Market News 9 min read

Fabrinet Stock Analysis 2026: FN Share Price, Earnings and Outlook

Washington put Thailand on the AI chip licence list. Fabrinet assembles almost everything there, and had just grown quarterly revenue 51% to $1.47bn.

A single ornate key resting on top of a sturdy wooden crate sealed with fresh export stickers, set in the middle of an otherwise empty warehouse.

On Monday 16 November the US Commerce Department's Bureau of Industry and Security issued an interim final rule that changed the unit of control on advanced AI chips from the chip to the place. Alongside closing the loophole that let Chinese-headquartered firms rent advanced compute in third-country data centres, the rule adds licence requirements for advanced chip exports to five named jurisdictions — Indonesia, Malaysia, Singapore, Thailand and Brazil — described by BIS as countries presenting a higher-than-normal risk of diversion. Thailand is where Fabrinet physically builds and ships almost everything it sells.

That is the starting point for any Fabrinet stock analysis this week, and it cuts straight across the company's bull case. Fabrinet closed at USD 311.01 on Monday, down 5.2%, two weeks after reporting a quarter in which revenue grew 51% year on year to roughly USD 1.47bn, with data-centre revenue up about 78%. The growth is not in question. What changed on Monday is that the single-country manufacturing base which made that growth cheap to deliver is now also a licensing variable — and a business concentrated in one jurisdiction is worth less when that jurisdiction moves onto an export-licence list, even if not one shipment is ultimately refused.

What the rule actually does

The rule has three limbs, and the first two are the ones that reach the supply chain.

  • A compute-access threshold. A licence is required to ship advanced AI chips to any data centre, anywhere, where the arrangement would give customers headquartered in China, Macau or other Country Group D:5 destinations access to more than 12 million aggregate total processing performance. Capacity is counted across linked facilities, not just within one building, which is what makes it hard to structure around.
  • Five new licence jurisdictions. Indonesia, Malaysia, Singapore, Thailand and Brazil now carry licence requirements for advanced chip exports in their own right, independent of who the end customer is.
  • Rescission of the AI Diffusion Rule. The Biden-era framework is formally withdrawn, replaced by this narrower but more geographically explicit regime.

One thing is genuinely unsettled and we are not going to pretend otherwise: the timing. Coverage of the rule gives conflicting accounts of when it bites — publication in the Federal Register versus a 90-day runway, with a separate 90-day window reported for data centres already serving covered customers. Until the Federal Register text settles that, the commencement date is not a figure anyone should be building a model on.

Equally, the attribution of Monday's share-price moves across the optical complex is contested. One account ties the selloff in Jabil, Coherent, Fabrinet, TTM and Lumentum directly to the export rule; another attributes the parallel fall in Coherent to Nvidia's leadership change instead, and the sector has in any case been de-rating for a month on AI-funding nerves rather than on export policy. The rule itself is not in doubt. The claim that it caused Monday's specific 5.2% move is.

What it means for Fabrinet

Fabrinet is a contract manufacturer of optical and electro-optical components, and its production is concentrated in Thailand, around Bangkok. That concentration is the business model: it is how a company assembling precision optics for the largest names in networking has kept its cost base low enough to win the work. The company's own filings already list US, Chinese, Thai and Israeli export controls among its business risks, so this is not a novel category of exposure — but Thailand's arrival on a licence list makes a previously generic risk disclosure specific.

Set against that, the September quarter — Fabrinet's fiscal first quarter of 2027, ended 25 September 2026 — was the strongest in the company's history, and the labels matter:

  • Revenue of approximately USD 1.47bn, up about 51% year on year, above the company's own guidance range.
  • Data-centre revenue up roughly 78% year on year, an increase of about USD 340m on the same quarter a year earlier.
  • Non-GAAP earnings of USD 4.34 a share, against USD 2.92 a year earlier.
  • GAAP earnings of USD 2.86 a share, on GAAP net income of USD 103.5m against USD 95.9m a year earlier.
  • Operating cash flow of roughly USD 21m and negative free cash flow of about USD 137m, on heavy capacity capital expenditure.
  • Second-quarter revenue guidance of USD 1.6bn-1.65bn.

Two of those lines deserve more attention than they usually get. The first is the width of the GAAP/non-GAAP gap: revenue grew 51% and non-GAAP earnings per share grew about 49%, but GAAP net income rose under 8%. Whatever sits in that gap, it is large relative to the quarter, and a reader who picks up the non-GAAP number while believing it is the reported one will carry a materially wrong picture of this business.

The second is free cash flow. Fabrinet burned roughly USD 137m of it in a record quarter, because it is building capacity ahead of demand. That is a defensible choice when the demand is 78% data-centre growth. It is a considerably less comfortable one when the capacity being built sits in a country that moved onto an export-licence list seven weeks later.

The Openbook read

The five-factor profile on the Fabrinet ticker page is unusually lopsided, and Monday's rule sharpens the lopsidedness rather than changing it.

Growth is the strongest score in the set and is not in dispute. Revenue up 51%, data-centre revenue up 78%, and forward guidance of USD 1.6bn-1.65bn that implies sequential growth again. There is no interpretive work required here: this is one of the faster-growing manufacturing businesses on a US listing, and the growth is demand-led rather than acquired.

Profitability is good but less emphatic than the growth, and the basis you choose decides how good. On a non-GAAP basis, earnings per share growth has roughly tracked revenue growth, which for a contract manufacturer is a real achievement — operating leverage in a business model that classically has none. On a reported basis, net income grew a fraction of the revenue line. Both are true. A contract manufacturer's structural position is thin margins and customer concentration, and nothing in this quarter changes that structural fact.

Solvency is where the capacity build shows up. Negative free cash flow of about USD 137m in a single quarter is not a solvency event for a company of this size and cash position, but it is the mechanism by which a growth business becomes a financing question if demand pauses. The relevant risk is not the balance sheet today; it is a half-built Thai capacity programme meeting either a licensing delay or an AI capital-expenditure air pocket.

Momentum is weak and has been for a month. The shares fell 5.2% on Monday to USD 311.01, and that followed a sustained de-rating across AI optics on funding concerns — a stock that reported record revenue in early November has spent the time since falling. Momentum that moves against a record print is the market disagreeing with the print's durability, not disputing its accuracy.

Reward/Risk is the factor that actually moved on Monday, and the direction is the interesting part. Nothing in the rule reduces Fabrinet's expected revenue. What it does is widen the distribution around that revenue, by inserting a licensing authority between Fabrinet's factory and its customers' delivery schedules. The growth has not been devalued; the discount rate applied to it has gone up. On the other side, the de-rating that preceded Monday means a good deal of bad news about AI optics demand is already in the price — so the reward/risk picture is less a deterioration than a genuine widening in both directions, which is exactly what an unresolved regulatory commencement date produces. Readers can compare how that profile sits against the rest of the complex on our stock screener.

Which other listed names sit in the same flow

Geography-based licensing creates a different exposure map than chip-based licensing did, and it sorts the sector by where things are physically made and shipped rather than by what they do.

The most directly exposed are the companies that assemble and ship AI optics and interconnect from the five named jurisdictions. Fabrinet, with its Thai base, is the cleanest case. Coherent, Jabil, TTM Technologies and Applied Optoelectronics all carry manufacturing or assembly exposure in the region and all fell alongside Fabrinet on Monday, though as noted the attribution of those specific moves is contested.

At the demand end sit the chip principals — Nvidia, AMD, Broadcom and Marvell. Their exposure is not to where a box is built but to whether a licence is granted for compute they have already sold into a regional data centre. The 12-million-TPP threshold is a constraint on their customers' business models, which makes it a revenue-timing question rather than a cost question.

One structural change in the optical sector sits across all of this: Marvell agreed on 2 November to combine with Lumentum in an all-stock deal with an implied enterprise value of about USD 115bn, with completion expected in the first half of calendar 2028. That consolidation reshapes the competitive position of every independent optics supplier, Fabrinet included, on a timeline much longer than any export rule's commencement date.

What to watch next

Federal Register publication of the interim final rule. This is the single highest-value item and it should arrive within days. It settles the commencement date, the comment window and — critically for Fabrinet — whether existing arrangements get the reported 90-day runway. Until it publishes, the rule's practical effect on shipping schedules is unknowable.

Fabrinet's fiscal second-quarter results, due late January 2027. Two numbers matter more than revenue: whether the company lands inside its USD 1.6bn-1.65bn guidance, and whether free cash flow turns back toward positive as the capacity spend moderates. Management commentary on Thai export licensing will be the first time the company addresses this rule directly.

Any licence-processing data from BIS. A geography-based regime lives or dies on administrative throughput. A licence requirement that is granted routinely in two weeks is a paperwork cost; the same requirement with a six-month queue is a demand problem for everyone building in Bangkok, Penang or Johor.

The question this rule poses about Fabrinet is not whether AI optical demand is real — the September quarter settles that. It is whether a business whose cost advantage comes from being concentrated in one country can keep being valued as a pure demand story now that the country itself is a variable.

Was this useful?
Sign in to vote
4 of 4
Compare FN, COHR, NVDA, JBL