Skip to content
primarysourced Photonics sector Lumentum
LITE
~4 min read · 945 words ·updated 2026-04-29 · confidence 44%

Capex cycle — InP fab expansion

Lumentum’s capex envelope is undergoing a step-function increase in FY2026–FY2028 driven by the NVDA-funded San Jose InP fab expansion announced concurrently with the March 2026 $2B Series A Convertible Preferred placement. This is the most material capex inflection in Lumentum’s post-Oclaro history and re-rates the long-term operating leverage profile of the business.

Pre-AI-cycle capex baseline

Lumentum’s historical capex run-rate (FY2021–FY2024) was approximately $80–150M per year — modest relative to a $1.4–1.8B revenue base. This included:

  • Maintenance capex for San Jose InP fab and Towcester UK fab
  • Modest expansion capex for VCSEL line and ROADM/WSS subsystem assembly
  • Cloud Light integration capex post-November 2023 acquisition

The FY2024 trough environment kept capex constrained; the FY2025 recovery began to lift capex back to roughly $150–200M annual run-rate as Lumentum prepared for the 200G/lane EML ramp.

FY2026–FY2027 capex inflection

Three discrete capacity programs compound:

1. San Jose new fab (NVDA-funded)

Announced March 2026 as part of the NVIDIA strategic partnership. The new fab is a US-based InP fab dedicated to 200G/lane EML and follow-on advanced laser components for NVDA AI infrastructure. Key parameters:

ParameterStatus
LocationSan Jose, CA (adjacent to existing fab)
FundingNVDA $2B Series A Convertible Preferred + LITE corporate cash + likely CHIPS Act ITC offsets
Approximate capex envelope⚠ Not publicly quantified; industry estimates suggest $1.5–2.5B+ over multi-year build
Construction start2026
First-volume targetLikely 2027–2028 ⚠
Capacity dedicationSignificant share dedicated to NVDA per partnership terms

The fab is dimensioned for the long-cycle AI capex demand. NVDA’s $2B equity injection is dimensioned to be a meaningful but not full coverage of the build cost; the residual is funded by Lumentum operating cash flow plus existing balance-sheet cash.

2. Towcester UK InP fab expansion

The legacy Oclaro fab in Towcester continues to be expanded for the 200G/lane EML node. This is incremental brownfield expansion within the existing UK facility:

  • Equipment upgrades (epitaxy reactors, lithography, MOCVD capacity)
  • Yield-improvement capex
  • Capacity expansion for non-NVDA-allocated 200G/lane EML demand

UK capex is funded from existing cash; not directly tied to the NVDA arrangement. Likely several hundred million dollars cumulative over FY2026–FY2027.

3. Cloud Light Thailand transceiver-assembly expansion

Hyperscaler-direct module assembly capacity expansion in Thailand. Less capex-intensive than InP fab capex (no clean-room semi-equipment), more labor-cost-driven. Cumulative FY2026–FY2027 expansion likely in the $100–200M range — supporting 1.6T-module ramp and selective CPO assembly capability.

Implied capex schedule

⚠ Inferred / management-framed estimates only; not formally disclosed:

Fiscal yearTotal capex ($M est.)NVDA-funded portion ($M est.)Notes
FY2024 actual~$100Mn/aPre-AI cycle
FY2025 actual~$200Mn/aRecovery; preparatory
FY2026 estimate$400–600M$0–500M (ramping in H2 FY2026)NVDA placement March 2026; capital deployment begins
FY2027 estimate$700–1,000M$500–1,000MPeak capex year
FY2028 estimate$500–700Mtail-end NVDA fund deploymentVolume capacity online

Implication: cumulative FY2026–FY2028 capex of $1.6–2.3B, of which a meaningful share is NVDA-funded. Lumentum’s free-cash-flow model needs to absorb this. The Q2 FY2026 operating-cash-flow profile (operating margin 25.2% non-GAAP × $665.5M revenue ≈ $168M operating income; cash-flow conversion lagged by working-capital build) suggests Lumentum can generate $150–300M+ of operating cash flow per quarter, which combined with the $3B+ post-NVDA cash position is sufficient to fund the capex envelope without external debt.

Capacity output milestones

Management-framed capacity-output milestones (CEO Alan Lowe — March 2026 OFC commentary):

MilestoneTarget window
EML chip volume “more than double” CY2024 → CY2025Achieved (referenced on multiple FY2025 calls)
1.6T optical chipset rapid commercializationCY2026
$1.25B quarterly revenue run rate (capacity-supported)9–12 months from March 2026
$2.0B quarterly revenue run rate18–24 months from March 2026
Order book filled through 2028As stated April 10, 2026

The path from current ~$700M quarterly run rate to $2B quarterly run rate is roughly a 3x throughput expansion in 18–24 months — implying both capacity additions and significant ASP / mix uplift contribute. Capex efficiency (revenue-per-capex-dollar) should be strong given the InP fab fixed-cost base and the high-ASP 200G/lane EML chip economics.

Capex risks

  1. Build-execution risk — InP fab construction and qualification timelines are non-trivial. Equipment delivery delays, environmental permitting, hiring of qualified fab engineers, and first-silicon-out timing all introduce slip risk.
  2. Demand-mismatch risk — if AI capex pauses before the FY2027–FY2028 capacity comes online, Lumentum has stranded capex with under-utilized fixed-cost absorption. The mitigation is the multibillion-dollar NVDA purchase commitment which creates a contractual demand floor.
  3. Capex-funding risk — minimal. Cash + NVDA proceeds + operating cash flow comfortably covers the envelope; no debt-issuance dependency exists for the capex plan.
  4. Technology-pivot risk — if the industry-architectural transition to CPO accelerates faster than expected, the pluggable-module-related capex (Thailand assembly) faces stranded-investment risk. The InP-fab capex is technology-agnostic between pluggable and CPO.

Sources