The argument that hindering and even harming the broader Chinese economy should be a US policy goal is, to me, obscenely irresponsible. Why should they relax control of their economy, society and politics if we gleefully discuss our plans to cut them down to size by any means necessary at the earliest opportunity.
Also, the argument (baked into this whole interview) that the US-China economic relationship is 0 sum... ("Economically, it also overlooks the fact that strengthening Chinese competitors will harm American industry for decades. We should be consolidating the technology that drives productivity, not ceding it to a rival.") makes no sense and hasn't since Ricardo. If competition harms America, why should we sell chips to Japan or Europe? Why should we let more than one tech company exist in the US?
What actually hurts American industry is failing to expose them to competition. Case in point: General Motors.
I struggle to think how US companies with better chips, better talent and more resources couldn't develop Chinese AI companies. Just because Deepseek delivered a low cost mode once l doesn't mean that US companies can't do it too and more.
If by giving China the H200 makes us lose the AI race, then we deserve to lose. But we are not going to lose. There is a lot of breathless alarmism in this article that I think is really unwarranted.
Because we're all operating under the assumption that China's gain is somehow our loss. And "gain" here is defined as "having limited access to US chips that are a minimum one--likely two--generations behind what American companies have access to."
It is not a zero-sum game. The PLA will have plenty of compute in any event and if AI helps China become more prosperous, it is not only a good thing for the lives of its 1.4 billion people, but likely a better thing for the world than a declining China desperate to hold on to its power.
DeepSeek delivered low cost AI, an LLM specifically, because used most of the hard work already done by others (training data, etc.) in typical Chinese fashion. That’s one of the most expensive parts, and their value proposition is based on actually believing their numbers.
Regardless, it’s dumb to strengthen your primary adversary. It’s not just about winning the “AI Race”. It’s about Trump selling out the US and our national security for financial benefit. Par for the course.
I'll just leave this for you. Not saying that you should believe it, but it makes a better counterargument than I can about Deepseek. I can't think of a single major newspaper that report on it.
But regardless I don't believe the zero sum mind set that China's gain is the US's loss and that it must be the US national mission to weaken China as much as possible. What does weaken the US, in my opinion, is hubristic attitudes like yours that instinctively dismiss things that other countries did well -- particularly countries that we view as competitors and rivals -- which prevents us from strengthening ourselves by learning from those things.
I personally think the US should aggressively try to lose the AI war. I think "winning" will leave the country in a worse position, and that we are so caught up with "beating China" that we haven't taken a step back to gauge if this is even worth winning.
Selling chips to China is good, actually. It causes them to put money into US companies to purchase largely useless tech because AI doesn't work. And China has already won on automotive and energy
It is sad to see little direct criticism of Trump and cronies in this piece. Instead the focus is on Nvidia and JH. No doubt the entire US corporate tech, military and indeed every other sector should be shouting from the rooftops about the madness of advantaging the enemy. Have they been silenced by the President’s well known penchant for going after critics? Has ChinaTalk?
My two cents is that the US president really does not care about China, but cares immensely about the stock market. So, allowing NVIDIA to sell these chips should create a bit of a stock boost, and NVIDIA is so important that it will drag the rest of the market into a short-term rally.
Whether this is good policy is up for debate, but that is my best guess at the logic behind the decision.
Why don't you have an episode with Ben Thompson of Stratechery? He has a well argued post in favor of this decision. He makes his thesis clear, in my mind it's hard to deny its validity.
Sounds like the Chinese are hesitant to buy any based on this rationale, so they now may look like the less cooperative side in the international perception game that is also being played out.
I never really felt the magic of AI until I heard the Jamieson Greer anti-H200 export song at the end of this pod.
if that's not AGI i don't know what is
The argument that hindering and even harming the broader Chinese economy should be a US policy goal is, to me, obscenely irresponsible. Why should they relax control of their economy, society and politics if we gleefully discuss our plans to cut them down to size by any means necessary at the earliest opportunity.
Also, the argument (baked into this whole interview) that the US-China economic relationship is 0 sum... ("Economically, it also overlooks the fact that strengthening Chinese competitors will harm American industry for decades. We should be consolidating the technology that drives productivity, not ceding it to a rival.") makes no sense and hasn't since Ricardo. If competition harms America, why should we sell chips to Japan or Europe? Why should we let more than one tech company exist in the US?
What actually hurts American industry is failing to expose them to competition. Case in point: General Motors.
'If competition harms America, why should we sell chips to Japan or Europe?' well, those two are treaty allies, and on the other side of the ledger we've got this https://www.csis.org/analysis/parading-chinas-nuclear-arsenal-out-shadows
but fwiw i am not personally as into slowing down chinese economic growth than dmitri
I struggle to think how US companies with better chips, better talent and more resources couldn't develop Chinese AI companies. Just because Deepseek delivered a low cost mode once l doesn't mean that US companies can't do it too and more.
If by giving China the H200 makes us lose the AI race, then we deserve to lose. But we are not going to lose. There is a lot of breathless alarmism in this article that I think is really unwarranted.
'If by giving China the H200 makes us lose the AI race, then we deserve to lose' but why take the chance?
Because we're all operating under the assumption that China's gain is somehow our loss. And "gain" here is defined as "having limited access to US chips that are a minimum one--likely two--generations behind what American companies have access to."
It is not a zero-sum game. The PLA will have plenty of compute in any event and if AI helps China become more prosperous, it is not only a good thing for the lives of its 1.4 billion people, but likely a better thing for the world than a declining China desperate to hold on to its power.
And, did I mention that the Chinese will literally be paying both Nvidia and the US Government*billions*?
DeepSeek delivered low cost AI, an LLM specifically, because used most of the hard work already done by others (training data, etc.) in typical Chinese fashion. That’s one of the most expensive parts, and their value proposition is based on actually believing their numbers.
Regardless, it’s dumb to strengthen your primary adversary. It’s not just about winning the “AI Race”. It’s about Trump selling out the US and our national security for financial benefit. Par for the course.
I'll just leave this for you. Not saying that you should believe it, but it makes a better counterargument than I can about Deepseek. I can't think of a single major newspaper that report on it.
https://www.nature.com/articles/s41586-025-09422-z
But regardless I don't believe the zero sum mind set that China's gain is the US's loss and that it must be the US national mission to weaken China as much as possible. What does weaken the US, in my opinion, is hubristic attitudes like yours that instinctively dismiss things that other countries did well -- particularly countries that we view as competitors and rivals -- which prevents us from strengthening ourselves by learning from those things.
I personally think the US should aggressively try to lose the AI war. I think "winning" will leave the country in a worse position, and that we are so caught up with "beating China" that we haven't taken a step back to gauge if this is even worth winning.
Selling chips to China is good, actually. It causes them to put money into US companies to purchase largely useless tech because AI doesn't work. And China has already won on automotive and energy
It is sad to see little direct criticism of Trump and cronies in this piece. Instead the focus is on Nvidia and JH. No doubt the entire US corporate tech, military and indeed every other sector should be shouting from the rooftops about the madness of advantaging the enemy. Have they been silenced by the President’s well known penchant for going after critics? Has ChinaTalk?
lol not sure how much louder we can be shouting but appreciate the note
Arghh! I knew I shouldn’t have said that. There goes my US trip:
US could ask foreign tourists for five-year social media history before entry https://www.bbc.com/news/articles/c1dz0g2ykpeo
My two cents is that the US president really does not care about China, but cares immensely about the stock market. So, allowing NVIDIA to sell these chips should create a bit of a stock boost, and NVIDIA is so important that it will drag the rest of the market into a short-term rally.
Whether this is good policy is up for debate, but that is my best guess at the logic behind the decision.
Why don't you have an episode with Ben Thompson of Stratechery? He has a well argued post in favor of this decision. He makes his thesis clear, in my mind it's hard to deny its validity.
fwiw i think he's not totally off base but that MAD argument is truly wild
he's got an open invitation!
Is there any chance these chips will have backdoor access built into them that allows the US to better monitor the Chinese?
that's what the MSS thinks!
Sounds like the Chinese are hesitant to buy any based on this rationale, so they now may look like the less cooperative side in the international perception game that is also being played out.