Via marketbeat.com
A beat-and-retreat pattern in AI infrastructure stocks is raising questions about how much growth is already priced in
Fabrinet posted a quarter that would make most CEOs pop champagne. Revenue hit a record $1.316 billion, up 45% year-over-year. Non-GAAP earnings per share came in at $4.10, comfortably topping the $3.81 consensus. The stock initially climbed about 5% on the news. Then, in after-hours trading, it cratered somewhere between 9% and 20%.
The numbers looked great on paper #
Fabrinet’s fiscal Q4 2026 results were strong by virtually any conventional measure. The $1.316 billion in quarterly revenue blew past analyst expectations of roughly $1.28 billion. Full fiscal year 2026 revenue reached $4.64 billion, reflecting 36% growth from the prior year.
For Q1 of fiscal year 2027, Fabrinet projected revenue in the range of $1.375 billion to $1.425 billion. Non-GAAP EPS guidance of $4.10 to $4.25 also exceeded what the Street had penciled in. But investors looked past the top-line fireworks and zeroed in on something less flattering: margins, capital expenditure requirements, and the question of whether the stock’s valuation had simply run too far, too fast.
Profit-taking hits the whole neighborhood #
The sell-off wasn’t contained to Fabrinet. Marvell Technology and Seagate each dropped approximately 8% on the same trading day, caught in the gravitational pull of souring sentiment around AI-linked semiconductor and optics names. Amphenol, the connector and cable giant that supplies critical components for data center buildouts, also felt the pressure.
What the sell-off signals about AI infrastructure valuations #
Fabrinet’s 45% revenue growth and above-consensus guidance suggest demand for optical networking components and precision manufacturing remains robust. The issue is valuation math. When a stock has already priced in several quarters of blowout growth, even a strong report can trigger selling if investors see any cracks in the narrative. In Fabrinet’s case, concerns about profit margin sustainability and rising capital expenditure needs gave sellers the justification they needed.
For Marvell, which derives a growing share of its revenue from custom AI chips and data center connectivity, the read-through is particularly relevant. The company’s exposure to the same end markets that drive Fabrinet’s business means sentiment shifts in optical networking tend to bleed into semiconductor names with AI exposure. Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our