In this report
February 2026 Revenue — A Record in Disguise
TSMC posted NT$317.66 billion (roughly $10 billion USD) in February revenue — a 22.2% jump year-over-year.1 On the surface, the month-over-month drop of 20.8% looks alarming — but that's entirely seasonal: the Lunar New Year holiday simply means fewer working days in February.2
Look at January and February together and the story is compelling: NT$718.91 billion ($22.6–22.9 billion USD), up roughly 30% versus the same period in 2025. That two-month run rate places TSMC squarely on track to meet its own first-quarter guidance.3
Plain English: Despite a headline dip in monthly revenue, the underlying trend is strong. Think of it like comparing a snowy February to December — the calendar, not the business, explains the drop.
| Metric | Figure | Change | Note |
|---|---|---|---|
| February 2026 Revenue | NT$317.66B (~$10B USD) | +22.2% YoY | Record February result1 |
| Month-over-Month Change | −20.8% | vs. January | Fewer working days (Lunar New Year)2 |
| Jan–Feb 2026 Combined | NT$718.91B (~$22.7B USD) | ~+30% YoY | On-track for Q1 guidance3 |
| 2026 Capex Plan | USD $52–56B | +27% YoY | Signals long-term AI confidence4 |
| Price Hikes (Leading Edge) | 3–10% annually | Guided through 2029 | Reflects pricing power4 |
Revenue at a Glance
The charts below put the numbers in visual context. Notice how the Feb dip is clearly seasonal — the year-over-year comparison (orange) tells the real story.
What's Driving the Numbers: AI, Everywhere
The short answer is AI. TSMC manufactures the chips that power virtually every major AI system — from Nvidia's H100/B200 accelerators to Apple's latest iPhones. Demand for its most advanced 3-nanometer process (and soon 2nm) is running at full capacity.5
Hyperscalers — think Microsoft, Google, Amazon, and Meta — are pouring tens of billions into AI data centers. Those data centers require vast quantities of AI accelerator chips. Those chips are made almost exclusively by TSMC.6
Key Growth Drivers
Long-Range Targets4
- ~25% overall revenue CAGR through 2029
- Mid-to-high 50% CAGR for AI accelerator sales through 2029
- $52–56B capex in 2026 alone — investing ahead of demand
- Annual price increases of 3–10% on leading-edge processes
- Global fab expansion (U.S., Japan, Europe) underway
What this means for investors: TSMC isn't just growing — it's compounding. A 25% revenue CAGR over five years would roughly triple annual revenue by 2029. The AI accelerator slice is growing even faster. The question isn't whether demand is there; it's whether supply chains and infrastructure can keep up.
The Risk Beneath the Surface: Energy & Chokepoints
Here's a risk that doesn't show up in most analyst reports: TSMC's extraordinary growth depends on an uninterrupted supply of energy — and that energy supply has a geographic Achilles' heel.
Taiwan's Energy Vulnerability
Taiwan produces over 60% of global semiconductors and more than 90% of advanced chips smaller than 7 nanometers.7 It imports approximately 98% of its energy, with LNG (liquefied natural gas) accounting for roughly 40% of electricity generation.8
Critical fact: Taiwan's LNG storage covers only 9–12 days of net imports on the island, with another 3–5 weeks' supply at sea.8 TSMC alone consumes an estimated 9–10% of Taiwan's total electricity — equivalent to 1.6 million households.9
The Strait of Hormuz Connection
The Strait of Hormuz — a narrow waterway between Iran and the Arabian Peninsula — handles approximately 21% of global petroleum liquids and 20% of global LNG exports.10 Any significant disruption there ripples directly into the economics of Taiwanese chip manufacturing.
The Hidden Input: Sulfur
Here's one most investors don't know: semiconductor manufacturing requires large quantities of sulfuric acid for wafer cleaning and copper processing. Over 90% of global sulfur — the raw material for sulfuric acid — is produced as a by-product of oil refining.11 An oil supply disruption doesn't just affect fuel — it can tighten the supply of a critical manufacturing chemical.
Historical data suggests a 10% decline in global sulfur recovery could raise sulfuric acid benchmark prices by 15–25%.11
Risk Concentration at a Glance
What History Tells Us: Oil Shocks & Chip Stocks
Energy shocks and semiconductor stocks have a documented, uncomfortable relationship. Two recent examples stand out:12
| Period | Oil Move | SOX Index Move | Timeframe |
|---|---|---|---|
| 2008 Oil Shock | +45% YoY (WTI) | −34% | Apr – Oct 200812 |
| 2021–2022 Energy Crisis | +75% (Brent) | −29% | 2021–202212 |
Rule of thumb: Every +10% rise in industrial energy costs translates to roughly −2 to −3% contraction in semiconductor operating margins, based on Goldman Sachs modeling of fab cost structures.12 Semiconductors historically exhibit a −0.6 beta to oil during 3-month price shock windows.
What This Means for Your Portfolio
TSMC is a remarkable business — but investors who own it (or AI/semiconductor ETFs) are also implicitly taking on energy and geopolitical risk. Here's a practical framework:
Short-Term Watch
A disruption that reduces LNG flows through the Strait by as little as 10% could add 5–7% to regional electricity prices in Taiwan and elsewhere in Asia, translating to 20–30 basis points of margin compression for energy-intensive fabs.13
Medium-Term Structural Shift
Prolonged energy volatility is already accelerating fab diversification — TSMC is building in Arizona, Japan, and Germany. That's good for resilience, but new fab construction in the U.S. or Europe runs 30–50% more expensive than equivalent capacity in Taiwan.13 That cost gap is a long-term drag on returns unless offset by government subsidies and pricing power.
Five Practical Steps for Investors
- Check regional power mix when evaluating chip companies — those in nuclear or high-renewable grids face less energy price volatility than LNG-dependent fabs.14
- Value power purchase agreements (PPAs) — companies with long-term, fixed-price power contracts have partially insulated margins from fossil fuel swings.14
- Be cautious about timeline assumptions for new AI data centers and fabs in grid-constrained regions — permitting and interconnect delays are common and often underestimated.15
- Consider diversifying into power efficiency plays — GaN and SiC power semiconductor firms (e.g., companies like Wolfspeed, onsemi) benefit as customers chase energy efficiency gains.15
- Run stress tests: model what a 10–20% spike in Asian industrial power prices does to your portfolio's semiconductor exposure.13
Bottom line: TSMC's AI-driven growth story is real and well-supported by data. But owning that growth means owning exposure to energy geography. The best-positioned investors will treat energy security as a first-class variable — not background noise — when sizing semiconductor and AI infrastructure positions.