05 August, 2022
I rewatched Game of Thrones last week. It's still just as great and just as terrible an ending as its ever been (in case you were curious if it changed). But rewatching and rereading always results in new insights, and several things struck me while watching this time around.
First, what a miracle we live in the modern world, without lords and ladies and kings. At the end, all of the lords laugh at Sam's suggestion that they try democracy. How incredibly presumptuous of them to think that they can resist the temptation to corruption with power.
I've written before on the nature of power, and how democracy and inclusive institutions are the only stable system of government. If that suggestion comes as a surprise to you, I suggestion you give Why Nations Fail a read.
But a more interesting result of rewatching was that I began to meditate on my own job and my place in the world.
In Game of Thrones there isn't really a war for resources. Everyone just wants power. Except there is. The resource is land, and in an agricultural society, land is the most important resource for one reason: food production. It's remarkable how much agriculture is disrupted by the wars for agricultural resources in GoT. But wars for resources are always destructive of those resources.
I started thinking about the present day, and, like I mentioned, how great it is that we don't have to deal with that feudal bullshit. Because its true, and awesome, but we do have wars, and wars of resources. But what is the most important resource of the today?
Well, obviously, semiconductors.
A year ago I knew nothing about semiconductors. I still know nothing about semiconductors, but I know a hell of a lot more than the average person now.
I work as a contractor for a company that does ellipsometry, providing tools to the big semiconductor houses (TSMC, Intel, Samsung, Hynix, etc., etc.). I have never done anything more interesting in my life.
To sum up what we do: imagine a silicon wafer. Onto this is printed an integrated circuit. The process of photolithography is basically of five steps (this is highly simplified but about as good as I can explain, and understand):
- Coat the wafer in "photoresist", a chemical can be baked and then etched.
- Etch your pattern into the photoresist with a laser.
- Use a "developer" to remove the parts of the photoresist coating that you "etched."
- Lay on your material for the structures.
- Remove the photoresist
When I say this is highly simplified, I really mean this could be completely wrong, I'm not really sure, and will be looking into it more again to understand better. (This job has required more repeated learning of concepts than anything I've heretofore tried to master.) For a better introduction to photolithography, see the amazing channel of Sam Zeloof, Garage Fab Extraordinaire
Regardless, this process produces, essentially, nanotechnology. That is, structures on the order of nanometers. You can't look at this stuff with a microscope, and a laser would be too coarse and probably destroy the chip, so you have to use special techniques to figure out if what you want on the chip is what you put on the chip. So we have Rigorous Coupled Wave Analysis, and ellipsometry.
There is no way in hell I'm getting into the math and physics of this, but the principle is pretty simple:
- Shine a light at an angle at the wafer.
- Position a camera at the opposite end of the reflection (remember, light will reflect at an equal Angle of Reflection to the Angle of Incidence).
- Capture that light, and do some fancy math (RCWA) to figure out what the dimensions of the structure on the board are.
This stuff is insanely math intensive, insanely performance dependant, and insanely important. Linear Algebra, Calculus, and Machine Learning are all pretty essential parts of this job. It feels like working towards a masters degree and getting paid a lot of money to do it.
But the thing that feels truly incredible: the level of precision required in these fabs is beyond anything the world has ever had to use. This is the most advanced manufacturing in the history of the world. Hands down. And I'm a part of the tool chain (in an incredibly small way, sure) that helps reach that level of precision.
I feel like a seventeen year old kid that lied on his military application and ended up at Midway. I've never formally studied mathematics or computer science, and I'm the dumbest guy in every Teams room, but I'm actually helping and producing.
One of our products recently had a release. I contributed some code on it that lead to a performance improvement of about 14% in one particular operation. I'm far from the only one who contributed to this, but that release (which had more than just my speed improvement) lead to apps people at TSMC being able to go home earlier than they used to. Christ that feels good to know. I actually did something that mattered, even if in a small way.
Helping Build the Arsenal of Democracy
The Great Wars of the future will not be fought over oil. This is increasingly obvious as the world transitions away from fossil fuels. They will be fought with, and over, semiconductors.
We all know what war I'm talking about. Taiwan is a jewel. TSMC and Intel are the twin kings of Semiconductors, and TSMC dwarfs Intel's market cap. (Samsung and Nvidia actually lead Intel and TSMC respectively, but Samsung is mostly specialized in memory and Nvidia doesn't have fabs). If China were to invade Taiwan, I can think of four results:
- China blockades Taiwan and a huge portion of the world's semiconductors are sidelined.
- China scuttles TSMC and a huge portion of the world's semiconductors are sidelined.
- Taiwan scuttles TSMC and a huge portion of the world's semiconductors are sidelined.
- America scuttles TSMC and a huge portion of the world's semiconductors are sidelined.
Even worse is when you start thinking about the semiconductor tool supply chain. Look at this map of ASML locations. (ASML, if you're unfamiliar, is a Dutch company that produces semiconductor manufacturing machines; as far as I understand, they're the only producers of EUV machines.) Six (Six!) of their sixteen locations are in China or Taiwan. I have no idea what those locations are specialized in (its easy to imagine the Taiwanese plants being specialized in EUV) but even pretending they're fungible locations, that's almost 40% of the world's semiconductor tool chain going offline overnight in the event of an invasion of Taiwan. My God.
Imagine the 2020-22 chip shortage on steroids. Christ. It would be an economic bloodbath.
Things are going to get interesting in the next 10 years. And I have a front row seat.
The Game of Sand
The Game of Thrones was about a seat of power for controlling the most important resource in the Seven Kingdoms: the Seven Kingdoms themselves. In a post agricultural society, land isn't anywhere near as valuable. We can produce massive amounts of wealth and value on small land footprints. Imagine the most valuable plot of land in an agricultural society, and now imagine the most valuable plot of land in the post-industrial world of today. In the former, we'd say the largest plot of arable land. In the latter, we'd say somewhere in downtown Manhattan.
Silicon wafers are created with thousands of chemicals, from the humble sand to the noble neon. None of these resources come from Taiwan. But there is one element that does: know-how. Kill all the Taiwanese and you have some useless clean rooms. Leave Taiwan intact and you have semiconductors.
The most important resource today is not physical. It's a type of know-how. The know-how to make integrated circuits.
Land can be controlled by just a single man, paying a lot of dumb powerless people to control it. But Know-How can only be controlled through the cooperation of all the people that know how. If there is a more succinct demonstration of how democracy has triumphed in the modern world, I haven't found it.