TL;DR · 30-second read
The Short Version
- TSMC, the Taiwanese company that manufactures most of the world’s most advanced computer chips, just had its biggest sales month ever — about 60% more than the same month last year.
- The demand comes mainly from chips used for artificial intelligence, the technology behind tools like chatbots.
- The same day, TSMC’s board approved roughly $41 billion to expand its factories, plus up to $500 million each to invest in Anthropic and OpenAI, two leading artificial-intelligence companies.
- More chips coming means the next squeeze is finding enough electricity and buildings to run them.
Taiwan Semiconductor Manufacturing Company reported net revenue of approximately NT$586.89 billion for October 2026, a record month, up 14.7% from September and 59.7% from October 2025, according to a 6-K the company filed with US securities regulators on November 10. Revenue for January through October 2026 totaled NT$4,485.62 billion, up 43.3% year over year. Forbes noted the record landed days before a broad selloff in AI chip stocks.
In a separate 6-K filed the same day, TSMC’s board approved third-quarter results — NT$1,494.24 billion in revenue and NT$763.06 billion in net income — along with capital appropriations of approximately US$40.93 billion, a NT$9.0 per share quarterly dividend, a US$500 million strategic investment company, and investments of up to US$500 million each in Anthropic and OpenAI.
Executive Summary
The world’s dominant chip foundry — the company that manufactures leading-edge processors on behalf of the designers who sell them — is reporting demand that is still accelerating, not plateauing. A 59.7% year-over-year jump in a single month, on top of 43.3% growth for the year to date, is the clearest operational signal available that AI chip orders continue to outrun capacity.
The board’s same-day decisions amplify that signal. Approving roughly $40.9 billion in capital appropriations for advanced capacity, advanced packaging, fab construction and 2027 R&D is TSMC committing to keep supply expanding into 2027 and beyond. And taking stakes of up to $500 million each in Anthropic and OpenAI puts the foundry’s own balance sheet behind the AI demand it is building for.
For the infrastructure industry, the implication runs downstream: if the chip supply keeps compounding at this rate, the binding constraint on AI deployment moves to the things chips need once they ship — powered land, grid interconnections, substations, cooling and data center shells, all of which move on multi-year timelines.
The 60% Month That Undercuts the Slowdown Narrative
TSMC’s October revenue report, filed in a 6-K on November 10, shows NT$586.89 billion in monthly revenue — up 14.7% sequentially and 59.7% year over year, both to a record. Year-to-date revenue of NT$4,485.62 billion is running 43.3% ahead of 2025, which means growth is accelerating into the back of the year rather than fading. The third-quarter figures the board approved the same day underline the quality of that growth: NT$763.06 billion of net income on NT$1,494.24 billion of revenue is a net margin above 50%, extraordinary for a manufacturer.
Because TSMC fabricates chips for nearly every major AI accelerator designer, its monthly revenue is one of the few hard, audited-adjacent data points on actual AI silicon throughput — upstream of announcements, bookings and stock sentiment. Forbes framed the record against the AI chip selloff that followed days later, and that contrast is the point: equity prices moved, but the filed operational numbers show production and billing still climbing steeply through October.
A $40.9 Billion Vote, and Where the Constraint Lands Next
The board approved capital appropriations of approximately US$40,926.70 million for advanced technology capacity, advanced packaging (the capacity-constrained step where AI accelerators are assembled with high-bandwidth memory), mature and specialty capacity, fab construction, and 2027 R&D and sustaining investment. That is the mechanism behind this article’s headline claim: TSMC is converting record demand into committed future supply, which means the flow of AI chips into the market keeps growing through 2027 and beyond.
Every advanced chip that ships has to land somewhere — in a server, in a rack, in a building with megawatts of power and cooling behind it. Fab capacity can be appropriated with a board vote and built in two to three years; grid interconnections, substations, generation and large data center campuses frequently take longer. If the foundry is compounding output at 43% a year while the power and real-estate side of the equation moves on utility timelines, the queue that matters shifts from wafer allocation to energized sites. The parties affected are the ones downstream: data center developers racing to deliver powered shells, utilities and grid operators fielding interconnection requests, and the financiers underwriting both. TSMC’s filing does not say this — it reports revenue and capex — but it is the arithmetic consequence of supply growing this fast at the top of the chain.
The Foundry Buys Into Its Customers’ Customers
The least expected item in the November 10 board resolutions is the last one: TSMC approved investing up to US$500 million in Anthropic common shares at Anthropic’s per-share initial public offering price, and up to US$500 million in OpenAI preferred stock, alongside a separate US$500 million vehicle for strategic semiconductor-ecosystem investments. The Anthropic language is notable in its own right, since purchasing at an IPO price presupposes an offering — something neither the filing nor Anthropic has detailed.
Read charitably, these are small, strategic alignments: $1 billion combined is modest against a company earning NT$763 billion in a single quarter, and a foundry has an obvious interest in proximity to the model developers whose demand ultimately fills its fabs. Read skeptically, it extends a pattern investors have been debating all year — capital circulating among AI suppliers and AI buyers in ways that can blur the line between demand and financing. The filing discloses no terms, board seats, or commercial commitments attached to either stake, so the fair conclusion is narrower than either reading: TSMC is signaling it expects these two model developers to remain central to the demand it is spending $40.9 billion to serve, and the details that would distinguish alignment from entanglement have not been disclosed.
Background
Founded in 1987, Taiwan Semiconductor Manufacturing Company invented the pure-play foundry model — manufacturing chips exclusively for others rather than selling its own — and has been the world’s leading dedicated foundry since. Per its filings, TSMC deployed 305 distinct process technologies and manufactured 12,682 products for 534 customers in 2025, with operations across Asia, Europe and North America, including subsidiaries in Arizona, Washington state and Japan. Because nearly every leading-edge AI accelerator is fabricated at TSMC, its monthly revenue reports have become a de facto barometer of AI hardware demand.
Its October report arrived amid an intense market debate over whether AI infrastructure spending has outrun near-term returns — a debate conducted largely in stock prices, against which TSMC’s filed operational numbers offer a rare hard data point. Source: TSMC Posts Record October Sales, Up 60%, Days Before The AI Chip Rout — Forbes coverage of TSMC’s October 2026 revenue report. Primary sources: TSMC 6-K filed November 10, 2026 — October 2026 Revenue Report; TSMC 6-K filed November 10, 2026 — Board of Directors Meeting Resolutions.Sources