In a memo outlining his predictions for China and AI, Bernstein said that app development, use cases, and the race to build AI-based platforms will continue to proliferate. The Biden administration's restrictions on what types of AI chips companies can export to China could pose some setbacks to China's ability to compete in the AI race.
Bernstein Managing Director and Senior Analyst Stacy Rasgon joins Yahoo Finance to discuss the AI race in China and how it stacks up against American companies.
When asked about the chip gap between China and America and what they must do to catch up, Ragson explained that there is a huge gap and said, “What does that mean? It means you have to scrape together more chips. For example, the Huawei Ascend is probably about the same as the Nvidia A100, and that part from Nvidia is 2-3 years old… If you want to get the same amount of performance that you're offering in a part, you have to put more silicon in place to get that, so the cost of the chip, the power, the infrastructure, everything else. What if they have to sell it elsewhere? Absolutely not, it's not competitive, but they have no choice. , will do so.”
For more expert insights and the latest market trends, click here to watch the full episode of Yahoo Finance Live.
Editor's note: This article was written by Nicholas Jacobino
video transcript
Julie Hyman: Let's move on to China for a moment. You and your team went out with a memo about some predictions for AI in China. Those are kind of specific, so I'd like to zoom out a little bit and talk about China versus the United States and where we are with that development. Can you talk about that and where we are and who stands to win? Or maybe it's not zero sum here?
Stacey Rathgon: Well, there are export regulations. Now, responsibility has been placed on American companies that sell AI components to China. Mainly he affects Nvidia. Most other revenues are so small that they don't have much of an impact on other companies. NVIDIA had a considerable amount of business in China. And due to current export restrictions, NVIDIA is essentially unable to sell these parts in China at all.
They are working on developing new parts that can be sold legally. However, given the performance thresholds around these sanctions, performance must be significantly reduced. And indeed, there are domestic Chinese alternatives from companies like Huawei that, at least on paper, perform better than the products NVIDIA is allowed to sell into China.
Nvidia delivers on software compatibility. And like a lot of the existing AI infrastructure and models that are being built in China, already he's built on Nvidia hardware, so it leverages Nvidia, the Nvidia software ecosystem. So using Nvidia parts allows for seamless integration with it. So I think these are probably more suitable for more Tier 2 developers. Large Chinese companies may use other alternatives if they have the resources to port everything.
But I would say in the long term, I have some view on this – my view is that in the long term, after they imposed sanctions a year ago, the American I think China's TAM for is going to be impaired because they have lowered the sanctions threshold. But even if they don't, they will never be brought up. So the gap between what is legitimate in terms of performance for U.S. companies to sell to China and the rest of the world will likely only get wider over time. So I think that as much as TAM is potentially faulty, it should ideally be considered.
Now, I think the overall long-term effects of TAM are huge. I think this is a big deal regardless of whether China is present or not. So I think it's easy to handle. But you have to. By the way, I'll leave aside the whole question of whether this was the right thing to do. So I can understand the national security implications of wanting to limit China's ability to enhance this. But at the same time, it's not just AI, it's AI and its equipment and everything else, and we're forcing China to be creative in every sense of the word.
Julie Hyman: Yeah, that's what I was going to ask you, Stacey. So how far will Chinese chip developers go before he rivals Nvidia? You say Huawei's chips are said to be able to do the same things that lower-performance chips can do. Ta. But I mean, it seems like there's still a pretty big gap between domestically produced chips and the chips that Nvidia makes elsewhere.
Stacey Rathgon: that's right. So what does it mean? That means you have to throw more chips at the problem. For example, Huawei's Ascend is probably about on par with his Nvidia A100, and that part of Nvidia predates him by 2-3 years. On a performance basis, they are about the same. Again, it doesn't have the software ecosystem and all that. But if you just want to receive it, that's fine.
So if you want to get the same amount of performance that companies like Nvidia are offering in their core parts, you're going to have to put a lot more silicon in place to do that. That means it costs money in terms of chip costs. There are huge costs involved in everything from electricity to infrastructure. So, is it cost-competitive as it would be if the product had to be sold elsewhere in the world? Absolutely not. It's not competitive.
But they have no choice. So they'll do it. The same thing happened with smartphones. Huawei is making smartphone chips again, presumably using the SMIC 7-nanometer process, but this won't be able to compete with what Qualcomm and other major suppliers can offer. And they will consume even more power. The cost will be even higher. But again, if you don't have a choice, you will. And over time, it tries to find ways to optimize within the constraints. Again, they need to get creative.
We don't know what this will look like in 10 or 15 years. But it will become even more difficult. And China in general is a big market for US and other suppliers. It's going to get tougher. It's going to get tougher. However, we don't know what this will look like when it is rolled out.
Josh Lipton: Okay. I will continue to monitor it. Dr. Stacey Rasgon, thank you for joining me on the show, friend. I love having you around always.
Stacey Rathgon: you're welcome.
