the tuesday report #1 // tech layoffs, china, tiktok knife fight, google's antitrust gambit (an unlikely cut on AI?), porn perfected, lynchian midjourney creations go viral, and the war of chips
I'm trimming my subsciptions (Taibbi, Fifth Column), because it's just too much content. I still have to work part of the day. But I'm keeping my Pirate Wires!
The replies in the chip thread were good. Your main question isn't super clear—I am interpreting it as if you're asking if we should allow China to buy advanced manufacturing tech but block anyone from buying chips from them. If you did that, the semiconductor tooling industry would be happy. China might be able to rip off the tools with more hands-on knowledge, and the actual chips produced in China could potentially surpass what TSMC can produce. Even if they couldn't be exported, that would pose economic and national security risks, especially given the possibilities of something like AI. In addition, it would reduce China's dependence on Taiwan for advanced chips, which would arguably increase the risk of conflict. The US and its allies have largely elected a strategy of staying ahead by at least a few generations of chips.
As for the news, the big 3 problems I see the West, Japan, and South Korea grappling with are:
1) How do we keep China from dominating advanced semiconductor manufacturing?
2) How do we protect our economy and national security interests from being economically beholden to China's influence (either domestically or through rattling Taiwan)?
3) How do we grow our own manufacturing base?
It is easy to conflate the different issues, as they spill into one another. There's also a lot of history. Let's start with #1 and incentives. Since rolling out the "Made in China 2025" plan in 2015, China has tried to dominate every critical aspect of semiconductor manufacturing. This is where the questions around the West shaping incentives for Chinese self-reliance rather fall apart. China wants to be self-reliant. You can choose the reason you believe they want self-reliance (CCP regime stability, nationalism, communist true believers, rugged individualism, etc.) but there was no Western incentive required. They've long aspired to this, and put about $100B on the line to make it happen.
Their plan was to buy or build everything they'd need to own every advanced manufacturing market, including semiconductors. The US government (rightly) got concerned and blocked the sale of several chip companies to Chinese firms that would harvest all their IP. They went a step further and actively blocked their participation in advanced chip manufacturing, thanks to a fortuitous quirk in the lithography market. After 2019 or so, making the highest-performing chips with the best economics requires a specific lithography tool known as EUV (extreme ultraviolet). It's only sold by ASML in the Netherlands. When it became clear this was the critical path forward, the US leaned hard on the Dutch to restrict its sale, and the Dutch did. (I know, it's amazing, the US government did something smart and effective.) Can China steal this tech? Yes, the principles are well known. Doing this is easier said than done, though, as these tools require mastering many different technologies and accessing suppliers (in the US, Europe, Japan, South Korea, and Taiwan) who have mastered theirs. Going alone here is hard. China has been trying to do this for at least 7 years with billions of dollars of grants available and have not succeeded yet.
The most recent ban is about DUV (Deep Ultraviolet). DUV was the best tool before EUV hit the scene. It works for advanced chips, but it's less economical. Of course, if China wants those advanced chips for, say, a hypersonic missile, economics may not be the priority. Again, this initiative adds to the degrees of difficulty to replicate the entire supply chain in China. The other notable initiative was to prevent China specifically from accessing the most advanced chips for AI. The feds set a technical ceiling on data transfer for AI chips (a critical bottleneck in training neural networks).
#2 and #3 are harder, because the US spent a long time not actually caring about manufacturing. From 2000 to 2020, China spent $100B on semiconductor incentives. The US spent $0 until the CHIPS Act. Intel is happy to take the feds' money, and TSMC is building a few fabs here that won't be the most leading-edge tech but will keep a few customers like Apple happy. This is already a stupid long comment, but we're playing for time and have found it easier to hit the brakes on China than to hit the gas on our own initiatives (as your boss highlights, the West seems to actually prefer stagnation and has gotten quite good at it).
Chips, and whether ‘tis better to sell to or buy from….
In 1987, Toshiba violated an international agreement NOT to sell advanced computers and machining equipment to the (then) USSR. The argument was very Hunt for Red October: with state of the art (SOTA) equipment, which required SOTA computers to drive them, it was possible to make propellers for nuclear subs that minimized or eliminated cavitation. Cavitation is what makes a sub passively trackable, i.e., without using active sonar. Why bother? Because active sonar gives away ~your~ location.
Toshiba provided the computers and a Scandinavian company, Kongsberg, the machining part of the equation. Depending on who you read, this was either already in the works in Russia (LA Times) or it wasn’t and it cost us several billion 1987 dollars—a bunch—to reacquire the technological superiority that the episode cost. The Cold War was still on, and as far as we knew, Dr. Strangelove might still have been advising the USSR’s nuclear policy. This included sub-based missiles, so knowing where they were at any given time was actually important.
So: not so good to sell the advanced stuff to enemies, even to potential enemies.
Moving right along, we now examine the question of buying them from enemies, or at least from potential enemies. Were I a patriotic chip manufacturer (I’m very patriotic, but know next to nothing about making chips), I’d embed controls in what I sell. Chips are enormously complex little guys, so hiding a bit of extraneous circuitry would be trivial. My extra circuitry would contain code that would, at minimum, cause the chip to undergo catastrophic failure when told to do so.
Why “at minimum”? Because there are potentially even better failure modes from my viewpoint. For example, a chip might reveal aspects of the code it’s running. That’s far from unrealistic. And to the degree that we know that code, we could potentially know an enemy’s strategic intent (where to send bombs) and how they plan to do it. Or we could potentially destroy a nuclear missile in flight. Or even turn it around with a “return to sender” message.
Better to sell than to buy, but only remote-able chips. I’m pretty sure this has already been done in several high profile arenas.
Ive heard David Sacks call them “excess elites”. They’re being cut from corporate but to end up where? Id like to see how many of them are single and without child. Id bet a large portion.
I wonder if people will speak up in protest of AI porn because only a human being can bring the soulful touch needed to fully pretend to be someone’s step mother.
I mentioned this in a thread on Twitter, but the only response I got was when River Page liked only the one tweet in which I expressed agreement with an aspect of his piece... the problem with the cage-free eggs story is in how it paints chickens as somehow naturally crazed cannibals as an excuse for why we need the inhumane conventional methods used by CAFOs. While I’m with you on the false promise of cage-free, as some of those large operations aren’t much better than a conventional CAFO, presenting the egg market as a choice between those two bad options is wrong. What are the cage-free chickens being fed? Why didn’t you mention pasture-raised methods at all? Read Joel Salatin’s and Michael Pollan’s books from 2006-7, or at least just check out the model of www.polyfacefarms.com. I’m a huge fan of yours, Mike Solana, but that piece was a tad off the mark.
The chips question is interesting. We’re fooling ourselves if we think they’re never going to get the tech if we don’t sell it to them. We might as well allow American companies to profit from the eventual IP theft. Companies can reinvest the export dollars into further R&D.
This brings up an interesting question: how much does China’s copycat behavior actually fuel American innovation? If your product will be copied and sold at a lower price, you have no choice but to continue to innovate.
“So, from now on, I’ll be doing both bangers and banger newsletters. You’re welcome.”
Yes. So much Yes!
I'm trimming my subsciptions (Taibbi, Fifth Column), because it's just too much content. I still have to work part of the day. But I'm keeping my Pirate Wires!
Its 3am! Are we nocturnal?
Loved this - happy Tuesday 🤠
The replies in the chip thread were good. Your main question isn't super clear—I am interpreting it as if you're asking if we should allow China to buy advanced manufacturing tech but block anyone from buying chips from them. If you did that, the semiconductor tooling industry would be happy. China might be able to rip off the tools with more hands-on knowledge, and the actual chips produced in China could potentially surpass what TSMC can produce. Even if they couldn't be exported, that would pose economic and national security risks, especially given the possibilities of something like AI. In addition, it would reduce China's dependence on Taiwan for advanced chips, which would arguably increase the risk of conflict. The US and its allies have largely elected a strategy of staying ahead by at least a few generations of chips.
As for the news, the big 3 problems I see the West, Japan, and South Korea grappling with are:
1) How do we keep China from dominating advanced semiconductor manufacturing?
2) How do we protect our economy and national security interests from being economically beholden to China's influence (either domestically or through rattling Taiwan)?
3) How do we grow our own manufacturing base?
It is easy to conflate the different issues, as they spill into one another. There's also a lot of history. Let's start with #1 and incentives. Since rolling out the "Made in China 2025" plan in 2015, China has tried to dominate every critical aspect of semiconductor manufacturing. This is where the questions around the West shaping incentives for Chinese self-reliance rather fall apart. China wants to be self-reliant. You can choose the reason you believe they want self-reliance (CCP regime stability, nationalism, communist true believers, rugged individualism, etc.) but there was no Western incentive required. They've long aspired to this, and put about $100B on the line to make it happen.
Their plan was to buy or build everything they'd need to own every advanced manufacturing market, including semiconductors. The US government (rightly) got concerned and blocked the sale of several chip companies to Chinese firms that would harvest all their IP. They went a step further and actively blocked their participation in advanced chip manufacturing, thanks to a fortuitous quirk in the lithography market. After 2019 or so, making the highest-performing chips with the best economics requires a specific lithography tool known as EUV (extreme ultraviolet). It's only sold by ASML in the Netherlands. When it became clear this was the critical path forward, the US leaned hard on the Dutch to restrict its sale, and the Dutch did. (I know, it's amazing, the US government did something smart and effective.) Can China steal this tech? Yes, the principles are well known. Doing this is easier said than done, though, as these tools require mastering many different technologies and accessing suppliers (in the US, Europe, Japan, South Korea, and Taiwan) who have mastered theirs. Going alone here is hard. China has been trying to do this for at least 7 years with billions of dollars of grants available and have not succeeded yet.
The most recent ban is about DUV (Deep Ultraviolet). DUV was the best tool before EUV hit the scene. It works for advanced chips, but it's less economical. Of course, if China wants those advanced chips for, say, a hypersonic missile, economics may not be the priority. Again, this initiative adds to the degrees of difficulty to replicate the entire supply chain in China. The other notable initiative was to prevent China specifically from accessing the most advanced chips for AI. The feds set a technical ceiling on data transfer for AI chips (a critical bottleneck in training neural networks).
#2 and #3 are harder, because the US spent a long time not actually caring about manufacturing. From 2000 to 2020, China spent $100B on semiconductor incentives. The US spent $0 until the CHIPS Act. Intel is happy to take the feds' money, and TSMC is building a few fabs here that won't be the most leading-edge tech but will keep a few customers like Apple happy. This is already a stupid long comment, but we're playing for time and have found it easier to hit the brakes on China than to hit the gas on our own initiatives (as your boss highlights, the West seems to actually prefer stagnation and has gotten quite good at it).
Chips, and whether ‘tis better to sell to or buy from….
In 1987, Toshiba violated an international agreement NOT to sell advanced computers and machining equipment to the (then) USSR. The argument was very Hunt for Red October: with state of the art (SOTA) equipment, which required SOTA computers to drive them, it was possible to make propellers for nuclear subs that minimized or eliminated cavitation. Cavitation is what makes a sub passively trackable, i.e., without using active sonar. Why bother? Because active sonar gives away ~your~ location.
https://en.wikipedia.org/wiki/Toshiba–Kongsberg_scandal
Toshiba provided the computers and a Scandinavian company, Kongsberg, the machining part of the equation. Depending on who you read, this was either already in the works in Russia (LA Times) or it wasn’t and it cost us several billion 1987 dollars—a bunch—to reacquire the technological superiority that the episode cost. The Cold War was still on, and as far as we knew, Dr. Strangelove might still have been advising the USSR’s nuclear policy. This included sub-based missiles, so knowing where they were at any given time was actually important.
So: not so good to sell the advanced stuff to enemies, even to potential enemies.
Moving right along, we now examine the question of buying them from enemies, or at least from potential enemies. Were I a patriotic chip manufacturer (I’m very patriotic, but know next to nothing about making chips), I’d embed controls in what I sell. Chips are enormously complex little guys, so hiding a bit of extraneous circuitry would be trivial. My extra circuitry would contain code that would, at minimum, cause the chip to undergo catastrophic failure when told to do so.
Why “at minimum”? Because there are potentially even better failure modes from my viewpoint. For example, a chip might reveal aspects of the code it’s running. That’s far from unrealistic. And to the degree that we know that code, we could potentially know an enemy’s strategic intent (where to send bombs) and how they plan to do it. Or we could potentially destroy a nuclear missile in flight. Or even turn it around with a “return to sender” message.
Better to sell than to buy, but only remote-able chips. I’m pretty sure this has already been done in several high profile arenas.
Ive heard David Sacks call them “excess elites”. They’re being cut from corporate but to end up where? Id like to see how many of them are single and without child. Id bet a large portion.
There’s always mom and dad to take care of.
I wonder if people will speak up in protest of AI porn because only a human being can bring the soulful touch needed to fully pretend to be someone’s step mother.
I mentioned this in a thread on Twitter, but the only response I got was when River Page liked only the one tweet in which I expressed agreement with an aspect of his piece... the problem with the cage-free eggs story is in how it paints chickens as somehow naturally crazed cannibals as an excuse for why we need the inhumane conventional methods used by CAFOs. While I’m with you on the false promise of cage-free, as some of those large operations aren’t much better than a conventional CAFO, presenting the egg market as a choice between those two bad options is wrong. What are the cage-free chickens being fed? Why didn’t you mention pasture-raised methods at all? Read Joel Salatin’s and Michael Pollan’s books from 2006-7, or at least just check out the model of www.polyfacefarms.com. I’m a huge fan of yours, Mike Solana, but that piece was a tad off the mark.
The chips question is interesting. We’re fooling ourselves if we think they’re never going to get the tech if we don’t sell it to them. We might as well allow American companies to profit from the eventual IP theft. Companies can reinvest the export dollars into further R&D.
This brings up an interesting question: how much does China’s copycat behavior actually fuel American innovation? If your product will be copied and sold at a lower price, you have no choice but to continue to innovate.
But, but, but if we don't give them our technology they might not buy our chips.......And their market is so huge.....