TSMC Posts Record Quarter as 33% Growth Shows the AI Chip Bottleneck Isn't Over
TSMC reported on July 16, 2026 quarterly revenue up 33%, topping $40 billion, and guides for further acceleration in Q3, with its CEO citing very strong AI capacity demand from cloud providers.
The quarter's numbers
TSMC reported second quarter 2026 results on July 16, 2026, with revenue up 33% year over year, topping $40 billion, and earnings per share of $4.31, a 77% jump from the same period a year earlier, according to Motley Fool. For the third quarter, the company guided revenue of between $44.6 billion and $45.8 billion, which would mark further acceleration if confirmed. Yahoo Finance also noted that the company's June revenue came in at roughly NT$442 billion, up 67.9% year over year for that month alone.
What the CEO said about demand
According to Motley Fool, TSMC CEO C.C. Wei described the signals coming from customers as very strong, especially from cloud providers buying manufacturing capacity to train and run AI models. The report's reading is that this demand reflects real computing capacity needs rather than mere market speculation, reinforcing the case that the AI chip production bottleneck remains far from resolved.
A packed foundry, falling memory stocks on the same day
Yahoo Finance highlighted a clear split within the semiconductor sector on the day of the announcement: while TSMC moves forward with manufacturing capacity effectively sold out, shares of memory makers such as SK Hynix fell 13% in the same session, amid concerns over memory oversupply. Nvidia and AMD depend both on TSMC to manufacture their accelerators and on HBM memory suppliers to build chips such as Blackwell and Instinct, meaning any swing in memory prices can affect the final cost of those components, even with the foundry running at full capacity.
What it signals for buyers of AI capacity
For SMBs and agencies that depend on cloud providers and AI model APIs to run automations, this kind of result acts as an indirect cost gauge. When the world's largest chipmaker reports capacity effectively sold out and guides for further acceleration next quarter, it suggests the cost of running AI infrastructure at scale will stay under pressure for longer, even with billions being poured into new capacity (TSMC itself announced an additional $100 billion expansion in Arizona). It is worth planning AI contracts and budgets on the assumption that computing capacity supply remains tight in the near term.
Analytical close
The contrast between a sold out foundry and a memory sector under price pressure shows that the AI hardware supply chain does not move as a single block: each link, from foundry to memory to assembly, has its own supply and demand dynamics. For those building products on top of this infrastructure, understanding where the real bottleneck sits today (in leading edge manufacturing capacity, not memory) helps predict more precisely where the next cost increase, or the next bit of slack, will show up.