3 ms·

For highly complex technologies that require tons of capex to get anywhere, Chinese firms are perfectly happy to depend on foreign companies if they are allowed

For highly complex technologies that require tons of capex to get anywhere, Chinese firms are perfectly happy to depend on foreign companies if they are allowed to. Chinese chipmakers weren't willing to use the first batch of domestic DUV machines because they simply weren't as good as foreign alternatives, and this made it doubly difficult for domestic machine manufacturers: they couldn't get revenue for selling machines, and they couldn't learn from their customers' experiences. The Chinese government, for example, is now apparently withholding Nvidia chips from being imported - despite the export ban being partially lifted from the US side - in order to incentivize local companies to use domestic NPUs instead. Similarly, on a personal level, most of the (technical) folks I know here in China are using Claude/GPT models because they want to use the best models available. But, companies like Bytedance (Doubao) cannot use such models to serve their customers, and ordinary folks also cannot access these models without workarounds. As such, the market for domestic AI models is huge, which has uplifted a lot of the current AI model providers. Ultimately, this creates a funny conundrum: sanctions mean that certain parts of the Chinese population can still access the "banned" tech - able to experiment with it and learn from it - while spurring domestic companies to seek domestic alternatives.

arjie 2mo ago
Okay, I think you’re right about some things. Coordinating a market economy is not straightforward. Individual participants will buy what is optimal for them. By embargoing them we provided a coordination mechanism. Chinese semiconductor manufacturers now had a captive market to sell into and the state was able to incentivize production without too much waste. But the thing that strikes me is that China was already trying to do this. SMIC et al were already funded by the Big Fund. And they generally succeed in industrial policy. It’s true that a GPU can’t be torn apart and replicated like a car factory can. But they were already trying and they were already attracting sufficient numbers. I suspect we couldn’t have stopped them. DUV tech like this has been already explored deeply. And the loss is in power and scale - both of which China can compensate for elsewhere. We are power constrained. They are not. If they want to run 3x DCs for every one of ours they can do it. It’s not like we don’t rent them GPUs. My friends do that. It’s entirely legal and Chinese companies don’t balk at needing to rent B300s. The whole thing isn’t resting on whether they want to capex it or opex it. It’s that they want domestic capability and we needed to trade off building them a coordination mechanism vs slowing them down. And we did this thing and it did that. In a world where we sell them top of the line GPUs and allow ASML to sell them EUV do we think that the geolock on those devices stops China when they decide to replicate it? Or do you think they succeed at industrial espionage, that their local techs get as good at using the machines as TSMC does, and that they eventually transfer whatever learnings they have to what they can produce?
jingpostmedia 2mo ago
[flagged]

arjie 2mo ago

Okay, I think you’re right about some things. Coordinating a market economy is not straightforward. Individual participants will buy what is optimal for them. By embargoing them we provided a coordination mechanism. Chinese semiconductor manufacturers now had a captive market to sell into and the state was able to incentivize production without too much waste. But the thing that strikes me is that China was already trying to do this. SMIC et al were already funded by the Big Fund. And they generally succeed in industrial policy. It’s true that a GPU can’t be torn apart and replicated like a car factory can. But they were already trying and they were already attracting sufficient numbers. I suspect we couldn’t have stopped them. DUV tech like this has been already explored deeply. And the loss is in power and scale - both of which China can compensate for elsewhere. We are power constrained. They are not. If they want to run 3x DCs for every one of ours they can do it. It’s not like we don’t rent them GPUs. My friends do that. It’s entirely legal and Chinese companies don’t balk at needing to rent B300s. The whole thing isn’t resting on whether they want to capex it or opex it. It’s that they want domestic capability and we needed to trade off building them a coordination mechanism vs slowing them down. And we did this thing and it did that. In a world where we sell them top of the line GPUs and allow ASML to sell them EUV do we think that the geolock on those devices stops China when they decide to replicate it? Or do you think they succeed at industrial espionage, that their local techs get as good at using the machines as TSMC does, and that they eventually transfer whatever learnings they have to what they can produce?

jingpostmedia 2mo ago

[flagged]