TSMC set new third-quarter records for both revenue and profit, driven by surging demand for AI chips, Chosunbiz reported.

The result marks the world's largest contract chipmaker's strongest-ever Q3 performance. AI-related silicon — the processors and accelerators behind the current wave of generative-AI infrastructure buildout — pushed the foundry's top and bottom lines past every previous third-quarter benchmark.

TSMC does not sell chips under its own brand. It manufactures them on contract for fabless designers — companies that design chips but outsource fabrication — which places it at the center of the AI hardware supply chain. Every major AI accelerator program that needs advanced process nodes (the manufacturing generations that define how small and dense a chip's transistors can be) runs through TSMC's fabs.

That position explains the records. When AI demand surges, designers order more wafers. When they order more wafers at leading-edge nodes, revenue climbs and capacity utilization tightens at the same time — a combination that historically lifts margins alongside volumes.

What does the record quarter change?

For TSMC's customers, it signals that advanced-node capacity will stay tight. Fabless firms competing for AI-relevant wafer allocations face continued pressure on pricing and lead times, and smaller designers risk being squeezed by the largest buyers.

For investors, the record confirms that AI infrastructure spending has not yet cooled. Foundry results are a lagging indicator of demand — chips ordered today ship months later — so a record Q3 reflects order strength placed earlier in the year, and suggests the AI capex cycle still had legs through that period.

For competitors, including Samsung Foundry and Intel's contract manufacturing arm, TSMC's record quarter raises the bar. Rivals have spent years trying to close the gap at advanced nodes; a record quarter for the market leader implies the gap has not yet translated into meaningful share loss at the leading edge.

Why AI demand hits the foundry first

AI accelerators are among the most demanding chips to manufacture. They are large, transistor-dense, and built on the newest process technologies available — exactly the wafers that carry the highest prices in a foundry's portfolio.

That mix matters for the P&L. Advanced-node wafers command premium pricing relative to mature-node parts (older, cheaper manufacturing generations used for analog, power and microcontroller chips). A demand wave concentrated in AI therefore shifts revenue upward faster than unit shipments alone would suggest.

The record also carries weight beyond TSMC's own books. As the industry's dominant foundry, the company's quarterly performance functions as a bellwether for the whole semiconductor chain — from equipment makers like ASML to packaging suppliers and memory producers. Record foundry results tend to be read as confirmation that the AI buildout is still in its spending phase, not its digestion phase.

The backdrop

The record comes amid a broad race to expand AI computing capacity. Hyperscalers — the large cloud providers building out AI data centers — have committed to multi-year capital spending programs, and much of that spending flows through to chip designers and, ultimately, to their foundry.

TSMC has spent years and tens of billions of dollars preparing for exactly this scenario: capacity expansion at advanced nodes, new fab construction, and process leadership at each new manufacturing generation. The Q3 records are the demand side catching up to that supply-side bet.

Sustaining the momentum is the open question. AI chip demand has outpaced nearly every forecast since the current cycle began, but capital spending at this scale eventually invites scrutiny — investors watch for any sign that accelerator orders are outrunning actual AI usage and revenue. A record quarter answers that question for now. It doesn't settle it for good.

Still, the headline is simple. TSMC posted its best third quarter ever, on both revenue and profit, and AI chips did the pushing. The foundry that makes the silicon the AI industry runs on is printing records because the AI industry keeps ordering more silicon.