FN · Fabrinet

Current Thesis

Fabrinet is the contract optical manufacturer (CMO) sitting under the AI/optical interconnect buildout, and post-Q3 (reported 2026-05-04) the narrative is broadening from a pure telecom story into a supply-constrained datacom inflection. Q3 FY2026 revenue hit a record $1.214B (+39% YoY, +7% QoQ) with non-GAAP EPS $3.72 (+48% YoY), both above guidance. The acceleration is real YoY growth stepped up from +29% in Q2 to +39% in Q3 and the Q4 guide of $1.25–1.29B implies ~40% YoY at the midpoint. The market sold the May 4 print on datacom supply worries; the stock pulled back ~16% from its $748.89 52-week high and is now consolidating in a tight flag near $627. Accelerating fundamentals digesting into a base, ahead of a datacom unlock that hasn't hit the tape yet that is the leg being bought, not chased.

Bull Case

• Telecom is already ripping and it's NOT just Nvidia. Q3 (2026-05-04) telecom revenue was a record $628M, +55% YoY, led by Data Center Interconnect at $197M, +90% YoY. The DCI line is the AI-datacenter-to-datacenter fiber buildout durable, multi-customer, and accelerating independent of any single GPU cycle.
• Datacom is supply-capped, not demand-capped. Datacom grew just +4% YoY / −6% QoQ in Q3 because of laser, ASIC and memory shortages; management stated shipments ran "well below demand." That is a backlog being deferred, not lost. As components unlock, datacom inflects the asymmetric upside the current base doesn't price.
• Direct hyperscale design wins. Two 800G transceiver programs are now shipping direct to a hyperscale customer (announced on the 2026-05-04 call), ramping in Q4 FY26 and through FY27, full ramp ~mid-FY27. This moves FN up the value chain from merchant-CMO to named-program supplier.
• Capacity is being added ahead of demand. The $150M Building 10 expansion lifts capacity ~50% and is explicitly aimed at 1.6T transceivers plus automotive lidar visible runway into the 1.6T cycle.
• Balance sheet is fortress. $946M cash and short-term investments at quarter-end (2026-03-27), no debt, record GAAP net income $135M. Funds the capex without dilution.

Bear Case

• It's a low-margin CMO, not a chip designer. Q3 non-GAAP gross margin was 12.1% (+10bps YoY) and operating margin 10.7%. This is a volume-and-multiple story; any revenue stumble compresses a thin margin hard, and the multiple has expanded well ahead of the margin.
• Customer concentration. Nvidia and Cisco dominate the book. A single program push-out, in-sourcing decision, or share shift (e.g. toward CPO or linear-drive optics that bypass pluggable modules) is a step-down, not a wobble.
• The market already sold one print on this. FN fell after the 2026-05-04 Q3 results specifically on datacom supply-constraint commentary proof the tape punishes the datacom miss even when the headline beats.
• Stretched on the long frame. At ~$627 the stock sits ~41% above its 200-day MA ($443.80) and has run ~190% over the trailing year. Extension alone isn't a sell, but it raises the bar on the next print.
• Supply unlock is a timing bet. If lasers/ASICs stay tight, the datacom inflection slides into late FY27 and the "demand far exceeds shipments" story gets stale before it converts.

Setup & Price Structure

• Last ~$627.52 (2026-06-05). 52-week range $231.51–$748.89; 200-day MA $443.80; trading ~41% above the 200-day, ~16% below the high.
• Structure is constructive: ran into the 2026-04 high pre-earnings, sold the 2026-05-04 print on datacom worries, and has since contracted into a tight flag/pennant near the top of the range a continuation pattern, not a distribution top.
• This is an ACCELERATING theme in a price-digestion phase. The pivot to watch is a clean break and hold above the consolidation high (toward the $748.89 prior high); failure case is loss of the rising 50-day (~$560) on a weekly close.
• Beginner-trap check: no earnings inside 3 days (Q4 ~mid-August), not at peak-froth retail (news flow is filler "$1,000 invested 15 years ago" pieces, not a WSB squeeze), and the recent ~16% pullback already flushed the weak-handed buyers who chased the April high.

Catalyst Calendar (next 30 days)

• No earnings catalyst in window. Fiscal Q4 FY2026 ended ~2026-06-26; results historically land mid-August (est. ~2026-08-17). That is the next binary, outside the 30-day window.
• Component supply read-throughs (ongoing, June 2026). Laser/ASIC/memory commentary from peers (COHR, LITE) and ASIC suppliers is the real-time tell on whether the datacom unlock is arriving watch peer prints and supplier guidance through June.
• Sell-side resets (rolling). Average 12-month PT clusters ~$659–749 with a high near $850 (June 2026); fresh upgrades/PT raises off the Q3 beat would confirm narrative acceleration. None scheduled, but cluster timing matters.
• No FDA/PDUFA, no scheduled investor day flagged in the next 30 days. Catalyst_date: null.

What Would Change Our Mind

• Price: weekly close below the rising 50-day (~$560) breaks the post-Q3 flag and turns the structure; loss of the 200-day ($443.80) would be a full thesis break.
• Fundamentals: fiscal Q4 (mid-Aug) revenue printing below the $1.25B guide floor, OR datacom revenue declining a second consecutive quarter that flips "supply-constrained, demand intact" into demand erosion.
• Structural: a major customer (Nvidia/Cisco) in-sourcing modules, or a decisive shift to CPO/linear-drive optics that designs out pluggable transceivers, removes the core volume engine.
• Theme: networking-optical flipping to SATURATED peer optical names (COHR, LITE, CIEN) rolling over together on a demand (not supply) narrative would mean the whole cluster has topped.

Correlation Notes

• FN is the manufacturing node of the networking-optical cluster: COHR (Coherent) and LITE (Lumentum) are its laser/component suppliers AND partial customers, CIEN (Ciena) and Corning (GLW) the systems/fiber expression, ALAB/CRDO the electrical-interconnect adjacency. Cluster moves are tightly correlated; a coordinated peer breakdown is a stronger sell signal than any single-name wobble.
• Because FN supplies COHR/LITE-built modules, weak datacom at FN can coincide with strong component bookings at suppliers (the inventory sits upstream) read the cluster as a chain, not a basket.
• High beta to the broad AI-capex tape (NVDA/hyperscaler capex headlines). A genuine AI-spend deceleration hits FN with leverage given its thin margins and concentration.