AI compute hardware remains the main theme in Greater China technology hardware, with servers, packaging, and electronic components benefiting while consumer electronics face pressure in the second half
AI summary card
AI compute hardware remains the main theme in Greater China technology hardware, with servers, packaging, and electronic components benefiting while consumer electronics face pressure in the second half
Morgan Stanley believes that in 2026, AI server racks, GPU/ASIC platform upgrades, and the CoWoS/ABF/PCB and liquid-cooling power supply chain still have upside opportunities, but rising memory and raw material prices will weigh on demand and margins for PCs and Android smartphones.
- AI server opportunities are concentrated in Vera Rubin, the Kyber architecture, LPX/Vera CPU racks, and upgrades to ASIC servers such as TPU and Trainium.
- The report forecasts 2026 GB200/GB300 racks at about 70-80K, significantly above about 29K in 2025; VR200 is delayed by 1-2 months due to component issues and is expected to begin ramping in 4Q26.
- Supply of advanced packaging and key components remains the core bottleneck, with TSMC CoWoS capacity expected to expand to 165Kwpm by end-2027, while AI compute wafer consumption could rise to US$26B in CY26.
- Risks are rising on the consumer electronics side: higher memory prices, increases in raw material prices such as copper and nickel, and tight supply may pressure demand and margins for PCs and smartphones in 2H26.
- The report is relatively more positive on Apple and its supply chain, believing the Android smartphone chain is more vulnerable to price increases, spec downgrades, and demand destruction.
Report interpretation
Overview
This report focuses on the building blocks of compute within Greater China technology hardware, covering servers, AI infrastructure, advanced packaging, electronic components, and the PC and smartphone supply chain. The core conclusion is that AI data center demand continues to drive expansion in GPU/ASIC servers, rack design, advanced packaging, PCB/ABF substrates, liquid cooling, power supply, and MLCC, while rising memory and raw material prices will put demand and margin pressure on the consumer electronics chain in 2H26.
Core views
The report’s main theme is a selective overweight on the AI hardware supply chain while remaining cautious on consumer electronics. On the AI server side, smooth GB300 deliveries, changes in GB rack share, rising ASIC server volumes, 800VDC power supply, tight PCB/ABF supply, interconnect/CPO upgrades, and liquid-cooling solutions are near-term catalysts for share-price upside. On the consumer electronics side, PCs may see margins supported in the short term by tight supply-demand conditions and a better product mix, but demand and cost pressures will emerge in 2H26; in smartphones, upgrade-rate surveys show stronger momentum for iPhone than Android, and the Android supply chain is more likely to face shipment declines, spec downgrades, and gross-margin pressure.
Analysis framework
The report uses platform roadmaps, rack shipment forecasts, BOM changes, advanced packaging capacity, supply chain division of labor, MLCC TAM estimates, the AlphaWise smartphone survey, and regional manufacturing migration analysis to assess the relative opportunities and risks in AI hardware and consumer electronics supply chains.
Methodology notes
GB200/GB300/VR200 racks, Rubin/Vera platform upgrades
By using GPU and ASIC server roadmaps, rack output, BOM changes, and ODM division of labor, the report assesses revenue elasticity and bottlenecks in the server hardware chain.
CoWoS capacity expansion, CoWoP feasibility, ABF and PCB constraints
The report compares the capacity, materials, line width/spacing, and yield risks of advanced packaging solutions, concluding that adopting CoWoP for Rubin Ultra in the short term is not reasonable.
Smartphone upgrade rates and brand switching rates
By examining upgrade intentions and net switching rates over the next 12 months across regions such as China, the United States, and EMEA, the report judges the demand divergence between iPhone and Android.
Incremental MLCC demand driven by cloud and edge AI
The report estimates that AI servers and edge AI will increase MLCC demand, creating an incremental market opportunity of about US$1B.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- AI server hardware supply chain: Wistron、FII/Hon Hai、Wiwynn、Quanta、Delta Electronics、AVC、BizLink、King Slide、Accton、Chenbro、Gold Circuit、Unimicron、NYPCB、Zhen Ding、Yageo、FIT、FositekThe report lists these as key stock ideas in AI server hardware.
- Strengths
- Benefiting from GPU/ASIC server upgrades, GB300 deliveries, rack volume ramp-up, and tightness in liquid cooling, power supply, connectivity, and PCB/ABF.
- Weaknesses
- Highly dependent on component supply, capacity expansion timing, and customer platform schedules.
- Comparison
- Relative to the consumer electronics chain, the AI server supply chain has higher growth visibility and stronger product-upgrade elasticity.
- Risks
- Supply shortages, component delays, raw material price increases, and geopolitics could affect shipment timing.
- Edge AI: Xiaomi、Luxshare、LenovoThe report lists these as edge AI-related stock ideas.
- Strengths
- Benefiting from the spread of AI functions to end devices and potential replacement demand.
- Weaknesses
- Edge AI has not yet significantly changed upgrade intentions overseas, and demand elasticity still needs to be verified.
- Comparison
- Compared with cloud AI servers, edge AI has greater uncertainty in commercialization timing and user willingness to pay.
- Risks
- Without truly personalized digital assistants or major feature upgrades, AI may have limited impact on replacement demand.
- Apple and Apple supply chainThe report relatively prefers the Apple supply chain over the Android supply chain.
- Strengths
- iPhone users have lower price elasticity, and surveys show Apple is the only smartphone OEM with a positive and improving net switching rate.
- Weaknesses
- Rising memory prices will still affect costs, and 2026 shipments could also be affected by changes in the iPhone launch schedule.
- Comparison
- Relative to Android, Apple and its supply chain are better positioned in terms of shipment resilience, product mix, and cost pass-through.
- Risks
- If rising costs need to be shared by non-memory suppliers, supplier margins may still come under pressure.
- Android smartphone supply chainThe report holds a cautious view on it.
- Strengths
- There is a large shipment base, but the report does not emphasize clear incremental catalysts.
- Weaknesses
- Lower gross margins, a higher share of memory costs in COGS, and greater customer price elasticity.
- Comparison
- Clearly weaker than the Apple supply chain, especially as the mid-range Android market is more vulnerable to demand destruction.
- Risks
- Price increases, spec downgrades, shipment declines, and margin pressure may occur simultaneously.
- PC/laptops and Taiwan OEMsSupported in the short term by tight supply-demand conditions and optimized product mix, but caution is needed for 2H26.
- Strengths
- 1Q revenue for Taiwan OEMs was better than expected, PC price increases outpaced cost increases, and OEMs shifted toward more high-end models.
- Weaknesses
- This margin improvement may be temporary, with subsequent demand and cost pressures gradually emerging.
- Comparison
- Relative to AI servers, PC growth is less sustainable and depends more on short-term supply-demand mismatch.
- Risks
- Demand slowdown and margin pressure may appear in 2H26.
Key data
- Report date2026-06-14 08:05 GMTTime shown on the report cover.
- 2026 GB200/GB300 rack forecast70-80K racksAbout 29K racks in 2025, with significant growth expected in 2026.
- VR200 mass production timingDelayed by 1-2 months, with rack ramp expected to begin in 4Q26Reasons for the delay include component issues such as the Midplane PCB.
- Potential CoWoS capacityCould expand to 165Kwpm in 2027The report says strong AI demand could drive TSMC capacity expansion.
- AI compute wafer consumptionCould rise to US$26B in CY26Used to measure the pull of AI compute demand on wafers and the advanced packaging chain.
- Memory share of Rubin BOM25%+Rising memory prices increase the share of memory costs in the Rubin BOM.
- Incremental MLCC market opportunityAbout US$1BCloud and edge AI are expected to bring an approximately US$1B-level TAM increase to the overall MLCC market.
- Global smartphone upgrade rate43%, up 5ppt YoYThe AlphaWise survey shows the highest level in more than a decade, mainly driven by iPhone strength.
- 2026 smartphone shipment forecastAndroid down 15% YoY, global down 13% YoYThe report body forecasts that price increases will lead to demand destruction, especially in the mid-range Android market.
- Trade and originServers and switches produced in Mexico qualify under USMCA and enjoy 0% tariffsThe current USMCA remains valid through July 2036.
Impact & implications
The investment implication is that AI compute expansion continues to favor server ODMs, liquid cooling, power supply, connectivity, PCB/ABF, advanced packaging, and passive component suppliers with integration capability and mass-production track records; however, the consumer electronics chain needs differentiation between Apple and Android, with the latter more exposed to price increases and demand elasticity pressure. Regional supply chains are also continuing to diversify, with Vietnam, Thailand, India, and Mexico becoming more important.
Risks
- Consumer electronics may be hit by rising memory prices in 2H26, putting pressure on demand and margins.
- Rising prices and tight supply of raw materials such as copper and nickel will create margin headwinds.
- Component shortages may delay AI server shipment timing; VR200 has already been delayed by 1-2 months due to issues such as the Midplane PCB.
- Geopolitical tensions may affect AI data center infrastructure spending.
- If CoWoP is used for Rubin Ultra, PCB line width/spacing would need to fall below 10/10µm, which could bring significant yield risks and supply chain restructuring risks.
- Smartphone brand price increases may lead to demand destruction, especially in the mid-range Android market.
- The report discloses that Morgan Stanley has or seeks investment banking and other service relationships with multiple covered companies, and investors should note potential conflicts of interest.
What to watch
- Whether GB300 server rack deliveries proceed smoothly, and the pace of VR200 ramp-up in 4Q26.
- The actual BOM changes from platform design upgrades such as Vera Rubin, the Kyber architecture, and LPX/Vera CPU racks.
- The scale-up of ASIC/GPU server platforms such as TPU, Trainium, and AMD MI400-series Helios.
- Whether TSMC CoWoS expansion, ABF substrates, PCB, and advanced packaging material supply remain tight.
- The penetration pace of 800VDC power supply, liquid cooling, interconnect, and CPO upgrades in server architectures.
- Trends in key costs such as memory, copper, and nickel, and their impact on PC and smartphone gross margins.
- Whether PC demand in 2H26 shifts from first-half shortage to weakness.
- Whether AlphaWise smartphone upgrade rates, Apple net switching rates, and Android shipment downgrades continue to be validated.
- Changes in the share of Vietnam, Thailand, India, and Mexico in the electronics manufacturing export chain.