last 12 months came in just over $3 billion, and estimates cited by Downey call for about $6.3 billion over the next 12 months, approaching $10 billion the year after. Earnings per share (EPS) estimates sit at $21.70 for fiscal 2027 and $47.88 for fiscal 2029.
A roughly 400% gain over the past year is exactly what makes latecomers nervous. But the stock has dropped about 15% in a week, and with estimates rising as the price falls, the forward price-to-earnings (P/E) ratio has compressed to about 42. The market is pricing a momentum unwind. Another round of upward estimate revisions could change that, so earnings follow-through is the thing to watch.
AI Server Demand Fuels a Blowout Beat and Raise
Dell Technologies (NYSE: DELL) spent decades as the personal computer brand everyone knew. Its infrastructure business, spanning servers, storage, networking equipment and software, has turned it into a cornerstone of the AI build-out, and the stock has climbed more than 200% in a year.
The latest quarter was a beat and a raise. Per Downey, Dell guided third-quarter EPS to $6.50 against estimates of $4.46 and revenue to $49 billion against a consensus just above $41 billion, then lifted its fiscal 2027 outlook. In his view, the stock effectively repriced on that report.
A surprise that big raises the bar, and priced-for-perfection risk is real. The counterweight is valuation. With forward estimates climbing while shares slide, the forward P/E has fallen to around 18, below the broader market. The market is pricing an AI hardware hangover. Next quarter's results are what to watch, since they test whether this guidance was a one-off.
Enterprise Flash Storage Lands Hyperscale Wins and a Higher Outlook
The smallest of the three is Everpure (NYSE: P), which builds all-flash data storage systems, the infrastructure companies use to manage data and run AI workloads, at enterprise scale. Management had flagged hyperscale customer wins, and those are starting to show up in results.
Everpure reiterated guidance and raised its preliminary fiscal 2028 outlook to $7.15 billion at the midpoint, against a Street estimate of $6.2 billion. That's a message of acceleration, not slowdown. The stock also scores an 88 on TradeSmith's proprietary Quantum Edge system, which blends fundamental and technical factors into a single rating; TradeSmith reads that as a high-quality stock.
Storage and memory names such as Micron Technology (NASDAQ: MU) and Sandisk (NASDAQ: SNDK) have been caught in the same downdraft, even as workloads keep growing. A smaller company brings bigger swings, and stocks don't move in straight lines. What could change sentiment is continued upward movement in analysts' EPS forecasts, and that trend is the one to watch.
Earnings Revisions Hold the Key to the End of the AI Sell-Off
Lumentum, Dell and Everpure sit at different links in the same chain: optics that move the data, servers that process it and storage that holds it. What connects them is a widening gap—falling share prices against rising forward numbers.
The risk is that the hyperscaler spending debate drags on, flows stay negative, and the smaller names take the hardest hits, much like software's two-month grind. The upside is that sold-out capacity, multi-year backlogs and raised guidance are company-reported numbers, and compressed valuations leave room to recover if earnings season confirms them.
None of this requires calling the exact bottom. It does require hyperscalers to keep spending and these companies to deliver on guidance already issued. Stay focused on earnings revisions, because that's what moves these stocks once the selling runs its course.
For more of Downey's AI research and TradeSmith's fundamental scoring tools, get the TradeSmith Investment Report and its AI stock recommendations.
Read this article online ›
The best investment opportunities don't wait. Get our research and stock ideas delivered straight to your smartphone—so you never miss a market-moving opportunity. Our text alerts ensure you see timely stock ideas and professional research reports instantly, whether you're in a meeting, commuting, or away from your desk.
Get Text Alerts from American Market News (free)
Buying a house just got more expensive...
