AI server upgrades and supply-chain bottlenecks are the main theme for Greater China hardware, while consumer electronics face pressure in 2H

Overview

This is a Morgan Stanley investor presentation on technology hardware in Asia-Pacific and Greater China, themed around the building blocks of compute, including servers, AI infrastructure, and electronic components. The report outlines investment themes for 2026 to 2027 from the perspectives of AI GPU and ASIC server rack design upgrades, supply-chain bottlenecks, advanced packaging, PCB/ABF, MLCC, PC and smartphone demand, and regional manufacturing and tariff arrangements.

Core views

The core view is that AI infrastructure remains the strongest structural growth source in the technology hardware supply chain, especially demand for servers, racks, PCB/ABF, thermal solutions, power supplies, interconnects, and MLCC driven by platforms such as GB200/300, VR200, MI400 Helios, TPU, and Trainium. At the same time, consumer electronics may benefit in the short term from tight supply in pricing and margins, but will face pressure from memory price hikes, cost pass-through, and demand destruction in 2H26. The report prefers the Apple supply chain over the Android supply chain and believes suppliers with integration capability, verified shipment records, and capacity at key bottlenecks are more likely to outperform.

Analysis framework

The report uses a combination of supply-chain breakdown, BOM comparison, capacity constraints, rack shipment forecasts, end-demand surveys, and regional trade flow analysis. On AI hardware, it focuses on GPU/ASIC roadmaps, GB200/GB300/VR200 rack ramps, TSMC CoWoS expansion, AI wafer consumption, PCB/ABF, and liquid-cooling suppliers. On consumer electronics, it combines the AlphaWise Smartphone Survey, PC supply-demand dynamics, and the impact of memory prices to assess the relative performance of the Apple and Android chains.

Methodology notes

• Supply chain analysisAI server rack and BOM breakdown

  Identify hardware increments and bottleneck links through the BOM, component upgrades, and supplier division of labor across rack platforms such as GB200, GB300, VR200, and MI400 Helios.

  The report pays particular attention to the rising share of memory in Rubin BOM, tight PCB/ABF supply, liquid-cooling and power-supply upgrades, and market-share changes among ODMs such as Wistron, Wiwynn, and Quanta.

• Capacity constraint analysisCoWoS and advanced packaging capacity tracking

  Measure AI infrastructure deliverability using TSMC CoWoS capacity expansion, AI compute wafer consumption, and packaging material bottlenecks.

  The report mentions that TSMC may expand CoWoS capacity to 165Kwpm by 2027 due to strong AI demand, and discusses the potential benefits, yield, and supply-chain restructuring risks of CoWoP relative to CoWoS.

• End-demand surveyAlphaWise Smartphone Survey

  Assess the relative demand resilience of Apple and Android supply chains through smartphone upgrade rates, replacement motivations, and brand switching rates.

  The survey shows the global smartphone upgrade rate over the next 12 months is expected to rise 5ppt YoY to 43%, but the momentum is mainly driven by iPhone, while Android is more vulnerable to price increases and demand elasticity.

• Market size estimationMLCC TAM and AI server demand breakdown

  Translate incremental demand for MLCC from cloud and edge AI into TAM growth, and distinguish between high-capacitance and low-capacitance MLCC demand.

  The report believes cloud and edge AI will bring about US$1B of incremental TAM to the overall MLCC market, with AI servers as an important driver.

• Regional trade and manufacturing analysisTariff, USMCA, and origin-shift framework

  Assess the geographic risks of the hardware supply chain through U.S. import sources, USMCA compliance, and capacity migration to Southeast Asia/India.

  The report notes that servers and switches manufactured in Mexico are considered USMCA-compliant goods and can enjoy tariff exemptions; Vietnam, Thailand, and India are becoming increasingly important in laptop, Apple device, and smartphone supply chains.

Asset mapping & comparison

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

• AI server ODM and rack supply chain

  Core beneficiary asset chain, covering suppliers such as Wistron, FII/Hon Hai, Wiwynn, and Quanta.

  Strengths

  GPU and ASIC platform upgrades, GB300 deliveries, VR200 ramp, multi-year cooperation with Meta/AMD, and AI data-center investment bring incremental demand.

  Weaknesses

  Shortages of certain components may delay shipment timing; VR200 has already been delayed by 1-2 months due to issues such as Midplane PCB.

  Comparison

  The report notes that Quanta may gain share in CY26 GB racks, Wiwynn is the main ODM partner for Meta MI400-series Helios rack, and Wistron participates in MI400-series Helios rack OAM/UBB.

  Risks

  Supply shortages, delivery delays, geopolitical impacts on AI data-center capex, and customer platform shifts causing share volatility.

• PCB, ABF substrate, and advanced packaging chain

  A key bottleneck in AI server and GPU/ASIC packaging upgrades, covering Gold Circuit, Unimicron, NYPCB, Zhen Ding, and others.

  Strengths

  AI rack upgrades are driving demand for PCB, ABF, and high-density interconnect, while tight supply supports pricing and order visibility.

  Weaknesses

  Advanced process line width/spacing, yield, and material supply constraints may limit volume ramp.

  Comparison

  The report discusses CoWoS versus CoWoP, believing CoWoP can alleviate substrate warpage, improve thermal performance, and enhance NVLink routing, but that it is not reasonable for Rubin Ultra to shift to CoWoP.

  Risks

  If CoWoP is forcefully introduced, it may bring significant yield risks and supply-chain restructuring risks; larger ABF substrate size and layer count may also amplify bottlenecks.

• Power supply, liquid cooling, interconnect, and MLCC

  Supporting beneficiary areas from upgrades in AI server power consumption, thermal management, and high-speed connectivity, involving Delta, Lite-on Tech, AVC, Yageo, and others.

  Strengths

  800 VDC power solutions, liquid-cooling integration, upgrades in data networking/interconnect/CPO, and AI server MLCC demand together drive incrementals.

  Weaknesses

  Customer certification cycles, integration capability, and mass-delivery capability determine supplier differentiation.

  Comparison

  The report emphasizes that suppliers with integration capability and verified shipment records stand out more, while MLCC benefits from the incremental TAM brought by cloud and edge AI.

  Risks

  Rising raw material prices, tight supply, and changes in customer platform timing may affect margins and shipments.

• PC and laptop supply chain

  Supported in the short term by tight supply-demand and rising prices, but facing demand and margin pressure in 2H26.

  Strengths

  Demand exceeded supply in the first half, PC price increases outpaced cost increases, and OEM product-mix optimization supports short-term margins.

  Weaknesses

  Rising memory and key component prices may pressure second-half demand and alter consumer purchasing timing.

  Comparison

  Compared with the AI server chain, the PC chain is more cyclical and driven by cost pass-through logic, with weaker structural growth.

  Risks

  Margin and demand impacts begin to emerge in 2H26, and reduced low-end products may create shipment pressure.

• Apple and iPhone supply chain

  The relatively preferred consumer electronics chain in the report.

  Strengths

  iPhone users have lower price elasticity, brand switching rates are improving, upgrade momentum is mainly driven by iPhone strength, and the product mix is more favorable.

  Weaknesses

  Even Apple needs to bear significantly higher memory costs and may share costs with non-memory suppliers.

  Comparison

  Relative to Android, the Apple supply chain is stronger in demand resilience, user stickiness, and premium model mix.

  Risks

  Changes in iPhone launch timing may lead to lower 2026 shipments; rising memory costs may still erode supply-chain profits.

• Android smartphone supply chain

  The relatively pressured chain in the report.

  Strengths

  Some markets still have upgrade demand, but price elasticity limits room for cost pass-through.

  Weaknesses

  Lower gross margins, a higher share of memory costs in COGS, and greater customer price sensitivity mean price increases are more likely to cause demand destruction.

  Comparison

  Compared with the Apple supply chain, the Android chain is more likely to see shipment declines, de-specification, and margin pressure.

  Risks

  The report forecasts Android smartphone shipments will fall 15% YoY in 2026, with the mid-range Android market hit most clearly.

• Regional manufacturing and tariff-related supply chain

  Supply-chain geographic configuration and tariff exemptions affect hardware companies' costs and delivery security.

  Strengths

  Servers and switches manufactured in Mexico can be regarded as USMCA-compliant goods and enjoy 0% tariffs; Vietnam, Thailand, and India are becoming more important across different hardware categories.

  Weaknesses

  Supply-chain migration takes time, and dependency structures on China, Vietnam, India, and Mexico differ across categories.

  Comparison

  The U.S. still mainly imports smartphones, laptops, and game consoles from China, but Vietnam has become an important export source to the U.S. for iPads, AirPods, and Apple Watch.

  Risks

  Geopolitics, tariff policy changes, and slower-than-expected regional capacity ramps may affect costs and deliveries.

Key data

• Report date2026-06-14The cover shows June 14, 2026 08:05 AM GMT.
• 2025 GB200/300 rack scaleabout 29K racksThe report says 2025 came in at about 29K racks.
• 2026 GB200/300 rack forecast70-80K racksMorgan Stanley forecasts about 70-80K GB200/300 racks in 2026.
• VR200 mass-production timelineDelayed by 1-2 months, ramp starts in 4Q26Reasons for the delay include component issues such as Midplane PCB.
• Memory share in Rubin BOM25%+The report believes recent memory price increases will make memory more than 25% of Rubin BOM.
• Potential TSMC CoWoS capacity165Kwpm in 2027The report title indicates TSMC may expand capacity to this level due to strong AI demand.
• AI compute wafer consumptionCould rise to about US$26B in CY26The report title states AI Computing Wafer Consumption Could Rise to US$26B in CY26.
• Incremental MLCC TAMabout US$1BCloud and edge AI are expected to add about US$1B TAM to the overall MLCC market by around 2027.
• Global smartphone upgrade rate43%, up 5ppt YoYThe 2025 AlphaWise Smartphone Survey shows the global blended upgrade rate reached the highest level in more than 10 years of surveys.
• Change in smartphone upgrade rates in China and the U.S.China +8ppt, U.S. +3pptUpgrade rates over the next 12 months are expected to rise YoY, mainly driven by iPhone strength.
• 2026 Android smartphone shipment forecastDown 15% YoYThe report believes price increases will lead to demand destruction, especially affecting the mid-range Android market.
• 2026 global smartphone shipment forecastDown 13% YoYThe main text indicates global smartphone shipments are expected to decline 13% YoY in 2026.

Impact & implications

The investment implication is that the AI infrastructure chain remains stronger than traditional consumer electronics, and investors should especially focus on ODMs, PCB/ABF, advanced packaging, liquid cooling, power supply, interconnect, and MLCC suppliers with platform validation, mass-production records, and key bottleneck capabilities. Meanwhile, cost pressure in consumer electronics may become increasingly visible in 2H26, with the Apple supply chain showing relatively better demand resilience and product mix, while the Android chain is more vulnerable to rising memory prices, price elasticity, and de-specification effects.

Risks

• Rising memory prices may compress PC and smartphone margins, and lead to end-price increases and demand destruction.
• Rising prices and tight supply of raw materials such as copper and nickel may become headwinds for hardware supply-chain margins.
• AI server component shortages may delay shipment timing; VR200 has already been delayed by 1-2 months versus the original plan due to issues such as Midplane PCB.
• Geopolitical tensions may affect AI data-center infrastructure spending and regional supply-chain arrangements.
• If packaging routes such as CoWoP are introduced too quickly, they may bring yield, line width/spacing, and supply-chain restructuring risks.
• Android smartphone customers have higher price elasticity, and price increases may significantly damage mid-range demand.
• Short-term tightness in the PC market does not mean the full year is risk-free; the report maintains a cautious view and expects the impact to start emerging in 2H26.
• Morgan Stanley discloses investment banking or other service relationships with many covered companies, and investors should use the research conclusions cautiously in light of conflict-of-interest disclosures.

What to watch

• Whether GB300 server racks are delivered smoothly, and whether 2026 GB200/300 racks can reach 70-80K.
• Whether VR200 begins ramping in 4Q26 as expected in the report, and whether component issues such as Midplane PCB ease.
• Whether TSMC CoWoS expansion progresses toward 165Kwpm by 2027, and whether advanced packaging capacity continues to be an AI supply bottleneck.
• Whether memory prices continue rising, and the impact of memory accounting for 25%+ of Rubin BOM on server and consumer electronics margins.
• Share changes for ODMs such as Quanta, Wistron, and Wiwynn in GB racks, MI400 Helios, and Meta/AMD projects.
• The extent of upgrades and volume ramp timing for TPU, Trainium, and other AI ASIC servers.
• Whether cloud and edge AI deliver the roughly US$1B of incremental MLCC TAM.
• Whether smartphone upgrade rates in China and the U.S. remain elevated, and whether iPhone's brand switching rate relative to Android continues to improve.
• Whether Android smartphone shipments in 2026 decline 15% as forecast in the report, and whether mid-range handset price elasticity amplifies demand pressure.
• Capacity ramp and tariff impacts in Mexico, Vietnam, Thailand, and India across server, laptop, Apple device, and smartphone supply chains.