Revenue grew 36%, management pulled a factory in eight months early, and free cash flow was negative $5 million.
Thesis: Fabrinet is a contract manufacturer printing the fastest growth since IPO while capturing none of the scarcity it describes: gross margin held at 12.4% through a documented component shortage, cash conversion was 38% of net income, and FCF was negative on a record quarter. The volume and HPC ramp are real, but the leverage in this value chain accrues to the EML laser vendor and the customer, not the assembler. Holding is correct until either cash conversion or the margin line breaks the tie.
Verdict: HOLD — Conviction: MEDIUM
Catalyst: HPC program crossing the $150M quarterly run rate over the next two quarters as the third and fourth automated lines qualify, plus the EML second source flowing into datacom sequentially positive for a second straight quarter.
Key Risk: Roughly $280M of incremental plant (Building 10 at $130M plus Pinehurst) committed against demand where Grady states plainly they do not ask customers for hard commitments. If hyperscaler CapEx decelerates, utilization drops and the operating leverage that is offsetting FX reverses.
The Tell: Unprompted, Grady admitted the entire capacity thesis rests on non-binding demand: 'we do not ask our customers to make a hard commitment. We have to have capacity in place ahead of demand. That is why we are moving so fast with this.' He volunteered this while defending $130M of capex.
Friction Level: MODERATE_FRICTION — Both sides agree the beat is real and driven by AWS HPC and the EML second source. Disagreement is over whether volume-driven operating leverage at a frozen 12.4% gross margin plus negative FCF is investable, or whether scale with zero pricing power just buys more of the same.
2026Q2 · Feb 2, 2026
3 of 5 readings agreed3 HOLD · 2 LONGRead the full call
Record Revenue, 12.4% Gross Margin, Negative Free Cash Flow
Fabrinet is a contract manufacturer printing the fastest growth since IPO while capturing none of the scarcity it describes: gross margin held at 12.4% through a documented component shortage, cash conversion was 38% of net income, and FCF was negative on a record quarter. The volume and HPC ramp are real, but the leverage in this value chain accrues to the EML laser vendor and the customer, not the assembler. Holding is correct until either cash conversion or the margin line breaks the tie.
Unprompted, Grady admitted the entire capacity thesis rests on non-binding demand: 'we do not ask our customers to make a hard commitment. We have to have capacity in place ahead of demand. That is why we are moving so fast with this.' He volunteered this while defending $130M of capex.
Record $1.13B versus guidance range, non-GAAP EPS $3.36 beating the high end despite a $3M FX loss, and Q3 guided to $1.15-1.20B with EPS $3.45-3.60. Consistent execution above the range, not a one-quarter fluke.
Grady: supply constrained at 200 gig per lane, customer approved a second EML laser source during the quarter, 'should benefit us this quarter and in future quarters.' Datacom revenue turned positive sequentially after a 7% YoY decline.
$130M committed to 2M sq ft Building 10, 250k sq ft pulled in six to eight months early, 120k sq ft Pinehurst conversion underway now. Zero debt, $961M cash, stated 40% ROIC. Management is spending against demand, not talking about it.
Demand outstrips supply on 200G/lane and management says they could ship more with more components, yet gross margin sat at 12.4%, up 10bps sequentially on 36% YoY revenue growth. Assembler passes component cost through; the constraint accrues to the laser vendor.
Grady on Building 10: 'we do not ask our customers to make a hard commitment. We have to have capacity in place ahead of demand.' Capacity north of $2.5B of revenue built against verbal demand.
Datacom revenue $278M, down 7% year over year, while the CEO describes demand as very robust and says supply, not demand, is the limit. Two sequential improvements do not reconcile a 7% annual decline against a scarcity narrative.
CPO (revenue 'relatively small'), OCS ('nothing to announce'), and hyperscaler direct transceivers ('quarters away') are all cited as catalysts with no revenue attached. Three engaged customers, zero material dollars disclosed.
+72
CONFIDENT"Adjective density high (excellent, remarkable, impressive, phenomenal), hedging near zero. Grady volunteered specific numbers unprompted: halfway to $150M, $130M Building 10 capex, 15bps downside, 40% ROIC, 120k sq ft at Pinehurst. Only concessions are autos decline and 20-30bps FX."
Specific segment revenue, ramp percentages, capex, ROIC and capacity math disclosed. Gaps: no DSI/DSO or inventory breakdown available on the call, no hard customer commitments on the capacity build, and revenue quality checked only against this transcript.
HPC program crossing the $150M quarterly run rate over the next two quarters as the third and fourth automated lines qualify, plus the EML second source flowing into datacom sequentially positive for a second straight quarter.
Roughly $280M of incremental plant (Building 10 at $130M plus Pinehurst) committed against demand where Grady states plainly they do not ask customers for hard commitments. If hyperscaler CapEx decelerates, utilization drops and the operating leverage that is offsetting FX reverses.
“Second quarter revenue was $1.13 billion, a new record for the company, which represented growth of 36% from a year ago and is the fastest year-over-year growth we have achieved since our IPO over fifteen years ago.”
“We expect total revenue to be in the range of $1.15 billion and $1.2 billion, representing approximately 35% year-over-year growth at the midpoint.”
Figures are extracted from the transcripts and checked in code: every quote appears word for word. Descriptive only: none of this feeds the verdict.
Record $1.13B, up 36%, fastest growth since IPO, beat and raise with EPS $3.36 above the high end despite 9 cents of FX drag. Operating margin expanded 30bps to 10.9% on flat gross margin, pure volume leverage. HPC went $15M to $86M in one quarter with management a little more than halfway to a $150M run rate over two quarters, and the EML bottleneck at 200G/lane cracked when the customer approved a second source. $961M cash, zero debt, 40% ROIC, $130M Building 10 capex pulled six to eight months early against stated visibility management has never had. Every square foot built at 40% incremental returns with a self-described downside of 15bps of gross margin. Five discrete catalysts within four quarters: HPC ramp, datacom second-source flow, hyperscaler direct transceivers, CPO across three customers, OCS engagement.
Gross margin is 12.4%, up 10bps sequentially on 36% YoY growth. That is zero pricing power despite a documented component shortage. They describe scarcity and cannot raise price. Datacom is $278M, down 7% YoY, against a 'very robust demand' narrative. Net income $122M against operating cash flow $46M is 38% conversion; $52M capex put FCF at negative $5M on a record quarter, cash down $7M, working capital absorbing ~$76M. The $5M buyback at $387 against negative FCF is cosmetic. The $86M HPC line is one AWS program where Fabrinet is second source to an incumbent 'very good supplier with a long relationship,' and the $2.5B capacity build has no hard customer commitment behind it. Q3 guide midpoint implies ~4% sequential growth against 16% delivered. CPO, OCS, and hyperscaler-direct are all 'quarters away' with no revenue to underwrite.
Zero Pricing Power Despite Constraints
Management says demand outstrips supply at 200G/lane and they could ship more with more components, yet gross margin held at 12.4% with revenue up 36% YoY. A genuine scarcity position prices. Fabrinet passes component cost through and keeps 12.4%.
Backlog Quality
Building 10 adds capacity north of $2.5B of revenue and Grady says they do not ask customers to make a hard commitment. Roughly $280M of incremental plant funded against verbal demand while FCF was negative $5M this quarter.
Say/Do Gap
Datacom revenue declined 7% YoY while the CEO describes demand as very robust and states supply, not demand, is the binding constraint. Two sequential gains do not close a 7% annual gap.
Unsustainable Trend Confidence
DCI grew 42% YoY but only 3% sequentially; an analyst flagged it as disappointing and management answered with a promise that DCI will grow faster off that 3% base. Adjective density high, hedging near zero, only two HPC data points.
AI Label Without Substance
CPO revenue is 'relatively small,' OCS has 'nothing to announce,' and hyperscaler transceiver programs are 'quarters away' after eighteen months of work. Named as catalysts without dollars attached.
HOLD at medium conviction. The beat is real and the volume math compounds, but this is a 12.4% gross margin assembler that captured zero pricing out of a documented shortage, converted 38% of net income to operating cash, and finished a record quarter with negative FCF while pulling a factory in eight months early against non-committed demand. Zero debt and $961M cash make it a defensible hold, not a short: the cash conversion gap or the datacom narrative has to break before the shorts have an edge, and neither has broken yet. Add on evidence of cash conversion normalizing or first-source status at AWS. Trim if gross margin breaks below 12.0% or hyperscaler direct programs slip past 'quarters away' again.