Greater China Technology Hardware: AI Becomes the Main Line of Hardware Rerating, with CPO, Servers, and High-End Passive Components Showing Clear Opportunities

Overview

This report is Morgan Stanley’s Greater China Technology Hardware investor presentation in the Asia Summer School series, covering several hardware chains including AI servers, CPO/NPO optical interconnect, AI phones, optical components, server rails, enclosures, power connectors, test equipment, electronic paper, ABF substrates, PCBs, and MLCC. The core view is that around 2026 technology hardware performance will be clearly differentiated, with AI content becoming a key catalyst for valuation rerating: data-center and AI-interconnect related assets have clearer growth visibility, while traditional smartphones and some consumer electronics remain under pressure.

Core views

First, AI servers and interconnect upgrades remain the clearest demand theme, with CPO, optical modules, PCBs, ABF substrates, MLCC, and server rack value per unit increasing. Second, CPO will not replace pluggable solutions immediately; it will coexist in layers with copper cables and pluggable optical modules, first coexist, then expand, then mainstream over time. Third, AI phones offer incremental catalysts; ZTE's AI-native phone with Doubao demonstrates market interest, but whether it can scale remains to be seen. Fourth, some stock-level opportunities come from share gains, product upgrades, and AI deployment tailwinds, including Accton, King Slide, Bizlink, Chenbro, Chroma, E Ink, and Crystal-Optech. Fifth, at the industry rating level the view remains In-Line, suggesting opportunities are more concentrated in structural differentiation rather than a broad sector-wide upgrade.

Analysis framework

The report uses a theme decomposition and bottom-up supply-chain modeling approach: it first starts from AI computing architecture, scale-up/scale-out, and comparisons between CPO and pluggable solutions, then maps through optical modules, PCBs, servers, rails, enclosures, power interconnects, test equipment, substrates, and MLCC; meanwhile it combines company-level revenue, profit, share, product mix, and capex items to identify beneficiaries.

Methodology notes

• Industry chain mappingBottom-up analysis of AI hardware supply chain

  Derive passive component value contribution from GPU racks, interconnect architecture, and BOM changes.

  The report compares NVL72, NVL576, Feynman, and VR200 architectures to analyze GPU count, bandwidth, OE count, rack value per unit, and the supply-chain benefiting segments.

• Technology route comparisonCPO/NPO/pluggable optical module comparison

  Compare CPO, NPO, and pluggable optical module paths on performance, flexibility, power efficiency, and ecosystem maturity trade-offs.

  The report argues that CPO has performance and power-efficiency advantages in scale-up scenarios, but thermal management, yield, maintainability, standards, and cost remain major challenges.

• Relative rating frameworkMorgan Stanley Stock Ratings and Industry Views

  Overweight, Equal-weight, and Underweight correspond to 12-18 month relative risk-adjusted return expectations versus the covered universe.

  The industry view in this report is In-Line, indicating the analyst expects Greater China technology hardware to perform broadly in line with relevant benchmarks over the next 12-18 months.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• AI servers and ODMs

  Directly benefit from higher GB200/300 and VR200 rack shipments and rising value per rack.

  Strengths

  Demand visibility is high, rack value is increasing, and the report notes ODM added value per rack rises by about 38% on VR200.

  Weaknesses

  Highly dependent on Nvidia platform cadence and cloud capex.

  Comparison

  Compared with traditional PCs and smartphones, AI server growth momentum is much stronger.

  Risks

  Platform migration, ramp-up capacity risk, customer concentration, and supply-chain bottlenecks.

• CPO/NPO and optical interconnect

  AI cluster scale-up and scale-out are driving demand for low-power, high-bandwidth interconnect.

  Strengths

  CPO has power and bandwidth advantages; the report expects fast growth in CPO switch shipments.

  Weaknesses

  The ecosystem is still immature, with significant thermal management, yield, and maintainability challenges.

  Comparison

  The report believes CPO, copper cables, and pluggable solutions will coexist in layers rather than one path winning in the short term.

  Risks

  Slow standardization, slower-than-expected cost declines, and insufficient yields.

• AI phones and smart endpoints

  Edge AI may become a new selling point and valuation catalyst for smartphones.

  Strengths

  Nubia M153 selling out on its first day shows market focus on AI-native phones.

  Weaknesses

  Global smartphone shipments remain pressured, with year-on-year declines of 3% in 1Q26 and further downside risk through 2Q-4Q26.

  Comparison

  Compared with AI servers, AI phone commercialization is still in an early validation stage.

  Risks

  Weak consumer willingness to pay, commoditization of AI features, and failure to meet expected shipment scale.

• ABF substrates, PCBs, and MLCC

  AI server BOM upgrades, optical transceivers, and Rubin systems are pushing demand for higher-grade materials and passive components.

  Strengths

  The report expects ABF substrates to become supply-constrained from 2027, with AI application share rising above 75% by 2030; AI servers increase demand for high-capacitance MLCC.

  Weaknesses

  Long lead time for capacity expansion, and some materials and processes are constrained by technical specifications.

  Comparison

  Compared with generic electronics, higher-end AI-related products have greater ASP and gross-margin potential.

  Risks

  Too-rapid supply expansion, delayed customer qualification, and a slowdown in AI capex.

Key data

• Industry viewIn-LineGreater China technology hardware, Asia Pacific industry view.
• CPO switch shipment growth2024-2030e CAGR 144%From the report section title, referring to total CPO switch shipment forecasts.
• 1.6T CPO power savingabout 180MW saved at million-GPU scaleThe report says 1.6T CPO can save about 180MW at million-GPU scale.
• VR200 NVL72 rack costabout US$7.8mnThe report estimates the cost of one VR200 NVL72 rack for hyperscale cloud providers.
• 2026e GB200/300 racks70-80KThe report projects GB200/300 rack counts for 2026e.
• ZTE AI phone first-batch salesabout 30,000 units sold out on day oneThe nubia M153 model co-developed with Doubao AI Assistant.
• AI optical transceiver re-estimate2026E 73mn, 2027E 141mn, 2028E 158mnIncludes 800G, 1.6T and 3.2T.
• AI transceiver PCB TAM growth2025-28E CAGR about 83%The report states AI transceiver PCB TAM is growing rapidly.
• AI-server MLCC shareabout 3% of global units and about 5% of global value in 2027EHigh-capacitance MLCC demand rises as systems such as Nvidia Rubin scale.

Impact & implications

For investors, the report suggests rotating hardware allocation from traditional consumer electronics cycles toward the incremental value chain of AI computing infrastructure. AI servers, CPO, optical modules, PCBs, ABF substrates, MLCC, and test equipment have stronger structural growth, while the core smartphone chain still awaits proof of scalable AI-endpoint commercialization. As the industry rating remains In-Line, stock selection and segment-level analysis are more important than a simple industry beta call.

Risks

• AI server demand or cloud-provider capex comes in below expectations.
• CPO ecosystem, standards, yield, thermal management, and maintainability progress slower than expected.
• Smartphone demand continues to weaken, and AI phones fail to create a scalable replacement cycle.
• Delays in capacity expansion, qualification, or delivery of key components.
• High customer concentration creates order volatility and bargaining risk.
• U.S. executive orders and related regulatory restrictions may affect certain securities transactions or investor suitability.

What to watch

• Actual shipment cadence of 2026 GB200/300 and VR200 racks.
• Commercialization signals for CPO switches moving from coexistence to expansion phase.
• Shipment, ASP, and supplier-share changes for 1.6T and 3.2T optical modules.
• Whether the ABF substrate and T glass supply-demand gap persists through 2028.
• Launch timing, pricing, and sales of ZTE’s second AI phone with Doubao.
• Order execution for MLCC and high-capacitance components driven by AI server demand.
• Subsequent rating, target price, and earnings forecast revisions for Morgan Stanley-covered companies.