Wall Street Reviews TSMC Earnings: Gross Profit Margin 'Slightly Weak', Unexpected Capex Increase Boosts Confidence in AI Supply Chain

TSMC once again delivered results that exceeded expectations, providing a clear answer to recent market debates over whether the AI investment cycle has peaked.

The company not only maintained high growth in Q2 revenue and earnings but also rarely raised its full-year revenue and capital expenditure guidance significantly. It explicitly stated that investment intensity would continue to increase in the coming years, sending a strong signal that AI demand is still in a phase of rapid expansion.

For Wall Street, more important than the quarterly performance is management's judgment on future demand. With increased capital expenditure, strengthened expectations for AI accelerator growth, and continuous expansion of advanced capacity, the market has received further validation regarding the duration of the AI infrastructure prosperity cycle.

Multiple investment banks believe that TSMC's upward revision of its full-year outlook not only signifies enhanced growth momentum for the company itself but will also boost confidence across the entire AI semiconductor supply chain. Segments such as semiconductor equipment, advanced packaging, and memory are expected to continue benefiting. Although the gross margin guidance was slightly below some optimistic expectations, institutions generally believe this primarily reflects transitional pressure from the ramp-up of new processes, without changing the fundamental long-term upward logic of AI demand.

Steady Performance, Full-Year Growth Expectations Raised Again

TSMC raised its 2026 capital expenditure guidance from $52–56 billion to $60–64 billion, an increase of more than 10%. The full-year USD revenue growth guidance was also raised from "over 30%" to "over 40%," significantly higher than the market's previous consensus expectation of 35% to 40%.

Chairman C.C. Wei stated at the earnings conference that AI demand remains "extremely strong" and announced an additional $100 billion investment in Arizona, USA, bringing the cumulative investment scale there to $265 billion.

UBS analyst Crystal Hsu noted that TSMC rarely raises its full-year capital expenditure target in Q2, which further strengthens market confidence in the prosperity of the AI supply chain and implies that semiconductor equipment manufacturers are likely to continue benefiting.

Gross Margin Slightly Below Optimistic Expectations, But Profitability Remains at Historical Highs

The gross profit margin was 67.7%, slightly above the market consensus expectation of 67.6%, but below JPMorgan's previous forecast of 69.5%. Operating profit margin exceeded 60% for the first time, reaching 60.3%, a year-on-year increase of more than 10 percentage points; earnings per share were NT$27.25, approximately 10% higher than JPMorgan's forecast.

Morgan Stanley analyst Charlie Chan pointed out that the gross margin falling short of some high market expectations was mainly affected by the accelerated depreciation of the 2nm process and the dilutive effect of overseas fab ramp-ups. JPMorgan characterized this as a one-time profit margin reset but believes long-term gross margins can still stabilize above 60%.

Regarding the Q3 outlook, the company expects revenue to grow by about 12% quarter-on-quarter, falling within Morgan Stanley's previously predicted range of 10% to 15%. JPMorgan expects that the mass production of 2nm in the second half of the year will bring about 300 to 400 basis points of pressure on gross margins, but the continuing improvement in the profitability of 3nm products will partially offset this impact. Morgan Stanley noted that the company's Q3 gross margin guidance of 66% is lower than its forecast of 67.5% and also fails to meet the optimistic expectations of some buy-side institutions close to 70%.

Nevertheless, both institutions believe this will not change the company's long-term growth logic and advise investors to use the stock price volatility caused by profit margin expectation adjustments to buy on dips. JPMorgan expects that driven by the upward revision of full-year revenue guidance, the market's EPS forecast for TSMC in 2026 may be raised by 2% to 3%, with even more significant room for earnings expectation upgrades in 2027–2028.

UBS also stated that although the company's profit margin guidance is relatively conservative, the market generally believes that management prefers to prioritize customer relationships and smooth out profit margins through subsequent product mix optimization, meaning the long-term fundamentals have not changed at all.

Significant Increase in Capital Expenditure, AI Demand Visibility Extended to 2030

The most watched aspect by the market this time was the substantial increase in capital expenditure. CFO Wendell Huang stated that the company maintains high confidence in the long-term development trend of AI, and capital expenditure for the next three years will be significantly higher than that of the past three years.

UBS pointed out that TSMC has rarely raised its full-year capital expenditure in Q2 historically, so this adjustment carries a strong signaling significance. JPMorgan believes that this round of capital expenditure increase is mainly used for early equipment procurement, expanding 3nm and 5nm capacity, building new fabs, and coping with factors such as rising equipment prices.

Although the company did not disclose the specific capital expenditure scale for the next three years, management explicitly stated that capital input will further increase in the future. This means that JPMorgan's current estimates of $78 billion in capital expenditure for 2027 and $84 billion for 2028 still have room for upward revision.

The demand side also released positive signals. Management stated that cloud computing customers are continuously increasing their capital expenditures, and the compound growth trajectory of the AI accelerator business from 2024 to 2029 has already exceeded the previously forecast mid-to-high 50% growth rate.

JPMorgan believes that from the perspective of customer orders and end-user demand, new demand brought by Agentic AI is rapidly spreading, covering not only AI accelerators but also infrastructure such as CPUs, network chips, and HBM. The related prosperity cycle is expected to continue until 2029–2030.