Taiwan Electronics & Semiconductors: AI demand supports continued supply chain tightness

Overview

This report tracks monthly revenue and AI-related developments across Taiwan’s electronics and semiconductor supply chain. The core conclusion is that while the market has concerns about the adoption pace of technologies such as 800V HVDC, CPO, and VR NVL72, Citi believes Taiwan’s technology supply chain revenue trend remains solid in the near term. AI demand continues to support advanced-process foundry, packaging and testing, test interfaces, substrates/CCL, and AI server-related supply chains, with some segments potentially seeing further price increases in 2H26.

Core views

Citi’s core view is moderately positive: in foundry and OSAT, TSMC and UMC are progressing on track, while ASEH and KYEC are broadly in line with expectations; test interface and equipment orders are strong, with MPI, Hon Precision, Chroma, CHPT, and WinWay benefiting from mass production of GPU/AI ASICs and specification upgrades; in IC design, AI-related Aspeed and GUC are outperforming expectations, while non-AI names Mediatek, Realtek, and Novatek are also better than market concerns; in substrates, PCB, and CCL, the impact of CCL price increases continues to emerge; ODM/OEM overall posted slight sequential growth, with AI server ramp-up remaining the main theme, although Wiwynn is temporarily weaker due to the ramp timing of AWS AI ASIC servers.

Analysis framework

The report uses monthly revenue tracking, MoM/YoY comparisons, supply chain segment breakdowns, and analysis of orders, capacity, and price pass-through. It divides Taiwan’s technology supply chain into foundry, OSAT, test interfaces and equipment, IC design, substrates/PCB/CCL, ODM/OEM, notebooks, power and thermal, and passive components, assessing May sales performance, 2Q26 revenue trends, and potential upside or downside factors for 2H26 in each segment.

Methodology notes

• Industry Monthly TrackingTaiwan Technology Supply Chain Revenue Trends

  Observe monthly sales, MoM changes, YoY changes, order flow, and capacity constraints by sub-segment.

  This method is used to judge whether the supply chain is still expanding under AI-driven demand and which segments have revenue momentum above or below expectations.

• Supply-Demand and Price Pass-ThroughSupply Chain Tightness and Price Increase Monitoring

  Observe supply-demand gaps and realized price increases in advanced process nodes, packaging, test interfaces, substrates, CCL, PCB, and key components.

  The report views tight supply and rising costs as important support for potential price increases in 2H26, while also warning that capacity constraints may limit short-term revenue realization for some companies.

• Short-Term Catalyst DisclosureCiti Catalyst Watch/STV

  A short-term view framework for stock price rises or declines driven by specific events within 30 or 90 days.

  The appendix explains this rating disclosure framework, but the main text is primarily an industry monthly tracker and does not introduce a new explicit Catalyst Watch or STV call on any single company.

Asset mapping & comparison

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

• TSMC / UMC / VIS (Foundry)

  Core beneficiary segment of AI chips, advanced process nodes, and upward pricing trends.

  Strengths

  TSMC’s May revenue was strong, and increased shipments of N3 AI chips and N2 smartphone SoCs in 2H26 could create room for upward revisions; UMC and VIS are also seen as having upward pricing trends in 2H26.

  Weaknesses

  Highly dependent on advanced-node capacity, customer production schedules, and end-demand realization.

  Comparison

  Compared with mature-node or non-AI demand chains, advanced nodes and AI-related demand are stronger.

  Risks

  Capacity constraints, changes in customer order timing, and slower-than-expected technology adoption.

• ASEH / KYEC (OSAT and Testing)

  Beneficiaries of AI chip packaging and testing demand and new testing businesses.

  Strengths

  The report believes ASEH and KYEC’s May performance was broadly in line with expectations, with orders improving in 2H26 driven by LEAP and new AI chip testing businesses.

  Weaknesses

  Short term may be affected by customer onboarding schedules and capacity allocation.

  Comparison

  Compared with general packaging and testing demand, AI chip testing brings higher specifications and stronger order flow.

  Risks

  Slower-than-expected order ramp-up, packaging and testing capacity bottlenecks, and customer project delays.

• MPI / Hon Precision / Chroma / CHPT / WinWay (Test Interfaces and Equipment)

  Direct beneficiaries of GPU and AI ASIC mass production, specification upgrades, and CPO testing certification.

  Strengths

  MPI, Hon Precision, and Chroma have strong potential to exceed market expectations for 2Q26 revenue growth; CHPT reached another record high in May sales; although WinWay is constrained by capacity, its Renwu leased plant is expected to begin contributing in June.

  Weaknesses

  WinWay’s short-term revenue is constrained by capacity, and CPO test equipment still requires certification rollout.

  Comparison

  Test interfaces and equipment benefit more directly from AI GPU and AI ASIC upgrades than the traditional consumer electronics supply chain.

  Risks

  Slower-than-expected capacity expansion, delayed customer certification, and changes in next-generation AI GPU mass production timing.

• Aspeed / GUC / Mediatek / Realtek / Novatek / Alchip (IC Design)

  Covers AI server control chips, ASIC design services, and non-AI consumer/communications chip demand.

  Strengths

  Aspeed and GUC performed better than expected; Mediatek, Realtek, and Novatek were also better than market concerns in non-AI markets, supported by improved product mix.

  Weaknesses

  Alchip remains relatively slow, with the Trn3 ramp-up more reflected in 2H26.

  Comparison

  AI-related IC design companies have stronger growth, while non-AI companies rely more on product mix and cost pass-through.

  Risks

  Capacity constraints, weak non-AI consumer demand, and delays in customer chip projects.

• Substrates, PCB and CCL Supply Chain

  Affected by AI servers, demand for high-spec materials, and pass-through of raw material price increases.

  Strengths

  CCL sales were stronger than expected, with price increases gradually implemented customer by customer; the PCB industry is passing raw material price increases downstream, resulting in better-than-expected sales growth.

  Weaknesses

  Although the overall substrate segment posted steady sequential growth, sales momentum was below bullish expectations.

  Comparison

  CCL outperformed the broader substrate segment, while PCB was supported by price pass-through.

  Risks

  Price increase magnitude or implementation speed falling short of expectations, customer inventory adjustments, and raw material cost volatility.

• Hon Hai / Quanta / Wistron / Wiwynn (ODM/OEM and AI Servers)

  Key manufacturers undertaking AI server and system assembly ramp-up.

  Strengths

  Hon Hai’s May revenue was strong and ahead of 2Q26 expectations; Quanta and Wistron were slightly better; demand for B300 and GB300 remains strong.

  Weaknesses

  Wiwynn’s May sales were weak, possibly reflecting a slower ramp-up in AWS AI ASIC servers.

  Comparison

  The AI server supply chain is clearly stronger than traditional PC demand, but different customer and project timing creates divergence.

  Risks

  Delayed AI ASIC server ramp-up, pull-forward demand cannibalization, and changes in data center deployment timing.

• Delta / AVC / Auras (Power and Thermal)

  Affected by data center power demand, HDVC solutions, and AI platform thermal needs.

  Strengths

  Delta grew 44% YoY, with HDVC power solution ramp-up still on track; AVC and Auras are expected to see further monthly sales improvement with the TRN3 platform ramp.

  Weaknesses

  Delta’s May performance was slightly below market expectations, partly due to tight component supply and data center deployment timing.

  Comparison

  Power and thermal are necessary support for AI server expansion, but revenue realization is constrained by deployment timing.

  Risks

  Component shortages, delayed data center deployment, and demand fluctuations during GPU platform transitions.

• Yageo (Passive Components)

  Driven by AI-related applications and growth in standard/specialty products.

  Strengths

  May sales hit a record high, implying upside potential to current-quarter expectations.

  Weaknesses

  The report did not provide more detailed data on product or customer mix.

  Comparison

  Although passive components are less directly exposed than advanced process nodes, they can still benefit from spillover AI hardware demand.

  Risks

  Slowing AI demand, price fluctuations in standard products, and customer inventory adjustments.

Key data

• Report Date2026-06-10The report timestamp is 10 Jun 2026 19:28:54 ET.
• TSMC May RevenueNT$417bnUp 2% MoM and 30% YoY, with YTD revenue up 30% YoY.
• Hon Hai May RevenueNT$859.4bnUp 3% MoM and 40% YoY, ahead of Citi’s estimate for 15% QoQ revenue growth in 2Q26.
• Delta May RevenueNT$59bnFlat MoM and up 44% YoY, slightly below market expectations for 15%-20% QoQ revenue growth in 2Q26.
• Overall Notebook May Shipments+2% MoMCustomers pulled in orders early due to concerns over further price increases in components such as memory, chips, and substrates; Citi estimates 2Q26E shipments at +2% QoQ/-10% YoY.
• AVC and Auras May Sales+3% MoM / +65% YoYThermal management component sales rose slightly from April, with further support from the TRN3 platform ramp-up.
• NVL72 2Q Shipment Forecast13-15kVR NVL72 is expected to begin gradually ramping in 3Q26.
• Yageo May PerformanceReached a new monthly sales highDriven by strong demand from AI-related applications, as well as solid growth in standard and specialty products.

Impact & implications

From an investment perspective, the report reinforces the view that AI infrastructure demand is still spreading across Taiwan’s semiconductor and electronics supply chain. Advanced process nodes, packaging and testing, test interfaces, CCL, PCB, AI servers, power and thermal, and passive components may all benefit from orders, specification upgrades, or price increases; however, the pace is not fully synchronized, and short-term performance may diverge as some companies are affected by capacity, customer project ramp-up, data center deployment, and weakness in end-consumer demand.

Risks

• Capacity constraints in advanced process nodes, packaging and testing, test interfaces, and key components may limit revenue realization.
• If adoption of new technologies such as 800V HVDC, CPO, and VR NVL72 is slower than expected, market confidence in supply chain growth could weaken.
• Consumer electronics demand remains weak, and visibility on notebook orders in 2H26 is limited.
• Data center deployment plans and GPU platform transitions could cause short-term monthly sales volatility.
• If raw material and component price increases cannot be smoothly passed downstream, profit margins may be squeezed.
• If customer AI ASIC projects or server ramp-ups are delayed, related ODM/OEM and testing supply chains may underperform expectations.

What to watch

• Whether TSMC raises its full-year revenue growth target, and the actual shipment pace of N3 AI chips and N2 smartphone SoCs in 2H26.
• Whether price increases in 2H26 are truly implemented across segments such as UMC, VIS, Mediatek, CCL, and PCB.
• Whether 2Q26 revenue for MPI, Hon Precision, Chroma, CHPT, and WinWay exceeds market expectations.
• The ramp-up timetable for AWS AI ASIC servers, B300, GB300, and VR NVL72.
• The extent of monthly sales improvement for Delta’s HDVC power solutions, AVC, and Auras during the TRN3 platform transition.
• After pull-forward notebook demand, whether order declines or inventory adjustments emerge in 2H26.