this week i'm thinking about – #14 – TSMC
Hey friends,
Welcome to the many new subscribers since last edition! Feedback and suggestions are very welcome, especially from the new folks. I've been traveling a lot recently so please pardon my absence.
I’m writing an occasional blurb about something I learned that’s broadly tech-related. If you have thoughts, I'd love to hear from you, and I'll paraphrase the best responses in the next newsletter. My goal is to start conversations with people thinking about similar topics through different lenses. If you know anyone who would be interested in the discussion, please do forward this along, send them to the archive, or have them subscribe here.
John
What I learned about: TSMC 🇹🇼🖥
Taiwan Semiconductor Manufacturing Company (TSMC) is a living counterexample to the common wisdom that successful tech platform companies are 1) software-only, and 2) run by young founders. TSMC has maintained a near monopoly on pure-play semiconductor manufacturing since pioneering the foundry model back in 1987. We now live in a world where computer hardware is abstracted away behind a nice LCD display, but let’s time travel back to when the tech industry was more electrical engineering than web design. TSMC was the first to solve what was a major cooperation challenge within the world of integrated circuit design: prior to TSMC’s founding, foundries were primarily used to manufacture their own circuit designs. In practice, this meant you couldn’t design circuits without owning a production line, or else you’d risk your manufacturer stealing your design IP. Consequently, only massive companies could design circuits because there was no such thing as a “small test run” of a new integrated circuit without building out your own clean room, production line, and specialized machinery. By solely focusing on fabrication and avoiding any chip design that would compete with their customers, TSMC became the Switzerland of chip production, and started what I’d argue is one of the least well-understood tech platform successes.
Despite being in what is perceived as a pure commodity industry, TSMC maintains 56% foundry market share, more than 5x the size of next-largest foundry. There are massive economies of scale given the extreme capital intensity, both opex and capex, which is increasing exponentially ($9B for their Fab 15 facility, $20B earmarked for their next 3nm facility, plus $10B of annual maintenance). TSMC’s neutrality enables companies like Apple to avoid chip fabrication at plants owned by arch rivals like Samsung. TMSC also operates at 35%+ net margins, higher than both Qualcomm and Apple – somewhat ironic given founder Morris Chang claims TSMC was created so that companies “don’t have to worry about the capital-intensive part of the business any more.” That said, given Apple’s scale, I wouldn’t be surprised if they start their own chip production lines to insource the highest-cost component of the iPhone.
What's impressive about TSMC is its founder: Morris Chang was 56 when he founded the company in the late 1980s, and served as CEO until he retired at 74 in 2005, leaving TSMC with a market cap of $46B. But even more impressive is in 2009 when, after revenue and margins were roughly flat for four years, Morris returned to the company where he served as CEO again for nearly a decade until he retired at age 86 (!) in June 2018. He left TSMC with a market cap of $200B.
See these links if you want to read more about:
Morris Chang, the founder
The foundry model of semiconductor production
The Economist on TSMC’s “Fab Success”
What I’m reading 📚
Alex Clayton on Zoom’s S-1, with interesting comps on growth, margins, and payback periods of SaaS companies.
Chinese surveillance and racial profiling – alarmist, but informative.
Rohan Pavuluri on conditional probability in your career.
The New Revolution in Military Affairs, depicting what defense could look like if we used 21st century technology. Written by Christian Brose of Anduril.
Uber S-1, specifically the “core platform contribution margin” waterfall on page 99.
Proposal to open-source the index of the web. Could be an interesting anchor point for discussion of anti-trust solutions for Google.
Thoughtful reflection from the founder of Gumroad.
Map wars (spoiler alert: Google is winning): Google Maps, Apple Maps
China’s Starbucks killer Luckin Coffee raised $150m
Tesla’s robotaxi network, launching in 2020 according to Elon
Feature on Masayoshi Son, with good insight into his personality and process.
What subscribers said about the last edition (Servicetitan):
Wing highlighted ChefHero as another company aggregating supply within a specific vertical. “Similar to Servicetitan, they've also managed to fly relatively under the radar by being headquartered in Toronto, which raises an interesting point about the pros and cons of starting a company and investing centrally in the Bay Area. Obviously there is an incredible density of capital and talent here, but perhaps when most people believe that the next big thing will come out of this geographical area it's time to shift attention to undervalued and underserved geographies."
Brad surfaced Ivy, a similar vertically targeted SaaS offering for interior designers. They’ve now been acquired by Houzz, which is interesting as it becomes increasingly competitive with Ivy’s customer base.
David highlighted Lango, which he says has parallels to Servicetitan but oriented towards language translators.