BofA Raises TSMC Price Target to $620 as 2nm Ramp-Up Boosts Earnings Outlook
nashnova research
Bank of America raised its TSMC ADR target from $590 to $620, driven by a faster-than-expected 2nm ramp — this means TSMC is converting its technology lead into profit delivery at an accelerating pace.
Why raise the target right before earnings?
Analyst Haas Liu lifted 2026–2028 EPS estimates by 2% to 9% and maintained a Buy rating.
The core logic: 2nm — TSMC's latest process node, with circuit widths of just two nanometers — jumped from 3% of revenue in Q2 to roughly 11% in Q3.
This means → 2nm is no longer a lab-stage technology reserve; it is contributing revenue at scale, ramping faster than the Street expected.
Demand came from two anchors: Apple mobile SoC inventory builds and AMD Venice server CPU volume growth.
How did Q3 revenue overshoot TSMC's own guidance?
BofA estimates Q3 revenue grew 18% quarter-on-quarter, above the Street's 13% expectation.
In dollar terms, Q3 revenue likely reached $46.67 billion, exceeding the company's $44.6–$45.8 billion guidance range.
In plain terms = TSMC drew the roadmap, then delivered results bigger than its own sketch.
Beyond the 2nm ramp, multiple GPU and ASIC products — chips custom-built for specific tasks — migrated from 5nm to 3nm, lifting advanced-node shipments across the board.
What is the biggest uncertainty for Q4?
Revenue looks solid: advanced nodes (5nm and below) are expected to rise from 71% to 75% of revenue; 2nm could reach 18%, priced at a 15%-plus premium over 3nm wafers.
But gross margin is where the Street disagrees: BofA forecasts Q4 margin at roughly 67.8%, below consensus of 68%–69%.
This means → top-line growth is not the issue — profitability may temporarily lag. Two drags: overseas fab expansion pushing depreciation up 25%–30% sequentially, and early-stage 2nm yield ramp diluting margins by 3 to 4 percentage points.
Put simply = earning more but spending more too — burning cash on new lines is a necessary phase.
How much capex, and who foots the bill?
BofA projects 2026 capex at roughly $63 billion and a further jump to about $83 billion in 2027 — a three-year CAGR of approximately 20%.
Where the money goes: 3nm expansion to 55,000–60,000 wafers/month, 2nm ramp to 60,000 wafers/month, and initial A16 capacity installation.
Equipment vendors plan 10%–15% price hikes starting in 2027. BofA believes TSMC can pass these through via wafer price increases — projecting a 5%–10% advanced-node price hike in Q1 2027.
This reflects TSMC's pricing power: upstream costs get pushed downstream because customers have no alternative.
How wide is TSMC's technology moat, really?
Measured by transistor density — how many transistors fit per unit area, a direct proxy for performance — TSMC's N3/N3E leads Intel 3/18A by roughly 30%.
Looking ahead, the A16/A12 nodes are expected to widen that gap to 50%–60% over Intel 18A-P/14A.
The market worries about Terafab — the chip fab Elon Musk's xAI plans to build. BofA sees the risk as overstated: based on a $119 billion ten-year plan, Terafab could build only about 96,000 wafers/month of advanced capacity by 2036, far below TSMC's scale.
In plain terms = Musk's fab looks more like a strategic posture than a real capacity substitute. BofA actually thinks SpaceX could become a key TSMC anchor customer, consuming roughly 20% of TSMC's advanced capacity by 2029.
Is the stock expensive at current levels?
BofA's target is based on 20x estimated 2027 P/E — a ratio that prices today's shares against future earnings.
TSMC currently trades at roughly 16x 2027 earnings, sitting in the middle of its historical 11x–21x range.
This means → under BofA's framework, the valuation is "not demanding" — not cheap, but nowhere near bubble territory.
TSMC's ADR has rallied more than 48% year-to-date. Whether Q4 gross margin meets the Street's expectations will be the key validation point in this earnings release.
市场有风险，内容仅供研究参考，不构成投资建议。