Before TSMC’s financial report, JPMorgan loudly shouted: Buy! The shortage of the company’s advanced manufacturing processes will continue until 2027.

Before TSMC’s financial report, JPMorgan loudly shouted: Buy! The shortage of the company’s advanced manufacturing processes will continue until 2027.

```

The explosive growth in AI computing power demand is pushing TSMC’s advanced process capacity to unprecedented levels of tightness.

According to Wind Trading Desk, JPMorgan released a research report on April 2, 2026, raising TSMC's target price from NT$2,250 to NT$2,400. The report anticipates that TSMC's gross margin in the first quarter to be announced will reach 66.8%, higher than the company's own guidance range of 63% to 65%.

The core logic behind the upward adjustment is: Advanced process capacity, represented by N3, has been pre-booked until 2027, and the supply-demand imbalance has further intensified over the past two to three months.

The strong growth in Agentic AI workloads has driven an exponential increase in computing power consumption. At the same time, demand for various products, including GPUs, CPUs, and AI connectivity, has risen synchronously, causing TSMC’s advanced process utilization rate to continuously surpass 100%.

This supply-demand dynamic will support TSMC’s gross margin to continue rising in the first half of 2026 and is expected to push the company's cumulative capital expenditure to about $190 billion from 2026 to 2028.

Surge in AI Computing Power Demand, N3 Capacity Pre-booked Until 2027

In the past two to three months, AI computing power demand has clearly accelerated. The boom in Agentic AI workloads—including the launch of products like Claude Code, OpenClaw, and other AI labs’ similar products, driving a surge in token consumption—is the primary driver of this round's unexpectedly high demand.

Meanwhile, demand for server AI connectivity products (including network chips, DSPs, retimers, etc.) is rising along with computing capacity needs, and server CPUs like Google Axion, Nvidia Grace, AWS Graviton, and Microsoft Cobalt are also seeing a clear rebound in demand.

Currently, the demand for advanced process wafers has tightened significantly, and N3 and N2 capacity has been pre-booked till 2027 for most HPC customers.

Major AI accelerators like NVIDIA Rubin, AWS Trainium 3, Meta MTIA, etc. are migrating to N3 process in 2026. Combined with the continuous volume production of Google TPU (Ironwood and v8 series) and AMD MI series, the supply-demand imbalance for N3 is further intensifying.

Although TSMC is striving to expand capacity by the end of 2026, N3 utilization rate will still remain above 120% and 110% in 2026/2027, respectively. N3 revenue is expected to almost double in 2026, accounting for over 30% of TSMC's total revenue, with HPC demand making up about two-thirds of N3's total demand.

On N2, Apple’s full iPhone lineup (A20/A20 Pro), AMD Venice and MI450, as well as flagship SoCs from MediaTek and Qualcomm, will adopt the N2 process in 2026. N2 revenue share is expected to grow from about 4% in 2026 to 15% and 26% in 2027/2028, respectively.

Active Capacity Expansion, Capital Expenditure to Reach $190 Billion from 2026 to 2028

TSMC has raised its capital expenditure forecast for 2027 and 2028 to $63.5 billion and $70.5 billion, respectively. Total capital expenditures from 2026 to 2028 are expected to reach about $190 billion, almost double the approximately $101 billion spent from 2023 to 2025.

This large-scale investment reflects TSMC’s strategic determination to fully support AI computing power expansion, backed by rapid client revenue growth and exponential token usage. Advanced process capacity (N7 and below nodes) is expected to achieve a 16% CAGR from 2025 to 2028.

Global Capacity Expansion Accelerating Simultaneously:

In Taiwan, TSMC is advancing N3 expansion at Tainan Fab 18 P9, N2 and A16 capacity build-out at Kaohsiung Fab 22 P1 through P5, and construction at Baoshan Fab 20 P1 through P4;

For long-term planning, Taichung Fab 25 will be set up for A14/A10 nodes. Fab 18 is expected to add phases P10 to P12 (pending land approval), and there is also early planning for another eight-phase wafer plant in Tainan aimed at A10/A7 nodes.

Overseas, phase II N3 mass production at Arizona Fab 21 in the USA may be advanced to mid-2027. Phase III N2 capacity could start between late 2028 and the first half of 2029; at Kumamoto, Japan, Fab 23 phase II N3 capacity is expected to begin mass production in the second half of 2028.

For advanced packaging, CoWoS capacity is expected to reach 115,000 and 145,000 wafers/month by the end of 2026 and 2027, respectively. SoIC will enter large-scale investment from 2027 to 2028, and N2 accelerators such as NVIDIA Feynman, Google TPU v9, Trainium 4, and OpenAI Titan 2 are expected to adopt the SoIC architecture.

Gross Margin Exceeds Expectations, May Reach Historic High in First Half

TSMC’s first quarter gross margin is expected to reach 66.8%, about 180 basis points above the company's upper guidance limit, and stay at a high 66.4% in the second quarter, jumping over 550 basis points from the second half of 2025.

The report lists five factors supporting the gross margin's outperformance:

First, continued N5 and N3 utilization rates above 100%;

Second, cross-fab operations improved N7 and 28nm utilization, reducing idle equipment costs;

Third, increases in "hot run/super-hot run" wafer orders from HPC clients, with such orders commanding 50% to 100% premiums;

Fourth, a 6% to 10% price hike across similar advanced process products has been fully implemented since January 2026;

Fifth, the Taiwan dollar depreciated over 1.5% against the US dollar in the past two months, improving gross margin by about 0.4 percentage points for every 1% depreciation.

It is conservatively estimated that gross margin will fall back about 400 basis points in the second half of 2026, mainly reflecting the dilution from ramping up N2 process yields, earlier-stage amortization from rapidly built overseas fabs (especially in the US), and the potential impact of electricity price increases.

Market Expectations Before Earnings: Q1 Exceeds Projections, Full-Year Guidance Unchanged

TSMC's first-quarter revenue is expected to grow 7% quarter-on-quarter, above the company's guidance range (3% to 6% sequential increase, or $33.5 to $35.8 billion). Q2 guidance is expected to be a sequential growth of 6% to 8% (estimate: 8%), with gross margin guidance at 64% to 66%.

Despite continued strong demand, the report expects TSMC will not adjust its annual guidance during this earnings call, and management is likely to maintain the following three guidance items:

Dollar revenue growth for 2026 "close to 30%" (forecast: 35%);

AI revenue CAGR from 2024 to 2029 to be "mid to high 50%";

And company dollar revenue long-term annual growth of around 25% from 2024 to 2029.

Management is expected to send the following signals during this call:

A more positive attitude towards AI demand;

A more aggressive stance on capacity expansion plans for 2027–2028;

Emphasis on strong demand in the Apple supply chain;

Accelerated progress in US N3 and N2 capacity expansion;

And continued confidence regarding the competitive landscape including Tesla TeraFab, Samsung, and Intel foundry businesses.

The report adopts a valuation benchmark of around 20x 12-month forward P/E ratio, corresponding to the new target price of NT$2,400. This exceeds TSMC's five-year historical average P/E, reflecting JPMorgan’s strong confidence in TSMC’s structural growth logic.

Risk Disclosure and DisclaimerThe market has risks, and investments must be cautious. This article does not constitute personal investment advice, nor does it consider the special investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Investment is at your own risk. ```