JPMorgan has released a research report on Taiwan Semiconductor Manufacturing (TSM.US), providing an outlook for its first-quarter 2026 performance. The firm anticipates a strong rise in Q1 gross margin, driven by increasing AI demand, which is expected to lead to further upward revisions in capital expenditure. The price target has been raised to NT$2,400. As the Q1 2026 earnings season approaches, the bank believes TSMC's gross margins for Q1 and Q2 2026 will significantly exceed expectations. This is primarily due to persistently tight 3nm capacity, high utilization rates, increased demand for expedited wafers, and the depreciation of the New Taiwan dollar. The bank forecasts Q2 2026 revenue to grow 6%–8% quarter-over-quarter, with the upper limit of growth constrained by 3nm capacity. It now projects full-year 2026 revenue growth, measured in U.S. dollars, to reach 35%, with the potential for another 30% growth in 2027. Following a significant increase in AI computing demand over the past 2–3 months, the bank suggests capital expenditure for 2027/2028 could become more aggressive, with cumulative capital expenditure from 2026 to 2028 reaching approximately $190 billion. Although a formal upward revision to the full-year 2026 revenue guidance may wait until the Q2 earnings call, the bank expects management's qualitative comments on AI and iPhone demand trends to be highly positive, sufficient to offset weakness in client PC and Android smartphone demand. JPMorgan has raised its 2026/2027 profit forecasts by 4% and 6%, respectively, reflecting stronger growth and better gross margins, and increased its end-2026 price target to NT$2,400, corresponding to a 12-month forward P/E ratio of 20x.
Over the past 2–3 months, AI computing demand has intensified significantly, leading to unprecedented tightness in advanced node capacity. This surge is driven by robust growth in autonomous AI agent workloads, following product releases such as ClaudeCode, OpenClaw, and others from AI labs. The bank's research indicates that demand for advanced node wafers is now tighter, with most high-performance computing clients having already secured 3nm and 2nm capacity for 2027. In addition to strong demand for AI accelerators and networking chips, a recent recovery in CPU demand is further straining advanced node capacity. Despite TSMC's plans to add new supply by the end of 2026, the bank still expects 3nm utilization rates to exceed 120% in 2026 and 110% in 2027. Revenue from the 3nm business is projected to triple in 2026, accounting for over 30% of total revenue, with HPC demand comprising about two-thirds of total 3nm demand. Revenue growth for both 2026 and 2027 is forecast to exceed 30%. Growth in 2026 is primarily supply-constrained amid strong demand and TSMC's accelerated capacity expansion.
Against this backdrop of robust demand and rapid capacity expansion, the bank projects USD-denominated revenue growth of 35% for 2026 and 30% for 2027. Growth drivers include rising 3nm demand, increased need for advanced packaging, and the faster introduction of 2nm technology in 2027. JPMorgan states that overall demand in 2026 will still outstrip supply. TSMC is addressing this through innovative methods to expand 3nm/4nm capacity, such as utilizing 7nm fabs to assist with back-end metal layer processing for 3nm, using 28nm fabs to assist with back-end metal layer processing for 5nm, and bringing forward parts of the 3nm capacity at Fab 18 Phase 9 to Q4 2025. Consequently, the bank expects wafer shipment volume to grow 8% year-over-year in 2026/2027, faster than the growth seen over the past two years. Benefiting from an improved product mix favoring 3nm/2nm, a higher proportion of HPC within the 3nm segment, and a 6%–10% price increase for equivalent specifications, the blended average selling price per wafer is expected to maintain a year-over-year growth rate of approximately 20%. Looking ahead to 2027, if Taiwan's electricity prices rise again due to higher energy costs, advanced nodes could see another 4%–5% price increase for equivalent specifications.
Comments