TSMC in the AI stack

The business and what the filings show

TSMC manufactures nearly all leading AI accelerators, Nvidia’s and the custom chips alike, and packages many of them with its advanced packaging technology. It reported $121B of revenue for 2025, +38% on the year before, at a 60% gross margin and a 51% operating margin, figures it files annually on Form 20-F.

Its customers are concentrated: its largest accounted for 19% of 2025 revenue and its second-largest for 17%. TSMC describes high-performance computing, including AI, as its key growth driver. It spent $40.6B on capital projects in 2025, and its capacity plans are one of the best leading indicators of how long the industry expects the build-out to last.

TSMC wins whichever designer wins the accelerator contest, which makes it the closest thing to a neutral bet on AI chip demand. Its risks are concentration of a different kind: most leading-edge production is in Taiwan, and advanced-packaging capacity has repeatedly limited how many accelerators can ship.

How AI shows up in the numbers

AI shows up at TSMC as demand for its most advanced process nodes and its advanced packaging, which places accelerators beside their high-bandwidth memory. TSMC reports revenue by platform, and high-performance computing, which includes AI accelerators and data-center CPUs, is the one it describes as its key growth driver.

For scale, TSMC’s revenue over fiscal 2025 equals 21% of the $566B the five largest buyers spent on capital projects over their latest four quarters. Their capex rose +81% on the year before, while TSMC’s revenue grew +38%: slower than the spending, so it is capturing a smaller share of each new capex dollar, or more of its business sits outside the build-out.

TSMC is a foreign private issuer, so it files an annual report on Form 20-F rather than quarterly 10-Qs, and its interim results are furnished on Form 6-K without the same structured data. TSMC also furnishes its monthly revenue on Form 6-K, which is one of the most timely public reads on chip demand. This page therefore uses fiscal-year figures, using the US-dollar figures TSMC tags alongside its New Taiwan dollar results, and the quarterly comparisons on other company pages don’t apply. The next 20-F updates the figures here.

How it talks about AI

As a foreign private issuer, TSMC doesn’t furnish the 8-K earnings releases we read for US companies, so its annual report on Form 20-F is the place it describes AI on the record. Statements in a 20-F carry more weight than a press release or an earnings call: they are part of a filing the company signs and can be held to. The sentences below are quoted from its latest 20-F.

Read them for two things: how directly the company ties its growth to AI, and how much it is willing to quantify. A company that names AI as its main growth driver is making a claim its revenue mix can be checked against in the next filing; one that says AI’s effect is hard to measure is telling you its numbers won’t isolate it.

The figure to set these statements against is TSMC’s own growth: revenue +38% in fiscal 2025 at a 60% gross margin. If AI is driving the business as described, the next 20-F should show it in the mix of revenue by product or customer, and in margins that hold up as volumes grow.

• On AI demand: “High Performance Computing ("HPC") : Driven by data explosion and AI application innovation, HPC has become the key growth driver for our business.” 20-F, April 16, 2026
• On its customers: “A vast majority of our revenue is derived from customers who use our products in HPC (including AI applications), smartphones, IoT, automotive, and digital consumer electronics.” 20-F, April 16, 2026

The trend in the numbers

Over the 5 fiscal years shown, TSMC’s highest revenue was $121B in 2025 and its highest gross margin 60% in 2025. The latest year set the revenue record.

Year by year, TSMC’s revenue changed +29% in 2022, −4% in 2023, +25% in 2024, +38% in 2025. Growth accelerated in the latest year.

Its gross margin moved from 52% in 2021 to 60% in 2025: it keeps more of each dollar than it did. It spent $40.6B on capital projects in its latest year, 33% of revenue, and generated $72.5B from operations.

Where it sits in the stack

TSMC sits in the foundry, equipment and design tools layer of the AI stack. It is paid by the chips and memory and networking layers.

Among the 10 companies in this layer on our map, it ranks first by revenue. Its gross margin of 60% compares with 52% for the layer as a whole, a sign that it keeps more of each dollar than its neighbors. The full layer analysis is on the foundry & tools page.

Latest four quarters for 10-Q filers; latest fiscal year for 20-F filers. Companies under $100 million of revenue are left out.

TSMC reports annually on Form 20-F, so it is compared with its layer on its latest fiscal year; the US-listed companies in its layer grew +21% over their latest four quarters combined.

What has to go right, and what could go wrong

What has to go right. Demand for leading-edge chips and advanced packaging has to keep growing, whoever designs the chips, and TSMC has to keep its technology lead while expanding capacity outside Taiwan at acceptable cost.

What could go wrong. Geopolitical risk is concentrated here, as is customer concentration; a slowdown in accelerator orders would also leave new capacity underused, though TSMC has historically absorbed cycles better than the layers above it.

These describe the conditions the business depends on, as the filings show them. They are not a view on the stock.

Through the lens of Investing in AI

Rob May’s book offers six mental models for telling AI moats from traps. The one most useful for TSMC’s layer is “AI changes markets and corporate strategy”: When creation gets cheap, choke points move. Like the internet shifting newspaper profits to discovery, AI pushes moats toward coordination and control points. The winners sit at the new choke point, not the best AI story.

The question it asks of any company is: What becomes scarce in this value chain as AI makes the work abundant? Applied to TSMC, the filings on this page are the place to start: the margin trend shows whether it holds scarce value or passes it through, the customer and financing disclosures show who it depends on, and the commitments show what it has promised before the revenue arrives.

Customers and concentration

US companies must disclose any customer that accounts for 10% or more of revenue, and foreign filers usually describe their largest customers. Concentration matters because a company’s growth is only as durable as the handful of buyers behind it, and in the AI stack those buyers are often the same few companies.

• “Our largest customer in 2023, 2024 and 2025 accounted for 25%, 22% and 19% of our net revenue in the respective year.” 20-F, April 16, 2026
• “Our second largest customer in 2023, 2024 and 2025 accounted for 11%, 12%, and 17% of our net revenue in the respective year.” 20-F, April 16, 2026

Commitments and backlog

• Backlog. TSMC doesn’t report a comparable total of remaining performance obligations in its XBRL data.

What to watch

TSMC’s next annual report on Form 20-F will cover the fiscal year after the one ending December 31, 2025; foreign private issuers must file it within four months of the fiscal year end, so expect it by around April 30, 2027. Interim results arrive earlier on Form 6-K, without the structured data this page is built from.

• Monthly revenue reports and annual capex plans, the most timely reads on chip demand and on how long TSMC expects it to last.
• Advanced-packaging capacity additions, since packaging has repeatedly limited how many accelerators can ship.
• The share of revenue from high-performance computing, which shows how much of TSMC’s growth AI is driving.
• Its place in the next quarterly AI Stack Report, which updates every company on the map from the new filings.

Figures from TSMC’s SEC filings, latest period ended December 31, 2025. Other companies: all company pages.