Fabrinet's AI-Optics Growth Is Real. So Why Did the Stock Fall 19%?

The question the headline wants you to ask — can Fabrinet's AI-optics growth let it take on Lumentum and Applied Optoelectronics? — is the wrong one, because Fabrinet is not one of those companies. It is the factory that builds precision optics for the OEMs who are. So start with the number that actually moved: Fabrinet just reported the strongest quarter in its history — revenue up 45% to a record $1.316 billion, with AI data-center revenue up 68% and, for the first time, more than half the company — and the stock fell 19% the next morning and kept sliding. It trades near $416 today, still about 44% below its May high near $749. A record quarter gets punished that hard only when investors were paying for something the quarter didn't deliver.

Treat the "challenge LITE & AAOI" framing as PR until you know what Fabrinet actually sells, because it sells manufacturing, not transceivers.

The factory, not the brand

Fabrinet is an electronics manufacturing services company — the world's leading precision optical and electronic contract manufacturer — and it is explicitly not a semiconductor designer. Its customers are the OEMs, the ones who own the product, the roadmap, and the margins. This distinction is the whole story, and it's written directly into the price. Lumentum trades at roughly 29 times sales and Applied Optoelectronics at about 16 times; Fabrinet trades at about 3. The gap is not the market being stupid about growth. It is the market correctly pricing a business whose gross margin sits near 12% against product designers who command far higher ones. When you have no pricing on the product because you never owned the design, you get a factory multiple, however fast the line runs.

That is not a defect in the stock so much as a description of the position. Fabrinet's growth is a proxy for how much AI-optical manufacturing its customers choose to outsource — which is why the story has been so clean as Nvidia and the hyperscalers leaned on the optical supply chain. The company said its data-center interconnect business reached a $1 billion annual run rate by the end of fiscal 2026, up 140% year over year, and it talks about growth broadening beyond Nvidia. All genuine.

The catch is that a factory's economics are one bad quarter of utilization away from looking ordinary — and the market just re-learned that lesson.

The quarter the headline loved and the market hated

The headline numbers were every bit as good as advertised: full-year revenue of $4.6 billion, up 36%, non-GAAP earnings of $14.09, up 39%. Against that, the selloff looks perverse. It wasn't. Investors focused on the three lines that define whether a factory's growth is cheap or costly.

First, gross margin actually narrowed — about 30 basis points year over year, to roughly 12%. Second, free cash flow went negative at minus $37 million, and the culprit wasn't weak demand; it was a surge in spending on new capacity in Thailand and California. That is the definition of a growth ramp eating its own cash to feed it. Third, guidance for the coming quarter calls for revenue above consensus but earnings per share declining sequentially, hit by seasonality and a new Thailand "top-up" tax.

The market didn't argue with the growth; it argued with the price of the growth.

Fabrinet entered the report at roughly 33 times earnings after a year that had already run the stock up sharply. At that multiple there is no tolerance for a 30-basis-point margin slip, a negative cash quarter, or a guided EPS decline — all of which arrived within the first few bullet points of the release. This is the standard pattern when a sector has run on forecasts rather than evidence: the first sign of a cost curve, not a demand problem, collapses the narrative. The optics complex as a whole is priced like this — remember Lumentum at 29 times sales — so when a 45%-growth quarter can't hold a stock up, the re-rating is telling you the whole sector is priced beyond what its own engineering can evidence.

There is a second, quieter signal worth holding onto. Inside the booming numbers, the traditional optical data-communications product line slipped slightly quarter over quarter; the growth is coming from the high-performance-computing, AI-interconnect side. That mix shift is exactly what you would expect from a factory riding the AI buildout — and it means Fabrinet's fate is now directly tied to whether hyperscale AI-optical spending keeps compounding, not to any product advantage of its own.

For the investor, the resolution is simpler than the volatility suggests. Fabrinet's AI-optics growth is real, and it is a legitimate way to own the buildout without betting on which transceiver design wins. But the recent fall was the market applying the cost lens that had been skipped while the stock tripled: a 12%-margin contract manufacturer that consumes cash to expand sells on volume, not on pricing power, and its multiple now has to justify itself against a quarterly tax hit, thinner margins, and negative free cash flow. The record quarter was the good news; the selloff was the bill for how a factory actually makes money.