Goldman Sachs sees AI demand, Arizona execution and durable manufacturing advantages supporting TSMC's long-term margin goals.

Overview

This conference-takeaways report presents Goldman Sachs' positive view of TSMC's ability to convert structural AI demand into advanced-node and packaging growth while preserving long-term profitability. The report emphasizes demand validation, Arizona manufacturing progress, capacity flexibility, and the value of TSMC's technology and customer trust.

Core views

TSMC management characterizes AI as a multi-year megatrend still at an early stage. It expects the transition from generative AI to agentic AI to broaden compute demand beyond accelerators to CPUs, while open-source models and greater computing efficiency reduce token costs and expand AI consumption. Management therefore expects incremental demand for advanced silicon, and its discussions with direct customers, customers' customers, and US cloud service providers reinforced its confidence that power and broader infrastructure deployment are being prepared for this demand. Management says capacity expansion is supported by demand validation rather than a simple short-term cycle call. Its top-down work considers macro conditions, semiconductor demand, fabless and system companies, and the foundry industry; its bottom-up work examines individual segments and customer and end-customer feedback. While management acknowledges that semiconductors remain cyclical, it argues that identifying the underlying megatrend matters more than forecasting every near-term fluctuation. N3 demand remains unusually strong: TSMC continues adding capacity and does not expect supply to be sufficient, unlike prior node cycles in which demand generally declined two to three years after introduction. Arizona is presented as proof that TSMC's manufacturing model can operate overseas. The N4-based Phase 1 is in production, has yields comparable to Taiwan, satisfactory quality and reliability, and is profitable. Equipment move-in for the N3-based Phase 2 is expected soon, Phase 3 construction has begun, and preparation is under way for Phase 4 and the first US advanced-packaging facility. TSMC has also acquired land for another five to six fabs alongside its recently announced incremental US$100 billion investment. It expects roughly 30% of N2-and-beyond capacity to be outside Taiwan in five years, primarily in Arizona, although leading technologies will initially ramp in Taiwan because early technology transfer requires close R&D and manufacturing collaboration. The report argues that geographic flexibility has become a priced source of customer value. Customers initially sought only limited geographically diversified supply, but their requirements have grown as Arizona execution has been demonstrated. Management intends to price this flexibility sustainably rather than react opportunistically to utilization. Its broader moat rests on leading technology, rapid high-volume ramps with attractive economics, sufficient capacity, and the trust that TSMC will not compete with customers. Management believes its roadmap execution, wafer and advanced-packaging investment scale, and record of delivering commitments cannot be replicated by capital spending or engineering recruitment alone. TSMC sees manufacturing flexibility and advanced packaging as important ways to support wafer growth. It can raise output through productivity gains and shift tools or capacity across nodes, including using portions of N5 capacity for N3 demand where feasible. This cross-node optimization is intended to improve responsiveness to demand and reduce capacity-obsolescence risk. CoWoS capacity remains tight, and TSMC has begun outsourcing selected packaging processes to OSAT partners to enable further growth. Management describes packaging primarily as a means to secure wafer demand, not a standalone profit pool, and cites a roadmap from approximately 5.5x to approximately 14x reticle sizes over the next several years. It does not expect packaging competitors to materially displace its leading-edge wafer business because foundry and packaging are different businesses. Management also states that disclosed AI revenue understates TSMC's broader exposure because its current definition mainly includes GPUs, custom AI accelerators, and memory base die, while excluding CPUs and networking products whose end uses are difficult to identify. Agentic AI could increase demand across x86, Arm and RISC-V CPUs as well as networking and supporting silicon. On technology, TSMC is ramping N2, expects A14 around two years later, and says development of the subsequent node is already under way, with no visible technical barrier to continued transistor scaling. On profitability, management reiterates a long-term gross-margin target of 56% and above, with emphasis on the higher end. It identifies new-technology ramps, pricing, cost, product mix, utilization and FX as the six main drivers. The N2 ramp is expected to dilute 2026 gross margin by about 2–3 percentage points, though N2 profitability should improve toward the corporate average and is structurally better than N3 at a comparable stage. Overseas expansion is expected to dilute gross margin by 2–3 points early in the current five-year period and by 3 points later as more overseas fabs start production. Management believes disciplined capacity planning, healthy utilization, cost reductions, productivity gains and value-based pricing can offset these pressures. Goldman Sachs maintains Buy. Its 12-month NT$3,100 target for 2330.TW is based on a 22x target P/E applied to 2027E EPS; the multiple is benchmarked at one standard deviation above TSMC's five-year trading average. The US$620 ADR target uses a USD/TWD rate of 30.0 and a 20% ADR premium.

Analysis framework

The report synthesizes management commentary from the Communacopia + Technology conference, testing the AI demand outlook through TSMC's top-down and bottom-up demand-validation process. It then connects manufacturing execution, capacity and packaging plans, competitive positioning, and gross-margin drivers to Goldman Sachs' target-price framework based on 2027E EPS and a target P/E multiple.

Methodology notes

• Industry AnalysisSupply-demand framework

  Top-down and bottom-up demand validation

  TSMC assesses demand through macro and industry conditions as well as segment-level customer and end-customer feedback before committing capacity expansion.

• Valuation methodsP/E and PEG Valuation

  Target P/E valuation

  Goldman Sachs applies a 22x target P/E multiple to 2027E EPS to derive the NT$3,100 target price for 2330.TW.

Asset mapping & comparison

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

• TSMC (2330.TW)

  Primary covered equity; positioned to benefit from AI-driven advanced-silicon and wafer demand.

  Strengths

  Technology roadmap, high-volume manufacturing execution, capacity scale, customer trust, cross-node flexibility and Arizona progress.

  Weaknesses

  N2 ramp and overseas expansion are expected to dilute gross margin during their early stages.

  Comparison

  Management argues that execution in high-volume manufacturing, rather than attractive competing roadmaps alone, will determine competitive outcomes.

  Risks

  End-demand weakness, slower node migrations, lower AI investment, execution or yield issues, competition, FX and cost pressures.

• TSMC (ADR) (TSM)

  ADR representing the same covered company; target price is linked to the Taiwan listing through FX and an ADR premium.

  Strengths

  Exposure to the same AI, advanced-node, packaging and manufacturing-execution drivers as TSMC's Taiwan listing.

  Comparison

  The US$620 target incorporates a USD/TWD rate of 30.0 and a 20% ADR premium.

  Risks

  Subject to the same operating and margin risks identified for TSMC, plus FX sensitivity reflected in ADR valuation.

Key data

• 2330.TW 12-month target priceNT$3,100Based on 22x 2027E EPS; 25.8% upside from NT$2,465.00.
• TSM ADR 12-month target priceUS$620Based on USD/TWD of 30.0 and a 20% ADR premium; 42.4% upside from US$435.36.
• Long-term gross-margin target56% and aboveManagement remains comfortable despite N2-ramp and overseas-expansion dilution.
• N2 gross-margin dilution in 2026Approximately 2–3 percentage pointsExpected to improve over time as N2 profitability approaches the corporate average.
• Overseas-expansion gross-margin dilution2–3 percentage points early; 3 percentage points laterExpected during the current five-year period as additional overseas fabs begin production.
• N2-and-beyond capacity outside TaiwanApproximately 30% in five yearsPrimarily expected in Arizona.
• Arizona additional site capacityFive to six additional fabsSupported by an acquired parcel alongside the announced incremental US$100 billion investment.

Impact & implications

The report views AI-led demand, durable N3 requirements, cross-node capacity flexibility and advanced-packaging expansion as reinforcing TSMC's leading-edge wafer growth. It also argues that proven Arizona execution can support customer demand for geographically diversified supply and associated pricing, although new-node ramps and overseas fabs create near-term margin dilution.

Risks

• Further deterioration in end-demand recovery could reduce capacity utilization.
• Slower customer node migrations could weaken advanced-node demand.
• A slowdown in AI investment could lower long-term semiconductor content growth.
• Poor yields or execution could result in worse-than-expected profitability.
• Stronger competition could erode average selling prices and profitability.
• An unfavorable FX trend or higher-than-expected costs could pressure the margin outlook.

What to watch

• AI adoption breadth, including demand expansion from accelerators into CPUs, networking and supporting silicon.
• N3 demand and whether incremental capacity remains insufficient to meet customer requirements.
• Arizona Phase 2 equipment move-in, Phase 3 construction, Phase 4 preparation and US advanced-packaging progress.
• Customer willingness to pay for geographically diversified supply.
• N2 ramp profitability, overseas-fab dilution, utilization, cost reductions and progress toward the 56%-and-above gross-margin target.
• CoWoS capacity tightness and the expansion of advanced-packaging capability.