Despite market noise, TSMC management maintained full-year revenue guidance and long-term growth guidance.
TSMC Q1 Earnings:
Revenue $25.5B in USD terms, up 35% year over year, down 5% sequentially. In NTD terms, NT$839.3B, up 42% year over year, down 3% sequentially, record high for the period;
Gross margin 58.8%, up 5.7 percentage points year over year, down 0.2 percentage points sequentially;
Operating income $12.4B in USD terms, up 56% year over year, down 6% sequentially; Operating margin maintained at 48.5%;
Net income $11.0B in USD terms, up 53% year over year, down 5% sequentially; Net margin maintained at 43%;
Equivalent 12-inch wafer shipments 3.259M (approx. 1.09M/month), up 8% year over year, 4th consecutive quarter of year-over-year growth, down 5% sequentially; ASP ~$7,832, up 26% year over year, 21st consecutive quarter of year-over-year growth;
Capex $10.06B in USD terms, up 74% year over year; Full-year capex guidance maintained at $38-42B;
By Process and Platform, Q1:
3nm 22%, 5nm 36%, 7nm 15%, 16/20nm 7%, 28nm 7%, 40/45nm 3%, 65nm 4%, 90nm 1%, 0.11/0.13um 2%, 0.15/0.18um 3%; Advanced nodes (3nm/5nm/7nm) 73%; 3nm+5nm combined 58%;
3nm revenue down 18% sequentially; 5nm revenue up 2% sequentially, 7th consecutive quarterly record;
HPC 59%, Smartphone 28%, IoT 5%, Auto 5%; HPC share exceeded smartphone for the 10th consecutive quarter; HPC up 82% year over year, up 7% sequentially; Smartphone up 4% year over year, down 22% sequentially; IoT down 9% sequentially; Auto up 14% sequentially; TSMC's 2024 top customer Apple 22%, second customer (likely NVIDIA) 12%; watching if 2025 sees NVIDIA overtake Apple as top customer.
Outlook:
Guiding Q2 revenue $28.4-29.2B, up 36%-40% year over year, driven primarily by strong HPC demand led by AI; Q2 gross margin 57%-59%, operating margin 47%-49%, Q2 tax rate 20%, Q3-Q4 falling back to 14%-15%;
Guiding 2025 full-year USD revenue up 24%-26% year over year (unchanged); 2024-2029 CAGR 20% (unchanged); Guiding 2025 capex $38-42B (unchanged);
Guiding 2025 AI revenue (GPU+ASIC+HBM controller) to double ($25.2-28.8B); 2024-2029 CAGR 44%-46% (unchanged); Guiding full-year advanced node (3nm/5nm/7nm) revenue share 80%;
2025 CoWoS capacity to double (unchanged); 2026 demand momentum to continue, with ongoing capacity additions aimed at achieving supply-demand balance.
2025 CoWoS capacity to double (unchanged); 2026 demand momentum continues, further expanding capacity to work toward supply-demand balance;
Long-term gross margin guidance is 53%+ (unchanged), even though overseas fabs will dilute margins by 2–3 percentage points annually for the next five years, with the dilution increasing to 3–4 percentage points later.
The company expects the global foundry 2.0 market to grow 10% in 2025 (unchanged) and continues to outperform the market.
Guiding 2025 global foundry 2.0 industry growth of 10% (unchanged); Company continues to outperform the market;
Cumulative planned investment in the U.S. is $165B. Beyond the originally planned three fabs, the company will add three more fabs, two advanced packaging plants, and one R&D center. Arizona Fab 1 (N4) began volume production in Q4 last year with yields comparable to Taiwan fabs. Fab 2 (N3) construction is complete, with volume production pulled forward to mid-2027. Fab 3 (N2) is under construction, and Fab 4 (A16) has begun early site work. Fabs 5 and 6 will adopt even more advanced nodes based on customer demand. This will form a GIGA Fab Cluster in Arizona, with 2nm becoming the mainstream node and eventually contributing roughly 30% of TSMC's total 2nm-and-below capacity.
Japan Kumamoto Fab 1 (28/22/16/12nm) began volume production at the end of 2024 with excellent yields; full capacity is 40,000 wafers/month. Construction of Fab 2 (focused on consumer electronics, automotive, industrial, and HPC) is planned to start this year. The Germany Dresden fab, focused on automotive and industrial applications, is progressing on schedule.
Planning 11 new fabs and 4 advanced packaging fabs in Taiwan over the next few years;
No joint venture, technology licensing, transfer, or sharing discussions with any other company (addressing Intel rumors);
Kumamoto Fab 1 (28/22/16/12nm) began volume production end of 2024, yields excellent, full capacity 40K wafers/month; planning to start Fab 2 construction this year (focus on consumer electronics/auto/industrial/HPC); Germany Dresden fab focused on auto/industrial, progressing on schedule;
Following TSMC earnings, both Ming-Chi Kuo and Lu Xingzhi published immediate commentary:
Ming-Chi Kuo believes TSMC's order visibility and competitive moat this year are more certain than its largest customers Apple and NVIDIA, giving the industry confidence; TSMC's long-term gross margin outlook already incorporates inflation and tariff impacts.
Lu Xingzhi argues TSMC should specifically explain how it maintains 53%+ long-term gross margin under these headwinds — whether Taiwan fab depreciation tailwinds offset US fab drag, or whether price hikes pass through costs.
One point I personally disagree with Ming-Chi Kuo: the entire semiconductor demand is now effectively driven by a single company, NVIDIA (though many are reluctant to admit this); TSMC's growth certainty is built on the certainty of many customers' demand growth, or rather TSMC's earnings certainty is more about the floor than the ceiling — similar to ASML to TSMC, or Arm to its downstream customers.
Fundamentally, at the upper end of full-year guidance, TSMC implies full-year revenue of $113.5B and net income of roughly $48B, putting the current market cap at about 14x PE. Of course, US equities currently have less to do with fundamentals and more to do with Trump.