TSMC Q2 2026 — KEY NUMBERS
● June revenue: NT$442.68B (~$13.8B) — largest monthly revenue in TSMC history, up 67.9% YoY, up 6.2% MoM
● Q2 total: ~$39.62B (NT$1.27T) — beat top of guidance range, up 36% YoY
● H1 2026 total: NT$2.4T (~$75B) — up 35.6% vs H1 2025
● The seasonal break: June has declined MoM for four consecutive years — this is the first June sequential gain in that window
● N3 capacity: Sold out for all of 2026 — targeted by all leading AI GPUs and CPUs this year
● CoWoS packaging: Also sold out through year-end — required for HBM stacking on Nvidia AI accelerators
● Market share: 73% of global pure-foundry market in Q1 2026 (Counterpoint Research)
● Full Q2 earnings + Q3 guidance: Thursday July 17 — same day as Gemini 3.5 Pro targeted GA
Why the Seasonal Break Matters More Than the Absolute Numbers
TSMC's June revenue of NT$442.68 billion — approximately $13.8 billion — is the largest in the company's nearly four-decade history. The figure breaks a four-year seasonal pattern where June revenue had declined month-on-month, as AI chip demand overrides consumer electronics cycles. The seasonal pattern break is more significant than the absolute record. Consumer electronics — smartphones and PCs — have a predictable Q2 demand dip as the back-to-school and holiday cycle buildups haven't begun. AI chips have no equivalent seasonal cycle. Data center operators and hyperscalers don't reduce chip orders in June because of back-to-school timing. The result: AI demand is now large enough to override the consumer electronics seasonal pattern that has defined TSMC's revenue rhythm for a decade.
Sravan Kundojjala of SemiAnalysis called the numbers 'quite robust,' noting that June revenue rising month-over-month runs counter to the seasonal pattern of the past four years. 'The demand supply situation in AI is still quite tight and TSMC is sold out on N3, which is targeted by all leading AI GPU and CPUs this year,' he said. SemiAnalysis estimates TSMC is on track to exceed $40 billion in AI-related chip revenue for the full year — roughly 25% of projected total revenue.
N3 Sold Out — What That Means for Every AI Company
N3 — TSMC's 3-nanometer process node — is the manufacturing process used for Nvidia's H100, H200, and B100 AI accelerators, Apple's M-series chips, and AMD's MI300X. TSMC's N3 manufacturing lines are fully committed for all of 2026. Sold out means every wafer TSMC can produce on N3 in 2026 is already assigned to a customer. For AI companies that haven't already secured N3 capacity through long-term supply agreements, 2026 production is simply not available — they must wait for 2027 capacity or use older process nodes with lower performance per watt.
CoWoS (Chip on Wafer on Substrate) advanced packaging is also sold out through year-end. CoWoS is the packaging technology that stacks HBM memory directly on top of Nvidia GPU dies — it is required for the HBM3E integration that makes H100 and H200 accelerators perform at their rated speeds. Sold-out CoWoS means sold-out Nvidia AI accelerators, which means sold-out AI training capacity for any company that doesn't already have committed supply. The SK Hynix IPO story from last week and the TSMC revenue story this week are two sides of the same physical constraint: the world needs more AI chips than it can currently make.
The Pattern — Hardware Compounding While Models Cut Prices
The clearest pattern of 2026: the model layer keeps cutting prices to compete, while the hardware layer keeps compounding. Last week it was SK Hynix's record stock debut; this week it is TSMC's record revenue. There is a concentration risk buried in the good news. The entire AI economy now depends on a handful of fabs on one island in a geopolitically tense strait. This week's TSMC revenue report is the strongest validation yet of the AI infrastructure investment thesis: the spending that Anthropic ($47B ARR), OpenAI ($35B+ ARR), Google, and Meta are doing on compute is translating into real revenue at the semiconductor layer — not hype, not capex announcements, but booked and delivered wafer orders that produced record quarterly revenue at the world's most important chipmaker.
The concentration risk is real. A single weather event in Hsinchu postponed a routine market-moving disclosure and illustrated, without ambiguity, a structural condition that no financial model can fully price: nearly every leading AI chip made anywhere in the world passes through one company, in one country. A typhoon moved a market. That concentration — 73% of global foundry market in one location — is why Intel's US foundry push, Samsung's expansion, and TSMC's own Arizona buildout all exist. None of them will meaningfully dilute TSMC's AI chip dominance before 2028 at the earliest.
Thursday July 17 — Full Q2 Earnings and Q3 Guidance
TSMC will announce its Q2 financial results on Thursday, July 17, 2026. The revenue beat is already confirmed. The questions for Thursday are margins and forward guidance. A single percentage-point swing in gross margin moves roughly three times more profit than TSMC's entire revenue beat did this quarter. Q1 2026 gross margin was strong but TSMC flagged that Arizona fab costs (higher labour, less operational experience) would pressure margins as US production scales. Q3 guidance will tell the market whether the AI chip demand runway extends into H2 2026 or whether hyperscaler capex is beginning to moderate. CEO C.C. Wei characterised AI chip demand as 'extremely robust' in Q1 and said TSMC would be unable to fulfill US customer demand for years even with new capacity online — watch for any change in that language on Thursday.
Sources: TechTimes, QZ, MacDailyNews, Seoul Economic Daily, SemiAnalysis via CNBC, Briefs.co — July 13-15, 2026 · Related: SK Hynix +13% on Nasdaq debut → · Anthropic IPO: Freshfields hired, $510B VC record → · Qualcomm in talks to acquire Tenstorrent →