Phase Space AI

Financial Model Notes

Lumentum Holdings [LITE]

Lumentum Holdings [LITE] — Financial Model Notes

No .xlsx was produced this run (time-boxed; the brief's priority order puts accounting-quality and valuation work above model artefacts). These notes carry every figure the model would contain with its source.

Per known-silent-failures.md §1, had a workbook been built: no defined name matching ^[A-Za-z]{1,3}\d{1,7}$. This is not hypothetical on this ticker — the record shows a prior LITE memo losing its volatility-derived risk ratio and its conversion trigger because the defined name VOL252 resolved as cell VOL252 and returned 0, silently. Use vol_252d, rev_fy27. And §2: the same prior LITE memo was one of three confirmations that concurrent agents sharing one Excel instance return another workbook's numbers as "verification" — open by absolute path and assert the filename matches the ticker before reading a cell.

Source hierarchy actually used

Verified inputs

Filed annual (10-K FY2025, $m) — including the last segment disclosure

Note the $34.9m Shenzhen facility gain sitting inside FY2025 operating expenses. It flattered the FY2025 GAAP operating margin by 2.1pp and is excluded from every forward figure here. It is a gain, not revenue, so it does not affect any growth rate — but it is exactly the class of item the 10x Genomics precedent (~$94m of settlements inside FY2025 revenue) exists to catch.

Quarterly, FY2025 onward ($m, AV normalized, tied to filings)

TTM: revenue $2,488.4m, gross margin 37.71%, operating income $251.3m (10.10%), D&A $254.5m, EBITDA $505.8m (20.33%). Prior-year TTM: revenue $1,645.0m, operating income −$201.9m (−12.27%). YoY: revenue +69.0%; operating margin +22.4pp.

Sum check: FY2025 quarters = 336.9 + 402.2 + 425.2 + 480.7 = $1,645.0m = filed FY2025 exactly. FY2026 9M = 533.8 + 665.5 + 808.4 = $2,007.7m = the 10-Q. Both windows tie.

Use same-quarter DSO. 51.1d (Q3FY26) vs 54.9d vs 62.2d — falling 11.1 days in two years while revenue grew 140%. The TTM-basis measure reads 66.4d vs 63.4d and is the acceleration artifact documented in the FN file.

Balance sheet and capital structure, 2026-03-28 ($m)

Convert detail — model each series separately:

Less cash-settled principal equivalent (3,198.4 ÷ 602.23) = 5.311m → net settlement 21.40m shares.

$102.0m of 2032 Capped Call Options was purchased and economically offsets 2032 dilution up to an unextracted cap. Capped calls do not enter GAAP diluted EPS. Range: fully diluted 95.4m–102.1m; EV $57.6–61.6bn. The 102.1m figure is used because it ties to the company's guided count.

Forward, from guidance (8-K and call, 2026-05-05)

Model structure that would be built

1. Revenue by product category, not by segment. Segments were retired in Q1 FY2026; Components and Systems are what LITE now discloses. Build Systems from cloud transceivers + OCS + CPO (all three quantified or named), and Components from merchant lasers (200G EML) + the declining consumer 3D-sensing and industrial laser lines.
2. Model the declining business explicitly even though it is no longer disclosed. Industrial Tech was $234.2m and falling 14.6% in FY2025 with segment profit of $12.1m. Carry it forward as a separate declining line at the last disclosed level and state that the assumption is unverifiable. Setting it to zero or folding it into growth would hide the censored metric inside the model.
3. Gross margin to 47.0% terminal, held at the Q3FY26 non-GAAP actual, not expanded. Note LITE's own disclosure that cloud transceiver growth is "partially offset by lower average selling prices" — model volume and ASP separately, because that is the bear case's mechanism.
4. Opex: R&D from 18.5% of revenue to 10.0% (doubling absolute spend); SG&A from 21.2% to 8.0%. Acquired-intangible amortisation on the disclosed schedule: $33.5m (rest of FY26), $123.6m, $83.0m, $52.6m, $46.5m, $21.2m — to effectively nil by FY2032.
5. Capital structure — the part most models get wrong.
6. Model each of the four note series separately with its own conversion price. Do not sum shortTermDebt and currentLongTermDebt — AV populates both for the same borrowing.
7. Model the diluted share count as a function of the share price, not a constant. It is reflexive: principal is cash-settled, excess is share-settled, so the count rises with the price. It was guided at 83.5m in November 2025 and ~102m in May 2026.
8. Model the $2.0bn NVIDIA preferred as 2.9m common-equivalent shares (1:1), participating in dividends.
9. Iterate any price target against the count it implies (see the valuation file — $839 naive, $827 converged).
10. Terminal margin capped at terminal gross margin as a hard model assertion that raises, not flags.

Defects to carry forward