Lumentum Is Sold Out Through 2029. That Makes 1.6T Optical, Not Silicon, the AI Bottleneck

The four biggest U.S. hyperscalers — Microsoft, Amazon, Alphabet and Meta Platforms — collectively plan to deploy about $725 billion of capital expenditure in 2026, a 77% jump from the $410 billion record set in 2025 (Yahoo Finance / industry coverage, Jun 3, 2026; Tom's Hardware, Apr 30, 2026). That capital flows first into GPU racks and then, with brutal arithmetic, into the optical cabling that lets those racks act as one machine.

The implication is mechanical: every additional rack of AI compute that hyperscalers stand up needs a roughly fixed ratio of 1.6T optical links, and every 1.6T link needs multiple 200G EML lasers. If Lumentum is sold out and Coherent's vertical integration only partially offsets the gap, hyperscaler capex dollars can no longer buy proportional compute. The bottleneck has moved upstream of the GPU.

Lumentum's fiscal fourth quarter, reported August 11, 2026, closed with $1.006 billion in revenue — up 109% year-on-year, the strongest quarterly growth print in the company's history. Gross margin reached 47.4% on a GAAP basis and 50.4% non-GAAP. Management guided Q1 FY27 to $1.225B–$1.275B at the midpoint ($1.25B), implying non-GAAP operating margin of 39.5–40.5% and non-GAAP diluted EPS of $4.05–$4.35 (Lumentum Q4 FY26 release, Aug 11, 2026).

Coherent is the closest substitute. Its TTM revenue of $7.12B is more than double Lumentum's, but its gross margin (~37.5% TTM) is structurally lower because it spans industrial lasers, materials and components, not just AI-grade optics. Coherent reported $2.046B of Q4 FY26 revenue (Coherent Q4 FY26 financials, Aug 14, 2026 filing), with capex-to-revenue at 27.5% — heavy investment in 6-inch InP capacity intended to take share from Lumentum. But Coherent and Lumentum together may not be able to fully offset Chinese supply for years (Reuters / Benzinga, Oct 1, 2026).

Zhongji Innolight is the volume engine of the 800G and 1.6T module market. First-half 2026 revenue of RMB 41.78 billion (+182% YoY) and net profit of RMB 13.65 billion put it on track for a full-year revenue near RMB 90B;2027 orders for 1.6T and 800G are already on the books, and the company has earmarked an HK$53.41 billion July 2026 Hong Kong listing to fund capacity and materials (simpletechtrend, Aug 23, 2026 earnings-call notes). The structural risk: a bipartisan U.S. legislative push introduced October 1, 2026 would prohibit federal use of Chinese-made optical transceivers (Eoptolink, Zhongji Innolight), accelerating the supply shift toward U.S. suppliers but capping Chinese-module access to U.S. hyperscalers.

On September 28, 2026, Citi projected that the optical circuit switching (OCS) market will grow to $11 billion by 2030 — roughly a 38× expansion from the ~$300M base of 2025 — driven by hyperscaler demand for high-radix, low-power fabric (Investor's Business Daily, Sep 28, 2026; 247 Wall St, Sep 28, 2026). Even after that call, optics stocks dipped as the market focused on near-term valuations rather than the long-horizon TAM. The thesis: every OCS port still needs EML lasers, and Lumentum's 50–60% share of 200G EML chips means the company becomes the de-facto toll taker on the next fabric build-out.

Near-term (days to quarters): Lumentum's October 9 disclosure reframes the equity narrative from \"AI optics play\" to \"AI optics monopoly with three-year visibility.\" With Q1 FY27 guidance implying $1.25B of revenue at 39.5–40.5% non-GAAP operating margin, the stock now trades at roughly 33× forward sales and 43× forward earnings — expensive on absolute multiples but consistent with a sold-out franchise. Watch the next Lumentum earnings call for incremental color on (i) the 200G EML production ramp at the Greensboro fab, (ii) the 30%/70% shortfall product mix, and (iii) any change to the 2027–2028 order book.

Long-term (1–3 years): The structural picture turns on whether alternative laser architectures (CW VCSELs, silicon photonics with external lasers, thin-film lithium niobate) can displace InP EMLs in 1.6T modules before Lumentum's new fabs come online. If they can, Lumentum's share can compress; if they can't, the 50–60% EML share persists into the 2.4T/NPO/XPO generation Innolight plans for 2H 2027 (simpletechtrend, Aug 23, 2026). The regulatory track is the second variable: an FCC listing of Chinese optical transceivers accelerates U.S. share gains for Lumentum, Coherent, Applied Optoelectronics and Corning, but raises cost-of-goods for hyperscalers that had been buying Chinese modules at scale.