Speaker 1: All right, a look at U.S. futures. Yes, in case you're just joining us right now and wondering why some of these contracts are down 2%, 1% on S&P futures. Yeah, well, you got to blame the, is it the bandwagoning? Is that the verb there on the deep-seek trade? So the opposite of this is obviously a lot of the related stocks in China are actually moving to the upside here. And might this give you something to think about as far as stretch valuations in the U.S.? And is there

Speaker 2: an alternative? I think that's the question here. Seeking deep meaning, going deep into this story, investors are too trying to figure this out. Is it a trend or is this, I mean, a short-term trend, or is it something a little deeper there to say? All right, over the past few days, a Chinese AI company has been making waves in Silicon Valley. And my pen just dropped onto the floor. I'll grab it for you. But yeah, go ahead. Keep reading. That's because AI models from deep-seek have been scoring impressively on global benchmarks. The open-source large-language model was unveiled in December. The lab behind it says it took two months and under $6 million to build. I think that's a key point right there. That's a fraction of what OpenAI and Google spent to train their models. Dave?

Speaker 1: Yeah, $6 million.

Speaker 2: How much has OpenAI and Meta and Google and Tesla spent combined?

Speaker 1: Well, I know of more expensive apartments in Hong Kong than the budget it took to get this up. Last week, the founder of deep-seek was the only AI leader selected to attend a meeting of entrepreneurs with Chinese Premier Li Chang. And over the weekend, Xinhua published this editorial on how China's AI advances are working despite Washington's tech curbs on the country. Robert Lee is with us, our Bloomberg Intelligence senior analyst. So, you told us so.

Speaker 3: Well, we've been writing on this company for the best part of six or seven months. And I think for anybody who follows the China AI sector closely, as we do, the profile has been relatively high within this part of the world. But what you've seen over the last week or so, you know, prompted by the coverage, and as you said, the meeting with the premier, it's helped raise the company's profile both nationally and also externally. And it's been, this story's been picked up by a number of external news outlets. But, you know, Bloomberg Intelligence, some lead, others follow. We've been, you know, very active on this for, as I said, for quite some time.

Speaker 2: It was the talk of the town in Davos, from what I understand. So, what is the secret sauce, if you will? Is it the low cost or do they have the processing power for the training

Speaker 3: and the deep learning? Yeah, well, taking a step back with the export restrictions you saw on the NVIDIA chips, etc., has been a major challenge. There is clear evidence, I think, that the big tech platforms, the Alibabas and Tencent, so this world both preemptively accumulated inventory of chips beforehand and have maybe, you know, gained a helping hand from their friends in other parts of the world to gain access to these chips. But then for smaller companies such as DeepSeek, although they still have access to a sufficient number of chips, you know, this challenge has prompted them to become more innovative. And so the way in which they've developed the models is far more computationally effective and therefore requires less chips and allows them to undercut their rivals cost-wise or price-wise significantly. So, without going into the technical detail on DeepSeek, it's something called a mixture of experts. It's the architecture they use, you know, one of the first in the industry to do this. They've really managed to forge ahead and, as you mentioned in your preview, you know, are ranked top 10 in terms of global large language models. So, I think, again, this just highlights and illustrates the point that whilst Chinese tech companies face, you know, problems on the semiconductor and chip side, they have a very strong track record in terms of innovation and software. Alibaba, Tencent are two of the world's leading tech platforms and that, you know, in numerical terms, China has more software developers than the U.S. by a ratio of approximately three to one. And what does this mean for U.S. AI in particular? Again, I think, you know, the Microsoft's etc. of this world were undoubtedly aware of DeepSeek, but, you know, it's the headlines that have probably prompted them to, you know, to have a little bit of a rethink and think how they can compete on this. I think one consistent point we've made over the last 18 months or so is that the technical barriers to entry on software are not as high compared to something like semiconductors. So, I think there's already evidence that some of the U.S. companies are moving to this mixture of experts architecture and I would expect the rest of them to follow. So, perhaps that could mean that DeepSeek's progress is forestalled to some degree, but I still see it as, you know, one of the leading large language model developers out there globally. But again, it just shows that China is very closely positioned as the number two competitor to the U.S. in AI globally on the software side.