Phase Space AI

Valuation

Fabrinet [FN]

Fabrinet [FN] — Valuation

Spot $414.39 (2026-07-29). Two outputs, as required: a 12-month target and the 5-year implied-path test.

---

1. COMPANY STATE — declared first

STATE A — mature and structurally stable

FN is the cleanest State-A case in this cluster, and the only one of the three names that qualifies.

Test, applied:

Not B: FN's revenue is cyclical through its customers, but its own margin is not cycle-driven — it held 9.5% through the FY2023–24 telecom inventory correction, which is exactly what a State-B business cannot do. Not C: it is not thin-margin-and-scaling; the margin is stable and the value is not concentrated in the terminal period. Not D: the reverse DCF solves across the whole range.

Consequence of State A, per valuation.md

• Terminal margin built from the company's own normalised economics plus an explicit forward bridge. Industry data is a sanity band, never an override — do not clamp to a percentile. This is precisely the rule the screen broke.
• Instrument: fix the underwritten margin, solve for price-implied growth, and show the neighbouring sensitivity surface.

evidence_grade: A. Filings current; av_vs_edgar VERIFIED with 6 periods compared and 0 disagree; customers named with a matching receivables table; guidance quantified and dated. The only material gap is consensus (unavailable — see §5) and the structurally undisclosed data-rate mix.

---

2. TERMINAL MARGIN — 9.8%, and the ceiling that makes it non-negotiable

Underwritten terminal EBIT margin: 9.8%.

The hard constraint comes first

m_EBIT,T  ≤  m_gross,T          →     9.8%  ≤  12.0%   ✓

FN's gross margin is 11.96% TTM and falling (12.71% FY2023 → 12.35% FY2024 → 12.09% FY2025 → 11.96% TTM). This is not a soft prior; it is an arithmetic ceiling, and it is only 2.2pp above the underwritten figure. There is essentially no room in this model for a margin surprise to the upside.

The opex bridge

The terminal margin sits 0.02pp above the TTM actual of 9.78% and inside the sixteen-quarter observed range of 9.0–10.1%. It is not a forecast so much as a statement that FN's margin does not move. That is the correct treatment for State A, and it is the opposite of what the screen did.

What the screen did

terminal_margin: 0.144, terminal_margin_basis: "max(own, sector peer median)".

14.4% against a 12.09% gross margin — a breach of the hard constraint by 2.31pp. Per valuation.md this fails the model; it does not merely flag. FN cannot earn 14.4% at the operating line without a negative total operating expense.

Independently confirmed mid-run by a separate audit: the same 14.4% was applied to 70 of 111 names while labelled sector-derived, and FN is one of seven names where it exceeds the company's own gross margin. This is the sixth recorded failure of this parameter (known-silent-failures.md §9.8) and it is the banned max(own, industry_median) construction working exactly as predicted: it can only ever flatter a low-margin business, and FN — at a 12% gross margin against peers earning 3–6x that — is the maximally exposed case.

Direction of the correction, stated so it cannot be read as motivated: cutting the terminal margin from 14.4% to 9.8% raises the required growth rate and moves FN from the screen's PASS toward FAIL. The correction is adverse to the name.

---

3. THE EXIT MULTIPLE — derived, then sanity-checked against a real EMS set

3.1 Derived from the identity

EV_T / EBIT_T  =  (1 − t)(1 − g/ROIC) / (WACC − g)

FN's effective tax rate is the load-bearing input here and it is unusually low: ~6.5% (FY2025 income tax $22.7m on pretax $355.2m), a durable consequence of Thai investment incentives. A low tax rate raises the warranted multiple mechanically.

Base exit multiple adopted: 14.0x.

3.2 Reference-class assertion, then the sanity band

valuation.md rule 6 requires n≥5 mature profitable firms, a matching operating model rather than a matching SIC code, and dispersion in the matched dimension.

The matched operating model is electronics manufacturing services: build-to-customer-specification, pass-through material content, gross margin under 13%, no R&D line.

n = 4 excluding FN, against a requirement of 5. The class is therefore marginally under-populated, and this is declared rather than papered over. It does have genuine dispersion in both dimensions being matched — growth spans 8.1% to 58.6%, margin 4.3% to 8.1% — so it is not the §9.9 failure (a precisely identified, uninformative anchor). It is a small but informative class.

FN's growth of +29.8% is bracketed by FLEX/JBL below and CLS/SANM above. The growth-matching requirement is satisfied.

The base 14.0x sits inside the anchor range (11.4x–30.3x), so valuation.md's prohibition on a base below every stated anchor is not triggered. It does sit below the median of 25.2x, and the reason is stated: the median reflects a market capitalising AI-EMS growth well above the 5% terminal rate the identity assumes. Both numbers are carried — 14.0x as the base, 25.2x as the upper sensitivity bound.

3.3 The second instrument — EV/gross profit, and what it says

The coordinator's instruction was that if an EV/EBIT exit multiple cannot be made to fit, say so and value the name another way. It does fit — FN's 9.8% operating margin makes EBIT meaningful — but a second instrument is produced anyway, because EV/Sales is genuinely contaminated for a contract manufacturer.

A CM's revenue includes customer-specified material bought at cost and passed through. FN's $4,235m of revenue contains roughly $3,728m of cost of revenues, much of it material FN neither designs nor marks up. So EV/Sales measures material content as much as it measures business value, and EV/gross profit is the correct normalisation.

On EV/gross profit, FN trades at 27.8x against an EMS median of 12.9x — a 2.14x premium.

Is the premium warranted? Partly, and the parts are nameable: - Operating margin 9.8% vs a 5.1% median — +1.9x, which alone explains most of it. - Effective tax ~6.5% vs peers paying full US or blended rates. - Net cash $944m and zero debt, against peers carrying leverage. - Growth +29.8% against a +32.7% median — no premium justified on growth; FN grows slower than the median of its own peer group.

Conclusion from the second instrument: FN is expensive within its own class, and the premium is justified by profitability and balance sheet but not by growth. That is a material qualification to both the screen's +2.5pp PASS and to this memo's own marginal result below, and it is the reason FN's verdict is stated as conditional rather than clean.

---

4. IMPLIED-PATH TEST — the Valuation Criteria

Terminal value exceeds 60% of EV, so the reverse DCF is the mandatory primary long-horizon output.

reverse_dcf.py --spot 414.39 --shares 36.217 --net-cash 943.8 --revenue 4235.0
               --years 5 --wacc 0.10 --terminal-margin 0.098 --solve cagr

Held fixed and named: terminal EBIT margin 9.8%; WACC 10.0%; horizon 5 years; TTM revenue $4,235.0m; shares 36.217m; net cash +$943.8m (verified against the 10-K: total debt $1.5m, interest expense zero); EV $14,064m.

The required parameter

Flip point: exit multiple 14.8x.

The margin the strategy ranks on

Demonstrated − required = 29.8% − 31.27% = −1.47pp.

Both growth windows are reported, as required:

The 3-year window spans the FY2023–24 telecom inventory correction, during which FN's revenue grew only 9.0% cumulatively. TTM YoY is used as the primary measure because the current regime — NVIDIA at 27.6% of revenue, DCI at +90%, FY2026 guided +34% against FY2025's +19% — is a different regime, and the 3-year figure is disclosed rather than discarded.

The implied compression, as a number

Current EV/EBIT 34.0x → base exit 14.0x: a compression of 20.0 turns, 58.8%. On EV/gross profit, 27.8x → an implied ~11.4x, −59%. FN must compound revenue at 31.3% into a multiple 59% below today's for the current price to be right.

Cross-check: solve for terminal margin instead

At the demonstrated 29.8% CAGR and a 14.0x exit, the required terminal margin is 10.37% against the underwritten 9.8% — a shortfall of −0.57pp. This is the tightest and most informative number in the FN file. It is not arithmetically impossible: 10.37% sits below the 12.0% gross-margin ceiling. But it requires FN to convert 86.4% of gross profit into operating profit, against 81.8% today — which means holding SG&A to roughly 1.6% of revenue. FN's SG&A has fallen 2.94% → 2.56% over three years; another 0.96pp of compression is a stretch, not an impossibility.

So the price requires FN to be modestly better than it has ever been on the one dimension where it has demonstrated near-perfect consistency for three years.

Verdict: PASS WITH ARGUMENT

The price requires 31.27% against 29.8% demonstrated — a shortfall of only 1.47pp, well inside the noise of a single quarter's revenue. valuation.md permits PASS WITH ARGUMENT where the price requires more than demonstrated and there is a specific, evidenced reason. The argument here is specific, quantified and filed:

1. Q4FY26 guided +40% YoY at the midpoint ($1.25–1.29bn) — above the required 31.27%.
2. FY2026 tracking +34% against FY2025's +19% — growth is accelerating, stated by management.
3. DCI modules at $197m, +90% YoY and +38% sequentially — the fastest line in the company, disclosed with a number.
4. A high-performance-compute programme approaching $150m per quarter, one quarter from that milestone, and "more than one family of products" — a dated, quantified new revenue vector.
5. The constraint is supply, not demand: "component availability that is not keeping pace with strong demand." A backlog-constrained order book is a materially better forward indicator than a demand-constrained one.
6. The exit multiple has to fall to 14.8x for the test to fail, and every EMS comparator trades above it — 22.5x, 27.9x, 30.3x, with only SANM at 11.4x below.

Against the argument, and recorded rather than buried: FN has no margin runway (§2 — the gross margin is falling and SG&A is already 2.56%), so every dollar of the required path must come from revenue; the required 31.27% implies FY2031 revenue of $16.5bn, roughly 4x today and larger than Celestica is now; and the second instrument (§3.3) says FN already trades at 2.14x its peer group on the one multiple that strips pass-through material.

PASS WITH ARGUMENT, at −1.47pp, with the exit multiple named as decisive and the flip point at 14.8x. This is the closest-to-the-line result of the three names and it should be read as "approximately fairly priced," not as an endorsement.

Sensitivity — on the exit multiple, never on scenario probabilities

The table above is the sensitivity. A 1.0x change in the exit multiple moves the required CAGR by ~1.5pp. A ±1pp change in the terminal margin moves it by ~3.0pp — but the terminal margin is boxed between the 12.0% gross ceiling above and three years of 9.5% actuals below, so its true range is ±0.5pp. The exit multiple carries essentially all the variance.

---

5. TWELVE-MONTH TARGET

Target $561. +35.3% to spot $414.39.

valuation.md step 1 requires starting from near-term consensus. AV EARNINGS_ESTIMATES returns an empty array for FN (and for COHR and LITE) while returning 22–41 records for CRDO, CIEN and AAOI on the same premium key in the same minute. The substitution is declared: filed company guidance replaces consensus. FN's guidance is quantified and dated, which makes it a legitimate substitute.

Build

Multiple percentile is stated as required, and it is honest in FN's case: FN's share count has been essentially flat (36.2m, with a $169m buyback authorisation only partly used and "no meaningful number of shares" repurchased in Q3), so the price/revenue metric is a clean proxy for EV/Sales across the window. That is not true of COHR (+26% shares) or LITE (+46% diluted), where the percentile is understated.

Named product-cycle events inside 12 months

Dated in FN_Catalyst_Calendar.md: Q4/FY2026 results and the first FY2027 guide (~mid-August 2026); the FY2026 10-K with the updated named-customer and receivables tables; the $150m/quarter HPC milestone (management: "probably a quarter away"); and the new customer programmes ramping in FY2027.

Caveat

The target requires the multiple to hold near the 78th–85th percentile of its own five-year range. At the median it roughly halves. And FN's target has a specific asymmetry the other two do not: because FN's margin cannot expand, the entire target is a revenue-and-multiple statement. There is no earnings mix-shift to fall back on if revenue disappoints.

No external professional target was available for the sanity band. Recorded as absent.

---

6. Liquidity Criteria and vehicle — PASS, equity by necessity as well as by default

Vehicle: EQUITY. Ladder rung 1.

FN's options chain contains no contracts within ±10% of spot at any expiration 350 days or more out. This is not a thin chain — it is an absent one. FN therefore fails the LEAP test at the first gate: a vehicle that does not exist cannot be constructed, and no IV−RV spread can be measured because there is no ATM implied volatility to measure.

Per the vehicle ladder, absent an argument the vehicle is equity, and here there is not even a candidate. Equity, unconditionally. Recorded as UNIDENTIFIED rather than as a zero or a peer-implied estimate — a missing input is INDETERMINATE, never a substituted number.

---

7. MEASURED Criteria

---

8. Summary card