Key Facts
Per the figures on TSMC's investor relations page: Q2 net revenue of $40.2 billion, a gross margin of 67.7% and operating margin of 60.3%, both at historical highs; per the earnings call and multiple foreign outlets, net profit grew 77.4% year over year, and the AI/HPC platform is already 66% of quarterly revenue. The outlook is equally aggressive: Q3 revenue guidance of $44.6–45.8 billion, full-year revenue growth guidance raised from the prior level to over 40%, and an additional $100 billion investment announced in Arizona, USA. Management also confirmed N3-process capacity is sold out through year-end.
Background
This report is the most direct thermometer of AI-infrastructure heat: TSMC fabricates nearly all flagship AI chips (NVIDIA, AMD, and the major cloud providers' in-house chips), and its AI revenue share going from "an important business" to "two-thirds of the empire" confirms that downstream cloud providers are still doubling down—Microsoft has said its 2026 capital expenditure is expected to reach $190 billion. The additional Arizona investment continues the route of advanced capacity migrating to US soil.
Impact Assessment
For the AI industry, two signals are worth noting: one, the "compute drought" has no short-term fix—N3 sold out through year-end means new chips launching in the first half of next year will still face tight supply, limiting room for inference cost cuts; two, TSMC's willingness to raise full-year guidance to 40%+ shows the customer orders it sees are enough to ride through the "AI bubble" debate—at least on the foundry side, the real money is still accelerating. For developers and enterprises, this can serve as a fundamentals reference for judging where GPU rental prices and cloud providers' quota policies are headed.
via: TSMC Investor Relations page (Q2 2026 Quarterly Results), AIToolsRecap and other foreign-media earnings summaries (2026-07-16/17); verified 2026-07-20