TSMC booked NT$511.86 billion in September revenue, a rise of more than 50% year-over-year, the strongest signal yet that AI chip demand is still accelerating rather than plateauing. The figure, reported by the world's largest contract chipmaker, extends a run of record months that began with the ChatGPT-driven surge and has yet to slow.

The number matters beyond TSMC's own ledger. September sales at the Taiwanese foundry function as a near real-time gauge of AI infrastructure spending across the technology industry, because nearly every major AI player must route its most advanced silicon through TSMC's fabs.

Who is buying the capacity?

The source describes a customer base that now spans far beyond the traditional anchor clients:

• OpenAI, scaling next-generation models
• Google, expanding AI infrastructure
• Microsoft, powering Azure AI services
• Nvidia, absorbing the most advanced manufacturing capacity for H100 and next-generation AI chips
• Apple, holding premium allocation for M-series processors

Beyond these names, enterprise software companies, cloud providers, and even traditional industries are rushing to build AI capabilities. That breadth is spreading demand across TSMC's entire product portfolio rather than concentrating it in a single product family.

What does the supply picture look like?

TSMC's 3nm and 5nm processes — the nodes that carry AI accelerators and high-performance computing chips — remain in critically short supply. Industry analysts estimate the company is operating at near-maximum capacity utilization, with waiting lists for premium manufacturing slots stretching months out.

The demand pattern has also changed in character. Rather than the cautious, incremental investments of earlier technology cycles, customers are making massive upfront commitments to lock in wafer capacity. That behavior converts TSMC's technological lead into direct pricing power, and the record revenue figures reflect it.

Why has the competitive gap widened?

The timing favors TSMC on two fronts. Samsung and Intel are both wrestling with their own advanced manufacturing challenges, which leaves the Taiwanese foundry with an effective monopoly position at the leading edge for AI-class silicon. The source characterizes TSMC as having become, in effect, the gatekeeper of AI progress — the one manufacturer able to produce the most sophisticated chips at scale.

That position explains why customers accept premium pricing. With no credible second source at equivalent nodes and utilization already near its ceiling, TSMC holds unusual leverage over allocation and price.

What comes next?

The September result positions TSMC for what could be an exceptionally strong fourth quarter. Q4 historically brings increased chip demand as companies prepare product launches and ramp production for the following year. With AI infrastructure buildouts continuing and new applications emerging across industries, the source expects TSMC to sustain its growth trajectory well into 2027 — a horizon that suggests the current demand cycle, however extreme, is still in its early innings.