LITELumentum Holdings Inc.

Investment Thesis

Lumentum occupies one of the tightest supply bottlenecks in the AI infrastructure buildout: the indium phosphide (InP) lasers that drive 800G and 1.6T optical interconnects between GPU clusters. These lasers are the physical layer that converts electrical signals to light and back — and Lumentum's lasers are sold out through 2027, potentially 2029. The CEO has stated the company is "significantly under-shipping demand." When a critical component is on allocation and capacity cannot be added quickly (InP wafer fabs take 2-3 years to build and qualify), the supplier owns pricing power and revenue visibility that few semiconductor companies can match.

The business has transformed under CEO Alan Lowe (formerly of MaxLinear). Revenue has roughly tripled in ~5 quarters as the AI datacom cycle ignited. Q2 calendar 2026 delivered approximately $426M (a beat), and full-year guidance was raised to $1.75-1.8B. Non-GAAP operating margins of 32-35% in strong periods reflect the pricing power of a sold-out supplier. The Q4 FY2026 report (Aug 11) is expected to show $960M-$1.01B in revenue — roughly half the annual total in a single quarter, underscoring the acceleration.

The LiDAR laser business (VCSEL and edge-emitting lasers for Ouster, Luminar, Innoviz, Hesai, and others) is a genuine but secondary thesis component. It is the picks-and-shovels logic in its purest form: Lumentum sells lasers to every LiDAR maker, so LiDAR adoption benefits Lumentum regardless of which architecture wins. But the shovel factory's main customer is AI datacom, not LiDAR — the LiDAR exposure is a volatility kicker, not the core thesis.

Physical AI / Value-Chain Relevance

Lumentum spans two layers of the physical-AI stack: Perception & Sensing (VCSEL and edge-emitting lasers for LiDAR and 3D sensing) and AI/ML Infrastructure (InP-based EML, CW, and coherent lasers for 800G/1.6T transceivers). The AI datacom business is the dominant revenue driver, but both layers are authentic physical-AI exposure.

The LiDAR relevance is real but should be sized correctly. Every LiDAR unit — whether from Ouster, Luminar, Innoviz, or Hesai — contains laser diodes that Lumentum or Coherent likely supplied. As LiDAR volume scales from millions to tens of millions of units, laser content per unit grows (higher channel counts, longer range, better resolution). However, LiDAR laser revenue is dwarfed by AI datacom revenue — the ratio is approximately 10:1. This is not a critique of the LiDAR thesis; it is a calibration. Buy Lumentum for the multi-year sold-out AI laser bottleneck. The LiDAR upside is a free call option.

Catalysts

1. Q4 FY2026 earnings (Aug 11, 2026). The immediate catalyst. Guidance of $960M-$1.01B would represent explosive sequential growth. Any LiDAR design-win commentary, InP capacity expansion updates, or guidance for FY2027 will move the stock.
2. InP capacity expansion reaching production. Lumentum has secured long-term InP wafer supply through AXT (agreement through 2031, $87M prepayment). As new capacity comes online, the sold-out constraint eases — but demand is growing faster than supply, so sold-out conditions may persist even with capacity additions.
3. 800G → 1.6T transition. As NVIDIA shifts from 800G to 1.6T optical interconnects for next-generation GPU clusters, the laser content per transceiver increases. Higher data rates require more sophisticated (and more expensive) lasers — a content-per-unit growth story.
4. LiDAR laser design wins at Western automakers. While secondary to the AI thesis, LiDAR design wins at volume OEMs (Toyota, Hyundai, GM) would validate the sensor-buildout option and could re-rate the LiDAR portion of the business.

Positioning / What the Market May Be Missing

At $779.89, Lumentum trades at a forward P/E of ~44x — reflecting the growth premium but not fully pricing the structural supply-demand imbalance in InP lasers. The stock is down -14.4% in the last week and -16.0% in the last month, caught in the AI infrastructure rotation alongside Coherent and other photonics names.

Three factors the market underappreciates:

1. "Sold out through 2027" is not marketing language. The InP laser supply chain is genuinely capacity-constrained at the substrate level (AXT, Sumitomo) and the device level (Lumentum, Coherent). Adding capacity requires years of fab construction and customer qualification. This is a structural shortage, not a cyclical spike.
2. The AXT agreement through 2031 is a strategic asset. $87M in prepayments locks in InP wafer supply for the better part of a decade. Competitors without similar agreements will be substrate-constrained.
3. LiDAR is a free option on a business that already works. The AI datacom thesis does not require LiDAR adoption to succeed. If LiDAR volumes disappoint, Lumentum's revenue and earnings still grow on AI photonics alone. But if LiDAR volumes surprise to the upside, it is incremental margin on existing manufacturing infrastructure.

Risks and What Invalidates the Thesis

1. AI capex cycle reversal. This is the number-one risk. If hyperscaler spending on GPU clusters slows, the core growth engine stalls. Book-to-bill provides a buffer (backlog converts to revenue even if new orders slow), but the multiple would compress sharply.
2. CPO (co-packaged optics) disruption. If co-packaged optics are deployed broadly before 2028, it could reduce or eliminate demand for the pluggable transceivers that Lumentum's lasers go into. This is a technology-transition risk, not an imminent threat — CPO is still in early adoption and faces thermal, yield, and serviceability challenges.
3. Competition from Coherent and Chinese suppliers. Coherent is the primary competitor in InP lasers, and Chinese photonics suppliers are investing aggressively. Lumentum's moat is manufacturing scale and qualification cycles, not unique technology — if Chinese suppliers close the quality gap, pricing power erodes.
4. Telecom cyclicality. The legacy telecom laser business (~15-20% of revenue) faces demand lumpiness and carrier capex cycles. This is a drag during telecom downturns, even if AI datacom is booming.

What to Watch Next

• Q4 FY2026 earnings (Aug 11). Revenue vs. $960M-$1.01B guidance. Datacom vs. telecom vs. 3D sensing segment splits. LiDAR design-win commentary. FY2027 outlook.
• InP substrate supply chain. Monitor AXT earnings and capacity announcements — Lumentum's growth is gated by InP wafer availability.
• NVIDIA 1.6T optical interconnect roadmap. Any shift in NVIDIA's transceiver architecture (pluggable vs. CPO, silicon photonics vs. InP) directly affects Lumentum's addressable market.
• LiDAR volume data. Ouster, Luminar, and Innoviz quarterly unit shipments — a proxy for Lumentum's LiDAR laser revenue trajectory.