InP supply bottleneck may persist; Citi is bullish on LITE and COHR and raises expectations for LITE catalysts
AI summary card
InP supply bottleneck may persist; Citi is bullish on LITE and COHR and raises expectations for LITE catalysts
Citi believes demand from AI data center 800G/1.6T, CPO, and NPO will continue to drive InP EML and CW lasers, and despite rapid capacity expansion, the supply-demand gap in 2026-2028 will still favor LITE's pricing power and COHR's vertical integration advantages.
- Citi built a new InP laser supply model and expects industry laser chip supply to grow from about 279 million units in 2025 to about 1.3 billion units by 2028E, implying a CAGR of about 66%.
- Even with capacity expansion by AXTI, Sumitomo, JX, COHR, and LITE, the report still expects EML demand to face supply gaps of 30%, 20%, and 11% in 2026-2028, respectively.
- LITE is viewed as a leader in 100G/200G EML and UHP lasers; the report raises FY2027E and FY2028E EPS by about 6% and 5%, respectively, and adds an upside Catalyst Watch.
- Management believes there is no delay in NVIDIA CPO; Lumentum is already shipping to NVIDIA, and laser shipments are expected to inflect in the fourth quarter of 2026 and continue growing in 2027.
- COHR benefits from its 6-inch InP wafer platform, rising internal capacity, and NVIDIA's roughly US$2 billion strategic investment, but still faces risks from customer concentration, the pace of CPO adoption, and competition from silicon photonics/LPO.
Report interpretation
Overview
This report focuses on bottlenecks in InP materials and lasers in AI data center optical communications. Citi believes InP is a critical material for high-speed EML, DFB, and UHP CW lasers used in 800G and 1.6T optical modules as well as CPO/NPO architectures. As GPU cluster bandwidth demand rises, 1.6T modules ramp in 2026-2027, and CPO gradually becomes the next-generation architecture in 2027-2028, supply of InP substrates, wafers, and laser chips will remain tight. The report believes supply shortages will bring better demand visibility and a stronger pricing environment, with LITE and COHR as the core beneficiaries.
Core views
The core views include: first, the InP supply-demand gap is not a short-term disturbance, and EML demand may continue to exceed supply in 2026-2028; second, with leadership in 100G/200G EML and high-power CW laser technology, LITE has room for further price increases and earnings upgrades; third, COHR is another important beneficiary as it lowers costs and expands capacity through 6-inch InP wafers and internal vertical integration; fourth, NVIDIA's CPO roadmap shows no substantive delay, and even if NPO becomes a transitional solution, it still requires similar ELS and laser chips, so it is not negative for LITE; fifth, China's export controls and capacity expansion by domestic suppliers have changed the supply chain structure, but advanced EML and CW lasers above 150mW still depend on a small number of leading vendors.
Analysis framework
The report uses a proprietary InP laser supply model and combines it with Citi's China network team's optical module demand model to assess industry laser chip supply, wafer capacity expansion, EML/CW demand, CPO/NPO roadmaps, and major supplier capacity in 2025-2028E. The analysis also incorporates LITE pre-earnings management communication, public roadmaps from NVIDIA and the supply chain, capacity expansion announcements from supply chain companies, changes in export controls, and revisions to company financial forecasts.
Methodology notes
Estimate the supply-demand gap based on wafers, laser chips, and end-market optical module demand
Citi estimates industry laser chip supply will grow from about 279 million units in 2025 to about 1.3 billion units by 2028E, and matching this against the optical module demand model shows that EML supply will still remain below demand in 2026-2028.
Compare the impact of pluggable optical modules, NPO, and CPO on laser placement, power, serviceability, and the supply chain
CPO improves power efficiency and bandwidth density but sacrifices serviceability; NPO is a practical near- to mid-term transitional solution. Both still rely on InP lasers and external laser sources in high-end scenarios.
Short-term event-driven view
Citi initiates an upside Catalyst Watch on LITE, focusing on company communication around pricing, the CPO roadmap, and technology leadership during the Jun-Q earnings release and the September 2026 ECOC period.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- Lumentum Holdings Inc (LITE.O)Core beneficiary
- Strengths
- Leader in 100G/200G EML and UHP lasers, vertically integrated, with strategic investment and purchase commitments from NVIDIA; both CPO and NPO require its high-end laser capabilities.
- Weaknesses
- The new Greensboro InP device plant is not expected to ramp until mid-2028, so near-term capacity remains tight; customer concentration and cross-border supply chain exposure are relatively high.
- Comparison
- Relative to new Chinese entrants, LITE has technological barriers in 200G EML and CW lasers above 150mW; relative to COHR, the report places greater emphasis on LITE's pricing power and Catalyst Watch.
- Risks
- Delays in CPO adoption, substitution by silicon photonics or LPO/LRO, supplier diversification by NVIDIA, tariffs, and export controls.
- Coherent Corp (COHR.K)Core beneficiary
- Strengths
- Advancing a 6-inch InP wafer platform with about 4x chip output potential and lower cost than traditional 3-inch platforms; has vertically integrated capabilities in InP, silicon photonics, VCSELs, and optical modules.
- Weaknesses
- Highly dependent on a small number of hyperscale customers and NVIDIA-related projects; CPO share depends on platform choices by Broadcom and others.
- Comparison
- Relative to LITE, COHR stands out more for its 6-inch wafer and internal manufacturing advantages; relative to Chinese competitors, its vertical integration and U.S. domestic expansion have strategic value.
- Risks
- Timing of CPO adoption, Broadcom in-house development or alternative supply, competition from silicon photonics/LPO/LRO, customer concentration, and geopolitics.
- AXT / AXTIUpstream InP substrate supplier
- Strengths
- Produces InP, GaAs, and germanium substrates in China through Tongmei, with plans for consecutive capacity expansion in 2026 and 2027 and a high order backlog.
- Weaknesses
- Highly exposed to the pace of Chinese export licensing, with significant revenue volatility.
- Comparison
- Together with Sumitomo and JX, it forms a major source of InP substrate supply, but its China manufacturing footprint brings higher policy risk.
- Risks
- Delays in export licenses, Chinese policy, capacity expansion by domestic competitors, and customers shifting to Japanese or U.S. supply chains.
- Sumitomo Electric IndustriesKey upstream substrate and laser supplier
- Strengths
- The report says it leads in InP wafer and CW-DFB laser diode capacity, with a manufacturing footprint centered in Japan and less impact from Chinese export licensing.
- Weaknesses
- The capacity expansion cycle is relatively long and may require 2.5-3 years to double capacity.
- Comparison
- Lower policy risk than AXT and more vertically integrated than pure substrate manufacturers.
- Risks
- Capacity expansion lagging AI demand, insufficient advanced laser supply, and customers building in-house capacity.
- JX Advanced MetalsUpstream InP substrate supplier
- Strengths
- Has more than 40 years of experience in InP substrate production and plans to invest up to about ¥120B over the next four years to expand InP substrate capacity.
- Weaknesses
- New capacity will mainly start coming online in the second half of fiscal 2027, providing limited relief for short-term shortages.
- Comparison
- Its business base is more diversified than AXT and pure laser manufacturers, with stronger revenue stability.
- Risks
- Capacity expansion execution, returns on capital expenditure, and competition from domestic Chinese suppliers.
- NVIDIADemand and ecosystem driver
- Strengths
- Quantum-X and Spectrum-X Photonics are driving CPO commercialization, and through investments in LITE and COHR, NVIDIA is locking in advanced laser supply.
- Weaknesses
- The supply chain requires coordination among multiple ELS and laser vendors, and CPO is also constrained by TSMC packaging capacity and yields.
- Comparison
- Its roadmap is the key external variable for demand visibility for LITE and COHR.
- Risks
- If NVIDIA adjusts the pace of CPO/NPO or supplier allocation, revenue and valuation expectations for related suppliers may fluctuate.
Key data
- Industry laser chip supplyAbout 279 million units in 2025, about 1.3 billion units in 2028ECiti expects a CAGR of about 66% in 2025-2028E.
- EML demand gap2026E 30%; 2027E 20%; 2028E 11%The demand-versus-supply gap is viewed as the core support for LITE's pricing power.
- LITE FY2027E/FY2028E EPS revisionRaised by about 6% and 5%, respectivelyCiti reflects an additional 15% price increase assumption.
- LITE rating and target priceBuy; target price US$1,100.00; expected return 46.3%The report adds an upside Catalyst Watch on LITE.
- COHR rating and target priceBuy; target price US$420.00; expected return 40.3%The report is positive on COHR's advantages in 6-inch InP wafers and internal capacity.
- Impact of China's InP export controlsInP was added to the export control list on February 4, 2025; lead times for some customers were extended to 2028The report says 6-inch InP wafer prices once rose about 250% to around US$5,000 per wafer.
- NVIDIA strategic investmentLumentum about US$2B; Coherent about US$2BUsed to support advanced laser components, CPO-related capacity, and U.S. manufacturing expansion.
- CPO timelineQualification and risk production in 2026, with rising laser volumes and scaled deployment in 2027-2028The roadmap is being advanced jointly by TSMC COUPE, NVIDIA Quantum-X/Spectrum-X Photonics, and Broadcom Tomahawk platforms.
Impact & implications
For investment, the report defines the InP shortage as a structural bottleneck in the AI infrastructure optical communications chain rather than a single-quarter supply disturbance. LITE's leading EML and UHP CW laser capabilities could drive price increases, EPS upgrades, and event catalysts; COHR's 6-inch wafers and vertical integration help expand internal supply, lower costs, and capture incremental demand from the NVIDIA CPO ecosystem. Upstream AXTI, JX, and Sumitomo benefit from tight substrate capacity, but Chinese export licenses, domestic substitution-driven expansion, and advanced laser technology barriers will determine profit distribution.
Risks
- Actual CPO volume ramp occurs later than expected, or NPO/pluggable solutions extend the transition period.
- Advanced packaging capabilities, yields, or capacity constraints such as TSMC COUPE affect the CPO roadmap.
- Silicon photonics, LPO/LRO, or VCSELs replace EML in some scenarios, weakening InP laser pricing power.
- China's InP export controls, tariffs, and cross-border supply chain policy changes cause supply disruption.
- Capacity expansion by Chinese domestic InP wafer and laser suppliers compresses the long-term share of AXT, Sumitomo, JX, LITE, and COHR.
- Rising concentration of LITE and COHR in NVIDIA and a small number of hyperscale customer projects means changes in order timing could amplify earnings volatility.
- If new capacity outpaces demand or technology roadmaps change, current shortage-driven pricing may prove unsustainable.
What to watch
- LITE's Jun-Q earnings call commentary on pricing, orders, NVIDIA/CPO, NPO, and 200G EML demand.
- Product updates from LITE and COHR on CPO, NPO, and 1.6T/3.2T roadmaps during the September 2026 ECOC period.
- Whether Lumentum laser shipments show the management-expected inflection in the fourth quarter of 2026.
- Actual progress of InP wafer capacity expansion and export license timing at AXTI, Sumitomo, JX, and Chinese domestic manufacturers.
- High-volume production progress of TSMC COUPE in the second half of 2026, and deployment of NVIDIA Spectrum-X Photonics and the Rubin platform.
- Yield in mass production of COHR's 6-inch InP wafers, cost reduction, and changes in the share of internal supply.
- Share changes of EML versus silicon photonics, LPO/LRO, and VCSEL in 1.6T, 3.2T, and scale-up links.