Semis 101 with Asianometry and Fabricated Knowledge
"Semiconductors are probably one of the purest transfers of information, from the thought of it into the physical world."
How do you get into chips? Doug O’Laughlin of Fabricated Knowledge and Jon Y of the Asianometry YouTube channel and newsletter run us through how they came to learn about semiconductors, and what's it like making a living by creating content about it.
We also discuss:
Why starting with something's history can help you understand how it works;
Who they talk to and what they read to understand their niches;
Keeping the YouTube algorithm happy.
ChinaTalk is a reader-supported publication. To receive new posts and support our work, consider becoming a free or paid subscriber.
Getting Into the Semiconductor World
Jordan Schneider: How, and why, did you guys get into chips?
Jon Y: It started with TSMC. I wasn't necessarily interested in technology at first. I wanted to learn about the history behind the technologies and how they came to be. A lot of the videos [at Asianometry] come from a story approach. The timeline of technology adoption from those industries over time was more interesting to me.
Jordan Schneider: What was most and least helpful for you in the process of leveling up knowledge?
Doug O’Laughlin: I feel like the context of history really makes everything more understandable. I was talking to a CTO of a semiconductor company, and he's like, “I don't understand how any of this works if I don't understand how it worked in the beginning and how we got here.”
Once upon a time, everything was understandable in a human-scale way. They did the first chips at Intel on a laptop. There were 35, or maybe a thousand, transistors; maybe you couldn't see it, but you [could] understand it. Fast forward 30 years, that process has just scaled down infinitely.
How do we understand billions of chips in tiny micro-structures with thousands of layers? You're never going to really understand it. But if you start at the beginning, you'd be like, “Hh, this used to just be this simple NPN gate, and then we just made it way smaller.” This applies to everything. I finally understood testing after I read a book about Teradyne’s history.
Like almost everything, if you start in the very beginning it's actually understandable.
For me, whenever I start a new research project into one of the verticals, I do try to start with stuff that's relevant right now, but I feel like the deeper understanding almost always comes when I start to do the history.
Jon Y: It’s like learning almost anything else: I tried to pick out the things I understand, take them, and expand. There's still so much I don't know, but it's just picking up things over time, advancing, and challenging myself. You might start with an article or blog post. Then you go to scientific articles, and then you go onto full-blown textbooks. The more that you start seeing things repeat over and over time, over and over, you get a better understanding.
Right now, I'm doing a video about the history of EUV development. Half of the terms didn't make sense to me, but over time, as I made more and more videos about this EUV system, I finally got to the point where I feel like I kind of understand how this machine works. There's no substitute for time and brute effort.
Doug O’Laughlin: If we’re going to do two [book recommendations], I would do one core history because it really helps: Makers of the Microchip or The Intel Trinity. Makers of the Microchip is better from a true-historical perspective and is closer to a textbook about what exactly happened. The Intel Trinity goes over a lot of the same things, but with a people-first approach. I would say it's not as intense [as] Fairchild, which really was the zero-to-one moment, but Intel was the 1-to-100 moment [as] the first prolific, successful company in the industry.
The other one I really like is Fabless. It's an industry overview of the way the world has come to be. Some of the predictions in the last half of the book are so obvious and so true: they said, “Oh, this advanced packaging thing is going to be a big deal.” Now, all we do and care about is advanced packaging.
Those two books, one from history and one explaining the entire fabless ecosystem, are the two foundational texts. If you want to go deeper, trying to find the most comprehensive history book about a company is always a great way to start.
Jordan Schneider: I'll make a pitch for a book that gets you from fabless to the present: Chris Miller's Chip Wars.
Engineers and Chinese Sources
Jordan Schneider: What's the value of talking to people in the industry relative to just reading stuff on the internet?
Doug O’Laughlin: On one hand, if we're talking about the day-to-day newsletter and mechanical content-making, I don't let good be the enemy of perfect. Especially from a financial perspective, which is almost everything I write from, I have a medium-level understanding I can apply to how it matters in the economics of these companies and push something out. So [for] my core content, I don't really think it's an extremely valuable part of my process. But if we're going to talk about the deeper technological primers, I read a textbook about metrology and I [didn’t] feel like I [understood] anything. So I talked to this one engineer in particular and he really, really helped set me straight. He was explaining these things that are so obvious to him — TEV, what a backscattered electron is — and I was like, “Oh, that's exactly what you mean.” That kind of conversation is very hard. The best way that I found [is to talk to] the marketing guys. Their job is communication. If you talk to an engineering guy, [he] might be the best in the entire world at his process step, but he is working in levels that are so arcane it's hard to explain without data and models.
The semiconductor industry is not image-based: there's no qualitative information anymore. We're past the minds of puny humans.
I think there's a lot of value in talking to people, especially when you want deep, foundational understanding. But for the most part, that's way past the horizon of what you need to know. I think an engineer can really make things really click, but you have to find the right person in the right space, and that's really rare.
The Semiconductor Industry Association, the voice of the US semiconductor industry, is hiring for two roles. The first is a China-focused gig where you'll be managing SIA's China policy, market research and analytical work. This position requires strong Mandarin reading skills.
The second will have you managing SIA's export control work, analyzing policy developments, coordinating with member companies on a policy agenda, and advocating in front of the federal government.
Both are looking for candidates with 2-7+ years' of experience.
For both positions, you’d be right in the vortex of the U.S. - China technology competition. These roles are both great opportunities to dive deeper into semiconductor policy surrounded by a top team.
Jordan Schneider: Jon, do you want to talk about Chinese language sources? What is your search process? Are there any particular one that you trust or think are terrible? How do you see the Chinese-language coverage ecosystem?
Jon Y: Mainland, I have no idea — it’s pretty iffy. I've tried to read several technical papers in Chinese, and it's a level of Chinese that I've never seen before.
Fortunately, a lot of the companies do their presentations in English. They'll speak in Chinese, but their slides will be in English and they'll be able to convey basic points in English. I think what you'll find is that many of the technical papers will be in English. The more interesting [issue] is Japanese. If you're trying to find technical papers for things that are being done in Japan, a lot of those will be in Japanese. For that part, Google Translate is sufficient. I think the challenge is more [in] finding those papers and making sure you're searching for the right phrases.
Doug O’Laughlin: If you actually understand the ecosystem and you're like, “these guys are the No.1 or No. 2 player, but all their financials are in Japanese,”. I don't think it really matters as much. A dog is a dog. The attributes of the step they do and the drivers make sense. If you can translate the financials, you can learn a lot about the company, but you’re kind of out of luck in terms of analysis or insight into the inventory or how this business is changing.
Oftentimes, if you talk to the competitors of the company, they'll tell you a lot about them as well. Everything is context-dependent; no one is truly doing anything alone in this industry. There's probably an upstream or downstream supplier that understands them really well, and you might have a shot at talking to them in English.
The Content Creation Dilemma
Doug O’Laughlin: I think there's actually an inverse relationship [between] things that I'm technically proud of and how well it does. I wrote one of the best things I've ever written in two days. It was super offhanded and it did amazing. Then I did this really nuanced piece about Cloudflare, Egress and data centers. A small minority really appreciated it, but 90% of [the audience] don't understand what I’m saying at all.
The hardest part about content creation is knowing your audience and writing toward them, or writing for yourself.
Writing for yourself is much more satisfying, sustainable, and true to you. I think that in the long run, it accrues back: you do pick up some really good, true fans when you do truly good work, and people see that because you’re going to do your best work when it's for yourself. But you also can't completely ignore your audience. There are times where I think [something] is going to do well because it's the topic du jour and on the zeitgeist. The stuff that I'm interested in, that’s real, meaningful, additive research, never really does that well.
Jordan Schneider: I had a very weird experience with this a few weeks back. I've been translating stuff on ChinaTalk for five years and happened to translate a Twitter thread, which was on zeitgeist. It used a lot of inflammatory language and doubled my Twitter follower count from 20,000 to 40,000. It got me more media attention than I've ever had before in my life. Government officials reached out to me afterwards. But I just translated something. It wasn't even my analysis.
I then wrote this 3000-word piece on export controls and the future of the US-China tech relationship that I'm really proud of and spent a lot of time on, and it gets 50 likes. It’s frustrating, but I think it's important mentally to remind yourself that this is why you do this in the first place. Don't turn into a junkie for the likes, retweets, or subscriber count, because the end of that game is turning into Logan Paul.
The whole point is to put out content that is grounded, thoughtful and adds to an important conversation.
Jon Y: I agree with that. Unfortunately, the tough thing about YouTube is that you get paid proportional to how many views you get. I'm really proud of this video I did about Taiwan’s 7-Elevens from the 1970s onwards and found all these great sources, but I know not many people are going to watch this. You can indulge yourself from time to time, but eventually, you have to pay the bills.
Jordan Schenider: Doing deep dives into semiconductors is not necessarily what people think about when they first think of YouTube. Jon, how have you tried to take this tricky content and turn it into something hundreds of thousands of people are interested in?
Jon Y: It's kind of crazy, don't you think? I'm a person that has somehow built an audience [by] making videos about something that most people never even directly interact with. A lot of people use YouTube as a way to learn things. I'm really fascinated by a lot of people who email me to say, “I watched your videos and chose semiconductors as a career.” It’s really fascinating, and also somewhat terrifying.
Then there's a portion of people in their fifties or sixties who reach out to me saying, “Thanks for this video about LSI Logic. I used to work there in the 80s, and thanks for helping me give some context about that.” That's also really fulfilling to me. I know that when some people watch it, they're enjoying something and revisiting a part of their life that they find important.
Why Semiconductors Matter
Jordan Schneider: If you worked in Hollywood, you can show your grandchildren the movies you were in. Are you going to show your kid a computer from 1979? It's not going to resonate. Especially because this isn't a field that gets a lot of great books [or] has a lot of good writing around it, it's important to memorialize and resurface [things], particularly since this is how you understand today — as Doug was saying!
Doug O’Laughlin: It can be really frustrating explaining to people why this matters. Oftentimes, when I explain what I do, I say “the chips in your phone.” But at the same time, these semiconductors have truly underpinned a lot of the technological progress over the last 40 years. We are talking to each other through the internet, webcams, and mics. This is all possible because of semiconductors. That’s something beautiful and profound in my opinion, and I feel very strongly about that.
What we've done with semiconductors and making information available to everyone is really core magic.
If people really understood and looked under the covers more, I think people would have that appreciation as well. A really good book about this is The Information by James Gleick, which really made me get more reverent about semiconductors. It talks about information theory, how we process information to each other, etc. Semiconductors are probably one of the purest transfers of information, from the thought of it into the physical world. That’s what keeps me coming back.
Outro music is Still Alive by Johnathan Coulton, performed by Ellen McLain:
ChinaTalk is a reader-supported publication. To receive new posts and support our work, consider becoming a free or paid subscriber.
Thanks for mentioning “Fabless” Doug. I am very proud of iteven ten years after we wrote it
Good convo, thanks guys! 💚 🥃