China Can Already Train a Model Like Mythos – Jensen Huang

Added: 2026-05-05

[00:00:00]If Chinese companies and Chinese labs and the Chinese government had access to the AI chips to train a model like Claude Methos with these cyber offensive capabilities and run millions of instances of it with more compute, the question is, oh, is that a threat to American companies, to American national security?

[00:00:19]First of all, Methos was was trained on fairly mundane capacity.

[00:00:26]And a fairly mundane amount of it.

[00:00:29]By an extraordinary company.

[00:00:32]And so the amount of capacity and the type of compute that it was trained on is abundantly available in China.

[00:00:40]And so you just have to first realize that chips exist in China. They manufacture 60% of the world's mainstream chips, maybe more.

[00:00:51]It's a very large industry for them.

[00:00:55]They have some of the world's greatest computer scientists.

[00:00:58]As you know, most of the AI researchers in all of these AI labs, most of them are Chinese.

[00:01:05]They have 50% of the world's AI researchers.

[00:01:11]And so the question is, if you're concerned about them, what is the considering all the assets they already have? They have an abundance of energy.

[00:01:22]They have plenty of chips. They got most of the AI researchers.

[00:01:27]If you're worried about them, what is the best way to create a safe world? Well, victimizing them, um turning them into an enemy likely isn't the best answer.

[00:01:44]They are an adversary.

[00:01:46]We want United States to win.

[00:01:48]But I think having a having a dialogue and having research dialogue is probably the safest thing to do.

[00:01:55]This is an area that that is glaringly missing because of our current attitude about China as an adversary.

[00:02:05]It is essential that our AI researchers and their AI researchers are actually talking.

[00:02:10]It is essential that we try to both agree on how to what not to use the AI for.

[00:02:18]With respect to the finding bugs in software, of course, that's what AI is supposed to do.

[00:02:25]Is it going to find bugs in a lot of software? Of course.

[00:02:28]There's lots and lots of bugs. There's lots of bugs in the AI software.

[00:02:34]And so that's what AI is supposed to do. And I'm delighted that that AI has reached a level where it could help us be so much more productive.

[00:02:44]One of the things that that um is is uh under under emphasized is the richness of ecosystem around cybersecurity, AI cybersecurity, and AI security, and AI privacy, and AI safety.

[00:03:02]That whole ecosystem of AI startups that are trying to create this future for us where where you have one AI agent that's incredible surrounded by thousands of AI agents keeping it safe, keeping it secure.

[00:03:19]That future surely is going to happen.

[00:03:22]And the idea that you're going to have an AI agent running around with nobody watching after it is kind of insane. And so we know very well that this ecosystem needs to thrive.

[00:03:34]It turns out this ecosystem needs open source.

[00:03:37]This ecosystem needs open models. They need open stacks so that all of these AI researchers and all these great computer scientists can go build AI systems that has are as formidable and can keep AI safe.

[00:03:52]And uh and and and so one of the things that we need to make sure that we do is we keep the the open source ecosystem vibrant.

[00:04:01]And um and that can't be ignored. That can't be ignored and and a lot of that is coming out of China.

[00:04:08]Um we we added we added not suffocate that.

[00:04:14]You know, with respect to to China, we want to have of course we want United States to have as much computing as possible.

[00:04:20]All right.

[00:04:22]What we're limited by energy.

[00:04:24]Um but you know, we got a lot of people working on that and we we added not make energy a a bottleneck for our country.

[00:04:32]Um but what we also want is we want to make sure that all the AI developers in the world are developing on the American tech stack and making the contributions, the advancements of AI especially when it's open source available to the American ecosystem. And it would be extremely foolish to create two ecosystems. The open source ecosystem and it only runs on the Chinese tech tech a foreign tech stack and a closed ecosystem and that runs on the American tech stack. I think that that would be that would be a horrible outcome for the United States. Mhm.

[00:05:08]Since there are a lot of things, let me just triage the um response. I mean, I think the concern going back to that flop difference in the hacking is yes, they have compute, but there's some estimates that because they're at 7 nanometer, they don't have EUVs because of chip making export controls, the amount of flops they're able to actually produce, they have like 1/10 the amount of flops that the US has. And so with that, could they train eventually a model like Methos? Yes, but the question is because we have more flops, American labs are able to get to these level of capabilities first and because Anthropic got to it first, they say, okay, we're going to hold on to it for a month while all these American companies we give them access to it, they're going to patch up all their vulnerabilities and now we release it. Furthermore, if they even if they trained a model like this, the ability to deploy at scale, you know, if you had a cyber hacker, it's much more dangerous if they have a million of them versus a thousand of them. So that inference compute really matters a lot. And in fact, the fact that they have so many AI researchers who are so good is the thing that makes it so scary because what is it that makes those engineer researchers more productive? Is compute.

[00:06:16]If you talk to any AI lab in America, they say the thing that's bottlenecking them is compute. So and there are quotes from DC founder coin leadership or whatever, they say like the thing we're bottlenecked on is compute. Um so then the question is, isn't it better that we get to get American companies because they have more compute get to get get to the level of Spot or Methos level capabilities first, prepare our society for it before China can get to it because they have less compute?

[00:06:42]We should always be first and we should always have more.

[00:06:46]But in order for that outcome for you to to what you described to be true, you have to take it to the extremes. They have to have no compute.

[00:06:55]And um and if they have some compute, the question is how much is needed?

[00:07:01]The amount of compute they have in China is enormous.

[00:07:06]I mean, you're talking about the country is the second largest computing market in the world.

[00:07:11]If they want to deploy, aggregate their compute, they got plenty of compute to aggregate.

[00:07:17]But is that true? I mean, there's like people do these estimates and they're like, well, SMIC is actually behind on the process nodes and they're they actually >> I'm about to tell you.

[00:07:23]>> Okay. The amount of energy they have is incredible, isn't that right?

[00:07:27]AI is a parallel computing problem, isn't it?

[00:07:31]Why can't they just put four, 10 times as much chips together because energy's free? They have so much energy. They have data centers that are sitting completely empty, fully powered.

[00:07:43]They've you know, they have ghost cities, they have ghost ghost data centers. They have so much capacity of infrastructure.

[00:07:50]If they wanted to, they just gang up more chips even though they're 7 nanometer. And their capacity of building chips is one of the largest in the world. The semiconductor industry knows that they monopolize mainstream chips.

[00:08:04]They they overcapacity, they have too much capacity.

[00:08:07]And so the idea that China won't be able to have AI chips is completely nonsense.

[00:08:13]Now, of course, if you ask me, um uh would would would a United States be be further ahead if the entire world had no compute at all? But that's just not an outcome. That's not a scenario that's true.

[00:08:28]They have plenty of compute already. The amount of threshold they need for the for the concern you're worried about, they've already reached that threshold and beyond.

[00:08:36]And so so I think the you misunderstand that AI is a five-layer cake.

[00:08:42]And at the lowest layer layer is energy.

[00:08:45]When you have abundant of energy, it makes up for chips. If you have abundance of of chips, it makes up for energy. For example, United States is scarce on energy.

[00:08:57]Which is the reason why Nvidia has to keep advancing our architecture and do this extreme co-design so that with the few chips that we ship, okay, with the few chips because the amount of energy is so limited, our throughput per watt is off the charts.

[00:09:13]But if your amount of watts is completely abundant, it's free, what do you care about performance per watt for?

[00:09:21]You got plenty you can use old chips to do so. So 700 meter 7 nanometer chips are essentially Hopper.

[00:09:29]The ability to for Hopper, um I got to tell you today's models are largely trained on Hopper.

[00:09:38]Yeah, Hopper generation. And so so Hopper 7 nanometer chips are plenty good. The abundance of energy is their advantage. But then there's a question of, okay, well, can they actually manufacture enough chips given their >> But they do. Uh What's what's the evidence? Huawei just had the largest single year in the history of the company. How many chips did they ship? A ton. Millions.

[00:10:02]Millions is way more way more than Anthropic has.

[00:10:07]So, there's a question of how much logic SMIC can ship, then there's a question of how much memory >> you what it is. They have plenty of They have plenty of logic and they have plenty of HBM2 memory. Right, but as as you know, the bottleneck often in training and doing inference on these models is the amount of bandwidth. So, if you HBM2 I don't have the numbers off hand, but like versus the newest thing you have, you know, you you can be almost an order of magnitude difference in memory bandwidth, which is Huawei's a networking company.

[00:10:34]Huawei's a networking company. But that doesn't change the fact that you need a EUV for the most advanced HBM.

[00:10:39]>> Not true.

[00:10:40]Not at all true.

[00:10:42]You could gang them together just like we gang them together with NVLink 72.

[00:10:47]They've already demonstrated silicon photonics sub- connecting all of these compute together into one giant supercomputer.

[00:10:54]That Your your premise is just wrong.

[00:10:58]The fact of the matter is their AI AI development is going just fine.

[00:11:01]And and the best AI researchers in the world because they are limited in compute, they also come up with extremely smart algorithms. Remember I just what I said.

[00:11:14]I said that Moore's law is advancing about 25% per year.

[00:11:18]However, through great computer science, we could still improve algorithm performance by 10x.

[00:11:25]What I'm saying is great computer science is where the lever is.

[00:11:31]There is no question. MOE is a great invention. There's no question all of incredible attention mechanisms reduce the amount of compute.

[00:11:42]We have got to acknowledge that most of the advance advances in AI came out of algorithm advances, not just the raw hardware.

[00:11:52]Now, if most advances came from algorithms and computer science and programming, tell me that their army of AI researchers is not their fundamental advantage, and we see it. DeepSeek is not inconsequential advance.

[00:12:08]And the day that DeepSeek comes out on Huawei first, that is a horrible outcome for our nation. If you enjoyed this clip, you can watch the full episode here and subscribe for more clips. Thanks.