AI and data center demand continues to drive Taiwan’s electronics and semiconductor supply chain

Overview

This report tracks June 2026 sales and 2Q26 revenue performance across Taiwan’s electronics and semiconductor supply chain. The key conclusion is that AI, data center, and networking-related demand remains strong, driving multiple subsectors to exceed revenue expectations. Citi believes that after earnings season begins, some companies may provide more positive 2H26 outlooks, with upward revisions potentially starting with TSMC.

Core views

The report’s core judgments are: first, visibility into advanced-process and AI-related orders is strong, and TSMC may raise its full-year revenue growth guidance; second, demand for advanced packaging, OSAT, testing interfaces, and testing equipment is strong, reflecting continued progress in AI GPU, HPC, ASIC, and CPU projects; third, CCL and ABF substrates are being driven by price increases and recovering demand for high-end AI products, outperforming traditional PCB; fourth, server ODMs and the AWS supply chain are supported by new-platform ramps and outperform notebook ODMs; fifth, power, thermal, and passive components are supported by AI data center infrastructure expansion and have relatively strong revenue visibility.

Analysis framework

The report uses a monthly sales-tracking framework, comparing June sales, 2Q26 revenue, sequential and year-over-year changes across Taiwan’s technology supply chain with Citi forecasts and Bloomberg consensus expectations, while decomposing trends across wafer foundries, OSAT, testing interfaces, IC design, substrates/PCB/CCL, ODMs, power and thermal components, and passive components.

Methodology notes

• Industry monthly trackingTaiwan technology supply chain monthly revenue tracking

  Assess industry momentum through companies’ disclosed monthly sales and quarterly revenue

  The report uses June sales and 2Q26 revenue as primary evidence, comparing year-over-year, month-over-month, quarter-over-quarter changes and performance relative to Citi and market consensus expectations.

• Expectation gap analysisCiti/BBG consensus tracking

  Measure beats or misses by comparing actual revenue with Citi forecasts and Bloomberg consensus expectations

  Multiple segments are described as beat, in line, or slightly below expectation; the investment implication comes from revenue realization and the possibility of subsequent guidance upgrades.

• Supply chain mappingAI data center supply chain tiered analysis

  Break down beneficiary segments by AI chips, servers, substrates, testing, thermal management, power, and passive components

  The report maps demand for AI GPUs, CPUs, ASICs, HPC, Trainium 3, GB racks, liquid cooling, and high-voltage power supplies to companies in Taiwan’s supply chain.

Asset mapping & comparison

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

• TSMC and advanced-process supply chain

  Core beneficiaries of demand for AI chips, CPUs, and advanced nodes

  Strengths

  Strong order visibility, N2 ramp-up, June and 2Q26 revenue ahead of expectations, and room for an upward revision to full-year revenue guidance.

  Weaknesses

  Capital expenditure is constrained by long WFE lead times, and near-term capex forecasts have not been raised accordingly.

  Comparison

  Compared with mature processes, advanced processes benefit more directly from AI demand; however, mature-node utilization at UMC and VIS is also improving.

  Risks

  Changes in AI order cadence, an advanced-node ramp-up falling short of expectations, and customer capex adjustments.

• OSAT, advanced packaging, testing interfaces, and testing equipment

  Driven by testing and packaging demand for AI GPUs, HPC, ASICs, and CPU platforms

  Strengths

  ASEH, KYEC, MPI, WinWay, CHPT, and Hon Precision reported strong revenue performance. Several companies announced additional capex or capacity plans, indicating improved backlog visibility.

  Weaknesses

  Some growth depends on a limited number of high-end platforms and key customer projects.

  Comparison

  Testing interfaces and testing equipment are outperforming the traditional consumer electronics chain in the current AI chip upgrade cycle.

  Risks

  Delays in AI GPU/ASIC projects, customer concentration, and utilization fluctuations after capacity expansion.

• CCL, ABF substrates, and PCB

  Beneficiaries of the upgrade cycle in high-end AI materials and substrates

  Strengths

  EMC significantly exceeded expectations, while TUC may benefit from price increases; ABF/BT demand is recovering month by month, and NYPCB, Kinsus, and Unimicron are preferred.

  Weaknesses

  Traditional PCB is affected by CCL shortages and delays in passing through raw-material price increases, leading to a relatively less positive view.

  Comparison

  CCL and substrates outperform standard PCB because pricing and demand for high-end AI products are stronger.

  Risks

  Rising raw-material prices, ineffective price pass-through, and some sales being deferred to 3Q26.

• Server ODMs and AWS supply chain

  Driven by the ramp-up of AI servers, GB racks, CPU servers, and Trainium 3

  Strengths

  Quanta and Wiwynn’s 2Q26 sales were slightly better than Citi and market consensus expectations; June sales at Wiwynn and Accton in the AWS supply chain recovered.

  Weaknesses

  Platform transitions and customer shipment cadence may cause monthly volatility.

  Comparison

  Server ODMs are significantly outperforming notebook ODMs; the report favors Hon Hai, Wiwynn, and Accton.

  Risks

  Cloud service provider order cadence, Trainium 3 shipments falling short of expectations, and delays in server platform transitions.

• Notebook ODMs and consumer electronics chain

  Affected by front-loaded shipments and rising component costs

  Strengths

  June notebook shipments rose 29% sequentially, and 2Q shipments were better than expected; product price increases may partially support revenue.

  Weaknesses

  End-market demand remains conservative, and 3Q26E shipments are expected to decline sequentially and year over year after front-loaded shipments.

  Comparison

  Compared with the AI server chain, the notebook ODM cycle has weaker sustainability.

  Risks

  Weaker downstream demand, component procurement constraints, and subsequent order declines caused by front-loaded shipments.

• Power, thermal management, and passive components

  Driven by AI data center infrastructure construction, liquid cooling, and high-voltage power demand

  Strengths

  Delta’s June sales showed strong growth, with power systems and liquid-cooling pipelines at full capacity; Yageo’s June sales reached a record high, with AI applications supporting growth in standard and specialty products.

  Weaknesses

  Some thermal companies reported 2Q26 revenue slightly below expectations due to GPU platform transitions and delays in customer shipments.

  Comparison

  Revenue visibility for power and passive components is stronger than for some thermal companies affected by short-term platform transitions.

  Risks

  Delays in GPU platform transitions, postponed customer shipments, and changes in the pace of AI data center construction.

Key data

• TSMC June salesNT$442.7 billion, up 68% year over year and 6% month over month2Q26 revenue rose to NT$1.27 trillion, up 12% sequentially and ahead of Citi’s expectations.
• TSMC full-year revenue outlookMay rise from approximately 30% to 35%-40% or higherThe upgrade would be based on strong order visibility, the N2 ramp-up, and AI advanced-process demand.
• TSMC 2026E capexUS$55-56 billionCiti maintains its capital expenditure forecast unchanged but does not rule out an increase due to strong demand for advanced nodes.
• UMC and VIS 2Q26 salesUp 13% and 14% sequentially, respectivelyThe report believes mature-node utilization and pricing outlook will improve in 2H26.
• ASEH and KYEC 2Q26 revenueUp 10% and 9% sequentially, respectivelyBroadly in line with market expectations; stronger advanced packaging demand provides room for an upward revision to ASEH’s LEAP revenue guidance.
• Hon Precision 2Q26 salesNT$13.8 billion, up 101% year over year and 29% sequentiallyA record high, supported by cold plate products and ATC handler sales to key GPU/ASIC customers.
• Delta June salesNT$65.6 billion, up 55% year over year and 11% month over monthDriven by AI data center power systems and liquid-cooling deployments; 2Q26 revenue was NT$183.0 billion, up 15% sequentially.
• Notebook ODM 3Q26E shipment forecastDown 6% sequentially and 16% year over yearFollowing front-loaded shipments in 2Q26, downstream expectations for 2H shipments remain conservative, although price increases may partially offset the shipment decline.

Impact & implications

For portfolios, the report reinforces the view that Taiwan’s AI data center supply chain cycle will remain strong. Relatively more attractive areas include advanced processes, advanced packaging, testing interfaces, testing equipment, CCL/ABF, server ODMs, AI power and thermal management, and selected passive components. Areas requiring greater caution include traditional PCB, notebook ODMs, and consumer electronics demand, which face delayed pass-through of raw-material price increases, demand declines after front-loaded shipments, or weak end-market demand.

Risks

• AI GPU, CPU, or ASIC platform upgrades and customer shipment cadence may fall short of expectations.
• Advanced processes, advanced packaging, testing equipment, and server ODMs have relatively high dependence on high-end AI projects and a small number of major customers.
• CCL shortages and rising raw-material prices may compress margins in the PCB segment, with potential delays in price pass-through.
• Notebook and consumer electronics end-market demand remains weak, and front-loaded shipments may pull forward 2H26 demand.
• Delays in GPU platform transitions, the Trainium 3 ramp-up, or Vera Rubin introduction may affect revenue across the server, thermal, and testing chains.
• Regulatory disclosures indicate that Citigroup or its affiliates have investment-banking, non-investment-banking, or significant financial-interest relationships with several covered companies; investors should monitor potential conflicts of interest.

What to watch

• Whether TSMC raises its full-year revenue growth guidance to 35%-40% or higher during earnings season.
• Whether TSMC raises its 2026E capex outlook due to strong demand for AI advanced nodes.
• Whether ASEH’s LEAP revenue guidance exceeds the previous expectation of more than US$3.5 billion.
• Whether CCL and ABF substrate price increases can continue, and whether PCB manufacturers can successfully pass through raw-material costs.
• The strength of AWS Trainium 3 shipments in 3Q26 and the revenue performance of supply-chain companies such as Wiwynn and Accton.
• Whether notebook ODM 3Q26 shipments decline sequentially as expected, and whether price increases can offset shipment pressure.
• The pace at which liquid cooling, high-voltage power, and Vera Rubin-related revenue materializes by the end of 3Q26 and thereafter.