Revenue grew 45% to $1.3 billion per quarter and the company generated $4 million in free cash flow for the entire fiscal year.
Thesis: Fabrinet is the bottleneck for complex photonics packaging in the AI datacenter buildout, with 12 consecutive record quarters and diversification away from Nvidia now visible. The problem is the unit economics: 12.2% gross margin down 30bps despite 45% revenue growth, $4M full-year FCF, and a capacity plan doubling to $14B on forecasts management concedes are 'not order commitments.' The growth is real but the business model converts almost nothing to cash, so a hyperscaler capex pause leaves them holding fixed obligations with no cushion.
Verdict: HOLD — Conviction: MEDIUM
Catalyst: Q1 FY2027 new transceiver programs ramping (merchant programs in December quarter, hyperscaler direct this quarter) plus Nokia business expansion beyond 10% could confirm FY2027 acceleration above 36% growth.
Key Risk: Hyperscaler capex slowdown hitting a business with $4M annual FCF and $12.5-14B in planned fixed facility commitments, converting them into a cash-burning capacity builder with no financial cushion.
The Tell: Grady walked through the entire capacity bridge from $5.3B to $14B with specific square footage, dollar-per-building figures, and revenue-per-sq-ft assumptions unprompted when asked about building 11 triggers. He then added 'not beyond the bounds of possibility' for FY2027 growth acceleration above 36%. Volunteering that level of forward specificity beyond a one-quarter guide reveals conviction that exceeds the conservative guidance stance.
Friction Level: MODERATE_FRICTION — Both sides agree revenue is accelerating and cash generation is weak. Disagreement is whether the capex-heavy investment phase resolves into pricing power and free cash flow, or whether 12.2% gross margin with a declining trend is structural for a contract manufacturer with zero pricing leverage.
2026Q4 · Aug 17, 2026
3 of 5 readings agreed3 HOLD · 2 LONGRead the full call
Beat Top of Guide, Full-Year Free Cash Flow Was $4M
Fabrinet is the bottleneck for complex photonics packaging in the AI datacenter buildout, with 12 consecutive record quarters and diversification away from Nvidia now visible. The problem is the unit economics: 12.2% gross margin down 30bps despite 45% revenue growth, $4M full-year FCF, and a capacity plan doubling to $14B on forecasts management concedes are 'not order commitments.' The growth is real but the business model converts almost nothing to cash, so a hyperscaler capex pause leaves them holding fixed obligations with no cushion.
Grady walked through the entire capacity bridge from $5.3B to $14B with specific square footage, dollar-per-building figures, and revenue-per-sq-ft assumptions unprompted when asked about building 11 triggers. He then added 'not beyond the bounds of possibility' for FY2027 growth acceleration above 36%. Volunteering that level of forward specificity beyond a one-quarter guide reveals conviction that exceeds the conservative guidance stance.
Non-GAAP EPS of $14.09 on FY2026 revenue of $4.6B, but full-year FCF was $4M ($257M OCF vs $253M capex). Q4 FCF was negative $37M. Cash declined $70M to $876M and they withdrew meaningful buybacks. Reported profitability and actual cash generation are wildly divergent.
Gross margin of 12.2% down 30bps year-over-year despite record utilization and 45% revenue growth. Management frames rising revenue-per-sq-foot as efficiency, but declining gross margin at peak demand means they cannot raise prices on customers despite capacity constraints.
CFO explicitly stated long-term customer forecasts driving capacity confidence are 'not order commitments.' Management is building to $12.5-14B revenue capacity against $4.6B current revenue on non-binding visibility.
Inventory jumped and analyst flagged it. CFO attributed to revenue growth and 'positioning for future ramps' but declined to disclose days sales in inventory. Inventory is building ahead of revenue that has not yet materialized.
Capacity ceiling raised from $11.5B to $12.5-14B in a single quarter based on forecasts management concedes are not binding. Building two additional 1.2M sq ft factories on non-committal customer visibility is confidence built on unsustainable trends.
12 consecutive record quarters, 6 consecutive accelerating YoY growth, beat above top of guidance range on both revenue and EPS. Q1 guide of $1.375-1.425B at 43% midpoint growth with EPS $4.10-4.25.
Management backing words with specific build schedule: Building 10 complete early 2027 adding 2M sq ft, Navanakorn commissioned, Santa Clara campus acquired, land for two more Chonburi factories. Cash deployed, not just promised.
+75
CONFIDENT"Grady volunteered the full capacity bridge from $5.3B to $14B with specific square footage and dollar figures unprompted, said 'staggering demand picture' and 'not beyond the bounds of possibility' for FY2027 acceleration. Direct answers throughout, no deflection."
Specific capacity math, segment revenue detail, and customer percentages provided. Deflected 1.6T timing to Nvidia and refused DSI disclosure on inventory. Segment reclassification obscures comparability.
Q1 FY2027 new transceiver programs ramping (merchant programs in December quarter, hyperscaler direct this quarter) plus Nokia business expansion beyond 10% could confirm FY2027 acceleration above 36% growth.
Hyperscaler capex slowdown hitting a business with $4M annual FCF and $12.5-14B in planned fixed facility commitments, converting them into a cash-burning capacity builder with no financial cushion.
The verdict is public. The case behind it is Pro: the forensic reading of guidance and language, the bull and bear cases, the red flags, and the play. It opens to everyone on Nov 15, 2026.