TSMC's 8% 2nm and 68% HPC: Two Mix Tables, No Joint Revenue Share
Two large percentages can produce a precise-looking answer that the evidence does not support. For a US investor reading TSMC's latest quarter, the useful question is which revenue population each percentage measures, and whether the filing connects the two categories.

The two reported lenses
TSMC's October 15, 2026 management report covers Q3, the quarter ended September 30. Its technology table reports 2-nanometer at 8% of wafer revenue. Its platform table reports high-performance computing, or HPC, at 68% of net revenue. The first classifies manufacturing technology; the second classifies end-market platforms. Keep both table headings attached to the numbers.
Why the tempting calculation fails
Multiplying 0.08 by 0.68 gives 0.0544, or 5.44%. That is our arithmetic, not a company-reported revenue share. The inputs use different stated denominators, so they first need a common revenue scope. Even with a common denominator, two separate category shares do not identify their overlap. You would need a technology-by-platform cross-tab, or an appropriate conditional share on the same basis.
Read the missing cell
Imagine a table whose rows are process nodes and whose columns are platforms. The cell at the 2nm row and HPC column asks how much revenue belongs to both. The two published mix tables provide separate views; they do not fill that cell. Labelling it 'not supplied by these tables' is more faithful than inserting a calculated point estimate. This schematic describes the research question, not customer-level data.
What an equity researcher can use
Keep the disclosed node share and platform share as separate observations. For any estimate of their overlap, list the revenue scope, period, and conditional allocation assumption first. Then look for additional company evidence to test that assumption. HPC is a platform label here; these percentages alone do not establish an AI-only revenue share, profit contribution, chip volume, or future growth rate.
A reusable filing check
Before combining percentages, write down four things: the reporting period, each denominator, the classification rule, and the evidence for the relationship between categories. If one item is missing, preserve that gap in your notes. A compact source table with an explicit unknown can be more useful than a polished estimate with an unstated assumption.
Research conclusion
The investable research task is to distinguish a reported fact from an allocation model before comparing companies or building a forecast. Here, the supportable statements are the two disclosed mix shares. A joint 2nm-HPC revenue percentage needs additional evidence. This analysis does not assess TSMC's valuation or recommend a trade.
Continue your research
SignalSage by MerlaTech | US Stock Research / AI Stock Analysis. English UI; some features require in-app purchases or subscriptions. iPhone + Android: https://merlatech-stock-analysis.blogspot.com/p/download-signalsage-iphone-android.html
TSMC Q3 2026 Management Report, page 2: https://investor.tsmc.com/english/encrypt/files/encrypt_file/reports/2026-10/f83780000ddc0c3b48e57a87eab6814c4ab35058/3Q26%20ManagementReport.pdf
TSMC Q3 2026 Earnings Release: https://investor.tsmc.com/english/encrypt/files/encrypt_file/reports/2026-10/526b65af7305df2dde6b9ece6812b1fdf1e49f9b/3Q26%20EarningsRelease.pdf
Original sourced diagram; AI-assisted writing. Educational research; no return promise.
Comments
Post a Comment