Nomura maintains Buy on TSMC: N2 ramps faster and AI advanced packaging demand continues to support growth

Overview

This report is Nomura's company research in the form of conference notes on TSMC's North America Technology Symposium. The report focuses on TSMC's updates in advanced logic process technology, advanced packaging, 3D stacking, automotive-grade processes, and global capacity expansion, and links these technology roadmaps to AI/HPC demand, data-center chip demand, and future supply constraints.

Core views

Nomura's core view is that TSMC remains a key enabler in the AI-driven upcycle. Although N2 is more complex technologically, its yield learning is faster and demand is stronger, with rapid expansion across five phases in Taiwan in 2026; A14, A13, and A12 extend the advanced logic roadmap; CoWoS, SolC, SoW, and COUPE together form AI system integration capability. The report also points out that incremental greenfield capacity additions around 2027 will be limited, and if data-center chip demand continues to strengthen, supply tightness may persist.

Analysis framework

The report uses a combination of symposium takeaway synthesis, process roadmap comparison, capacity build-out pace assessment, packaging economics analysis, and valuation multiple methodology. The authors map TSMC's disclosed node performance, yields, tape-outs, capacity CAGR, advanced packaging sizes, and overseas fab progress to the medium- to long-term business implications for AI/HPC, automotive, RF, DDIC, and mature-node foundry services.

Methodology notes

• Valuation methodsP/E multiple method

  Target price based on 25x average 2026-27F EPS

  Nomura derives the TWD2,820 target price using TWD112 average 2026-27F EPS and a 25x target P/E, a multiple near the high end of the historical 10-30x range.

• Technology roadmap analysisProcess node roadmap

  N2, N2P, A16, N2X, N2U, A14, A13, A12

  By comparing mass-production timing, improvements in performance/power/area, yield curves, and tape-out momentum, the report assesses TSMC's advanced logic competitiveness and demand visibility.

• Capacity analysisFab construction and conversion pace

  9 fab-site phase constructions/conversions in each of 2025 and 2026

  Combined with an approximately two-year minimum fab construction cycle, this suggests most greenfield capacity will come online after 2027, and near-term supply may still remain tight.

• Supply chain mappingAI system integration roadmap

  CoWoS, SolC, SoW, COUPE

  Starting from AI training and inference needs for bandwidth, latency, HBM integration, and optical interconnect, the report evaluates the growth potential of advanced packaging and 3D integration.

Asset mapping & comparison

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

• Taiwan Semiconductor Manufacturing Corp (2330.TW)

  Main subject of the report and core beneficiary

  Strengths

  Position as an AI enabler, strong N2 demand, leading advanced packaging roadmap, expanding global capacity footprint, and strong yield improvement capability.

  Weaknesses

  High capex requirements, economic challenges for ultra-large advanced packaging solutions, and capacity release constrained by construction cycles and human resources.

  Comparison

  Compared with mature-node foundries, TSMC benefits more from advanced logic and AI/HPC demand; compared with potential alternative packaging solutions, TSMC still needs to accelerate the maturity of new routes such as CoPoS.

  Risks

  Macro and geopolitical factors, weaker-than-expected demand, slower-than-expected technology migration, and intensifying competition in advanced nodes.

• Besi (BESI NA)

  Potential read-through beneficiary of SolC expansion

  Strengths

  The report notes that TSMC's high SolC capacity growth is one reason Nomura recently upgraded its rating on Besi.

  Weaknesses

  Affected by the pace of TSMC SolC mass production and the intensity of equipment adoption.

  Comparison

  Relative to traditional packaging equipment, hybrid bonding and 3D stacking equipment have more direct exposure to SolC growth.

  Risks

  Slower-than-expected SolC development or mass-production timing.

• Mature-node foundries such as Vanguard and UMC

  Indirect beneficiaries under resource reallocation

  Strengths

  If TSMC shifts more resources from traditional fabs to advanced logic, the supply landscape for mature nodes may improve.

  Weaknesses

  Demand and pricing remain subject to cyclicality.

  Comparison

  TSMC's strategic focus is more tilted toward advanced logic, and mature-node foundries may gain support during pricing adjustments.

  Risks

  Weak mature-node demand or intensifying capacity expansion competition.

Key data

• RatingBuyNomura maintains a Buy rating on TSMC.
• Target priceTWD 2,820Based on 25x average 2026-27F EPS of TWD112.
• Closing priceTWD 2,185As of April 24, 2026.
• Implied upsideApproximately 29.1%Calculated based on the target price and closing price.
• N2 demandSecond-year new tape-outs are 4x N5Reflecting accelerated adoption by AI/HPC products.
• N2 capacity70% CAGR in 2026-2028TSMC plans rapid ramp-up of N2/A16 capacity.
• First-year N2 wafer output45% higher than first-year N3TSMC plans to expand N2 across five phases in Taiwan in 2026.
• AI/HPC wafer demand11x growth from 2022 to 2026TSMC estimates a significant expansion in AI/HPC wafer demand.
• Large-die wafer shipments6x growth from 2022 to 2026Large die is defined as die size greater than 500mm².
• Combined N3/N5 capacity25% CAGR in 2022-2027Used to meet current demand from major AI chip platforms.
• CoWoS capacityMore than 80% CAGR in 2022-2027TSMC plans to continue expanding advanced packaging capacity.
• SolC capacityMore than 90% CAGR in 2022-2027The report believes SolC is an important capability for future logic stacking and AI inference.

Impact & implications

For TSMC, the symposium information reinforces its pricing and share advantages amid dual bottlenecks in advanced logic and advanced packaging. N2 and the A14 series support long-term node migration, while CoWoS/SolC/COUPE support the system-level evolution of AI chips from training to inference. For the supply chain, advanced packaging equipment, HBM, materials, testing, and the OSAT ecosystem may benefit; mature-node foundry players may also receive some pricing support as TSMC allocates resources toward advanced logic.

Risks

• Top-down macro risks such as U.S.-China trade tensions.
• End demand weaker than strong supply-chain demand expectations.
• Technology migration slower than expected.
• Competition in advanced 5nm/3nm nodes stronger than expected.
• Uncertainty over the economics of ultra-large CoWoS and the mass-production progress of CoPoS.
• Capacity ramp at overseas fabs and greenfield sites constrained by construction cycles, human resources, and execution progress.

What to watch

• Yield, output, and customer tape-out delivery for N2's five-phase expansion in 2026.
• Whether the mass-production nodes for A16, N2X, N2U, A14, A13, and A12 advance as planned.
• Cost, yield, and customer design trade-offs as CoWoS expands to 14x and above 14x reticle size.
• Development progress after construction of the CoPoS mini-line in mid-2026 and the high-volume production timeline.
• The actual adoption pace of SolC logic stacking, DRAM-on-logic, and COUPE in AI inference.
• Whether data-center chip demand in 2027 continues to outperform Asia supply-chain indicators.
• Capacity ramp and yield performance at overseas fabs in Arizona, Kumamoto, Dresden, and elsewhere.