TSMC (2330) Report Interpretation
JPMorgan remains Overweight on TSMC, arguing that strong AI accelerator demand, constrained N3 supply and technology leadership support sustained growth, margins and capacity expansion.
Summary
JPMorgan remains Overweight on TSMC, arguing that strong AI accelerator demand, constrained N3 supply and technology leadership support sustained growth, margins and capacity expansion.
- August revenue was NT$514.8bn, up 53% year on year and 10% month on month.
- JPMorgan expects 3Q26 revenue at the mid-to-high end of US$44.6bn-US$45.8bn guidance.
- N3 demand-supply tightness is expected to worsen in 2027 despite roughly 40% capacity expansion.
- Gross margin is forecast at about 66.5% in 3Q26 and expected to remain in the mid-to-high 60% range.
- The firm expects TSMC to retain over 95% leading-edge market share for the foreseeable future.
Report Interpretation
Overview
This company update argues that TSMC's August sales and sustained AI-driven demand put it on course for the mid-to-high end of its 3Q26 guidance. JPMorgan sees tightening N3 capacity, expanding advanced packaging demand and continued process-technology leadership as supporting a long-duration growth outlook.
Core views
TSMC's August revenue reached NT$514.8bn, rising 53% year on year and 10% month on month. This represented about 67% of JPMorgan's and Bloomberg consensus 3Q26 revenue expectations. JPMorgan therefore expects 3Q26 sales to reach the mid-to-high end of management's US$44.6bn-US$45.8bn guidance range, versus its own US$45.5bn forecast. The stated driver is robust AI-accelerator demand, with N3 and N5 utilization running above 100%. The report expects margin pressure from the N2 ramp to be manageable. JPMorgan forecasts 3Q26 gross margin of about 66.5%, down about 1.2 percentage points sequentially as N2 depreciation ramps. It nevertheless expects gross margin to remain in the mid-to-high 60% range, forecasting 66.9%, 67.5% and 68.7% for 2026, 2027 and 2028, respectively. The rationale is that N3 gross margin should exceed the corporate average in 2H26, leading-edge utilization should remain high, and expected early-2027 price increases of more than 10% for N3 and N2—along with milder increases for N5, N7 and CoWoS—should offset N2 dilution. JPMorgan's central supply-demand thesis is that the N3 shortfall will deepen into 2027. It estimates an N3 demand-supply gap of about 600,000 wafers in 2026 and expects it to widen despite roughly 40% N3 capacity growth in 2027. NVIDIA is expected to consume 55-60% of total N3 wafer supply in 2027, up from less than 10% in 1H26, as Rubin and Vera ramp. Incremental demand is also expected from agentic-AI CPUs and rising ASIC volumes, including TPUv8 and Trn3. JPMorgan forecasts TSMC capital expenditure of US$81bn in FY27 and US$90bn in FY28, with N3 capacity reaching 178,000, 223,000 and 243,000 wafers per month by end-2026, end-2027 and end-2028. It notes that capex expectations could be revised higher over coming quarters if the shortfall persists. Advanced packaging is presented as a complementary capacity and technology advantage. CoWoS capacity is expected to reach roughly 190,000-200,000 wafers per month by end-2027, up about 65% year on year. JPMorgan expects SoIC to become the primary focus in 2028, forecasting 65,000 wafers per month, based on its expectation that NVIDIA's Feynman could adopt 3D SoIC for GPU-to-GPU stacking. While Intel could win selected packaging projects, such as EMIB-T for TPU v9, JPMorgan expects CoWoS to remain the dominant platform and argues that packaging project wins do not readily translate into foundry wins because their development timelines differ. On competition, the report argues that TSMC's front-end technology lead remains intact. JPMorgan sees no meaningful Intel project wins in front-end manufacturing and says it has not identified HPC products being made at Intel 18A or 14A. By the time Intel 14A reaches mass production in 2028, JPMorgan expects TSMC's N2 family to be operating at a US$50bn-US$60bn revenue run rate and A14 to be entering mass production. It expects TSMC to maintain more than 95% share of the initial N2/A16 ramp and of the leading-edge market for the foreseeable future. JPMorgan remains Overweight on TSMC, citing higher confidence that AI demand visibility extends through 2029-30, an accelerated capacity plan and limited room for potential competition. Its Jun-27 NT$3,200 price target is based on about 20x 12-month forward P/E, above TSMC's five-year historical average multiple, reflecting anticipated stronger AI-led revenue growth and faster capacity expansion.
Analysis framework
JPMorgan combines monthly revenue tracking and quarterly guidance with utilization, gross-margin and capacity forecasts. It then uses industry checks on leading-edge wafer demand, customer ramps, foundry competition and advanced-packaging capacity to assess TSMC's growth and competitive position, and applies a forward P/E multiple to derive its price target.
Methodology notes
Leading-edge wafer supply-demand analysis
The report compares expected N3 demand from AI accelerators, CPUs and ASICs with planned capacity additions to argue that shortages and pricing power will persist.
12-month forward P/E valuation
JPMorgan bases its NT$3,200 Jun-27 target price on about 20x 12-month forward earnings, reflecting its expectations for AI-driven growth and capacity expansion.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- TSMC (2330.TW)Primary covered company; beneficiary of AI-driven leading-edge foundry and advanced-packaging demand
- Strengths
- High N3/N5 utilization, expected pricing strength, technology leadership, capacity expansion and 95%+ expected leading-edge share
- Weaknesses
- N2 ramp depreciation is expected to dilute gross margin in the near term
- Comparison
- JPMorgan expects TSMC to remain ahead of Intel and Samsung Foundry in leading-edge process technology
- Risks
- Potential shortening of the AI capex cycle and weak PC or smartphone demand in 2H26
- NVIDIAMajor demand driver for TSMC's N3 capacity through Rubin and Vera production ramps
- Strengths
- Expected to take 55-60% of total N3 wafer supply in 2027
- Comparison
- Its projected N3 demand is a major contributor to the supply-demand imbalance
- IntelFoundry competitor and potential advanced-packaging alternative
- Strengths
- May obtain selected packaging wins, including potential EMIB-T work for TPU v9
- Weaknesses
- JPMorgan sees no meaningful front-end wins and no signs of HPC products at Intel 18A or 14A
- Comparison
- TSMC's N2 family is expected to have a US$50bn-US$60bn revenue run rate when Intel 14A reaches mass production in 2028
Key data
- August revenueNT$514.8bnUp 53% year on year and 10% month on month
- 3Q26 revenue guidanceUS$44.6bn-US$45.8bnJPMorgan expects the mid-to-high end; its estimate is US$45.5bn
- 3Q26 gross margin forecast~66.5%About 1.2 percentage points lower quarter on quarter due to N2 depreciation ramp
- N3 supply-demand gap~600k wafersEstimated for 2026 and expected to widen in 2027
- FY27/FY28 capex forecastUS$81bn / US$90bnSupports aggressive leading-edge capacity expansion
- Leading-edge market share95%+JPMorgan's expectation for TSMC in the foreseeable future
Impact & implications
The report argues that AI demand and constrained leading-edge capacity should support TSMC's revenue growth, utilization, pricing and gross-margin recovery after N2 ramp dilution. It also views TSMC's process and packaging roadmap as reinforcing its competitive position against foundry challengers.
Risks
- The duration of the AI capital-expenditure growth cycle could be weaker or shorter than JPMorgan expects.
- Weak PC and smartphone demand in 2H26 could negatively affect the rating and price target.