#TSMCQ2NetProfitSurges77%

TSMC's Record-Shattering Quarter Left Markets Cold

Most investors read earnings the wrong way. They see record profits and assume the stock must surge. But markets aren't a scoreboard — they're a prediction machine. TSMC just posted its most profitable quarter ever, and the stock still slipped. That gap says a lot about where the AI cycle actually stands.

What Happened

TSMC reported Q2 2025 revenue of $30.07 billion, up 44.4% year-over-year, with gross margin at 58.6% and operating margin near 49.6%. Revenue came in above the top of guidance, driven almost entirely by AI-related demand.

Why TSMC Sits at the Center

TSMC manufactures roughly 90% of the world's most advanced chips. Nvidia, Apple, AMD, Qualcomm, Broadcom — they all send designs here. TSMC doesn't bet on which AI company wins; it profits from all of them, selling the picks and shovels of the AI gold rush.

High-Performance Computing now drives the majority of revenue, a real shift from the company's old smartphone dependence. In Q2, 3nm made up 24% of wafer revenue and 5nm another 36% — advanced nodes now account for 74% of the business. 2nm has entered commercial production. Smaller nodes mean faster chips and lower power draw, which is exactly what AI workloads need, and customers are paying a premium for it.

Why the Stock Fell Anyway

Management raised 2025 CapEx guidance to $38–42 billion, and 2026 spending is now expected at $60–64 billion — several billion above prior estimates. Markets read the higher spend as a sign that supply constraints will persist longer than hoped, and priced-to-perfection expectations meant even record results weren't enough to move the stock higher.

Three new Arizona fabs are underway (total investment reportedly near $165 billion), alongside expansion in Japan and Germany — a hedge against geopolitical concentration risk, though the bulk of advanced capacity still sits in Taiwan.

Bull Case

TSMC isn't just riding AI demand — it's capturing the value AI creates. It's raising prices while running at full utilization, which is the textbook definition of pricing power. Competitors can't easily match the technology, and customers can't easily switch. That moat compounds with every process generation.

Bear Case

Valuation is the obvious risk — the stock has to keep delivering to justify its multiple. Taiwan concentration remains a real structural exposure. The 2nm ramp carries normal early-yield risk, and if the current AI buildout turns out to be a one-time infrastructure cycle rather than a durable trend, today's CapEx could become tomorrow's excess capacity.

The Overlooked Point

TSMC is raising both prices and CapEx because it cannot meet all existing demand — that's the opposite of a bubble. Bubbles show up as excess supply and falling prices; this is undersupply with rising prices. The buyers here — Nvidia, Apple, AMD, the major cloud players — aren't speculative startups burning cash on unproven models. They're some of the most profitable companies in the world, already converting this demand into their own record earnings.

Bottom Line

TSMC is clearly a strong company — the real question is whether the price already reflects that. Short-term traders will keep reacting to margin guidance and CapEx headlines; long-term investors are watching a company quietly cementing itself as the indispensable layer of the AI era. Both views can be right at the same time.

What's your framework for telling genuine supply constraints apart from demand bubbles in tech infrastructure? Do you think TSMC's pricing power holds, or does competition eventually catch up?

Not financial advice — do your own research before trading.

Dragon Fly Official