How DeepSeek Shocked Silicon Valley & Crashed Nvidia | Pivot

1. KSKara Swisher0:00

   There's a new AI model on the scene that's smart, cheap, and made in China. It's called DeepSeek, and it's causing a panic in Silicon Valley, which is paying a lot of attention, and also on Wall Street. DeepSeek has reportedly outperformed models from OpenAI, Meta, and Anthropic in some tests, and it operates at a fraction of the cost of those models using fewer high-end chips. This is the ones that are made by NVIDIA and are hard to get, and the incumbents have been pricing them up heavily, uh, by grabbing all of them. The markets are not reacting well to DeepSeek. As of this recording, NVIDIA is down 16%, Oracle is down 10%, Microsoft Soft is down nearly 4%. Obviously, Meta is gonna be affected, all the others. So there's a lot to talk about, and I've seen different analysis of exactly what DeepSeek does. Yann LeCun from Meta was making an argument that it isn't as, what they're re- they're doing sort of a cheap and dirty version, and it's not nearly as... The stuff they're doing is much more advanced by the US companies. Uh, it's currently number one on Apple's, uh, free top apps chart. Uh, again, China invading i- in this country in a very different way. So thoughts on this situation? 'Cause you and I have talked about this quite a bit. Is this money ill-spent by US, uh, companies, and is it being relegated to the en- rich incumbents?

2. SGScott Galloway1:12

   Well, first, you just have to temper the, or put some context to the... I mean, NVIDIA's down 15 or 16%. It's shed something like a half a trillion dollars, which basically, if you take out Tesla, it shed today the value of the entire global automobile industry sans Tesla. So this is pretty dramatic. But at the same time, that just takes it back to its valuation in October. And when you look at market dynamics, when these companies have experienced these type of run-ups, it is like a balloon inflating beyond its natural capacity, and the slightest, the slightest touch can pop it. And so in some ways, the market was probably looking for an excuse to take these stocks down a bit, and it got it. Because what's interesting is NVIDIA will have a pretty interesting argument on, on Capitol Hill saying, "When you refused to let us sell into these countries, they come up with workarounds." And in this case, this workaround might tank the (laughs) US economy. And everyone's excited about the fact that these models... OpenAI, supposedly, their models, their LLMs cost 100 million to train, and they're claiming this thing costs... and they've been public, it's open source, costs a little over five million to train. So whereas the majority of LLMs and, uh, AI companies have been taking sort of this brute force strategy where it's buy as many chips as possible, this is saying maybe you don't need as many chips. The thing I find equally interesting is the second-order effects here, and that is Constellation Energy and some of these nuclear stocks have skyrocketed because the choke point was supposing it'll be energy. But now with this, this model, which appears to have chips speaking to each other in a more efficient, less energy consumptive way, nuclear stocks are crashing. Y- uh, electric, Constellation Energy, all these things that have had incredible run-ups are saying, "Wait, the entire supply chain or the assumptions we made about the supply chain in terms of the, the kind of the brute force of chips that we're gonna need, the amount of energy," it's all now coming into a little bit of question. But to be clear, the correction here is like, it's taken them back three months. And all of the stocks that have crashed, quote-unquote crashed, are, are only up, you know, 70% for the year now, not 98. And a lot of, uh, analysts, the smart analysts I've read have said, "Like every community or every sector, it's gonna bifurcate into the cheap layer and then the high-end layer, which will still go hard at massive computing and massive energy and do more sophisticated things." And this will be sort of... You know, everything eventually goes Walmart, Tiffany, right? And they're saying this might be the Walmart, and it's the Chinese, and they'll come up with cheaper models. But I, it's fascinating to see that basically this notion, this, this kind of conventional wisdom that you would need massive GPUs and massive energy may not be, um, kind of the written in law-

3. KSKara Swisher3:55

   Yeah, let me-

4. SGScott Galloway3:56

   ... that we thought it was gonna be.

5. KSKara Swisher3:57

   Let me read Yann LeCun, who's the head of Meta. I just recently interviewed him, and you can go listen to that long interview about this. But he's writing, "To the people who see the performance of DeepSeek and think China is surpassing the US in AI, you're reading this wrong. The correct reading is open-sourced models are surpassing proprietary ones. DeepSeek has profited from open research and open source, uh, for example, PyTorch and LLaMA from Meta. They came up with new ideas and built on top of other peoples' work. Because their work is published and open source, everyone can profit from it. This is the power of open research and open source." Obviously, this is the way-

6. SGScott Galloway4:27

   He's talking his own book.

7. KSKara Swisher4:29

   That's correct.

8. SGScott Galloway4:29

   Right.

9. KSKara Swisher4:30

   I was just gonna make that point.

10. SGScott Galloway4:31

    L- LLaMA is open source. Yeah, yeah.

11. KSKara Swisher4:32

    Yes, that's correct. That's what I was gonna say.

12. SGScott Galloway4:33

    Yeah.

13. KSKara Swisher4:33

    But it's interesting. He's having really interesting arguments, and he said... He, he's having a bunch of them, which is just interesting. And one of them that he just did... 'cause Gary Marcus, this guy who's somewhat of a, a crank a little bit, um, was saying that Congress needs to bring in Zuckerberg and LeCun to discuss how their unilateral open-sourcing decision rapidly undermined the US advantage in generative AI. He goes, "An absolutely hilarious take revealing the complete misunderstanding of the fact that open research and open source accelerates progress for everyone, from someone who's repeatedly claimed that deep learning was hitting a wall." But one of the things he just wrote again, 'cause he's, he's be- he's getting in there very deeply, "Major misunderstanding about AI infrastructure investments. Much of those billions are going t- into infrastructure for inference, not training. Running AI assistant services for billions of people requires a lot of compute. Once you put video understanding, reasoning, large-scale memory, and other capabilities into AI systems, inference costs are gonna increase. The only real question is whether users will be willing to pay enough directly or not to justify CapEx and OpEx." I think that's, that's probably... He thinks these reactions are woefully unjustified, and at the same time, he's sort of arguing that they aren't, right? I- which is interesting, interesting, interesting times.

14. SGScott Galloway5:39

    It's just so typical of the Chinese to come out... I mean, the, the entire Chinese economy-

15. KSKara Swisher5:43

    Mm-hmm.

16. SGScott Galloway5:43

    ... was sort of built on more ****** less.

17. KSKara Swisher5:45

    Yeah. Mm-hmm.

18. SGScott Galloway5:46

    And my guess is they had a mandate, or they've said, "All right, we're not gonna have access to the same level of, you know, high-end chips. We need workarounds." And it's-

19. KSKara Swisher5:56

    Yeah. That's what they're doing.

20. SGScott Galloway5:56

    ... it, it appears they just found really interesting innovation. And-

21. KSKara Swisher5:59

    Using open source *******-

22. SGScott Galloway6:00

    Yeah, using open source. The, I mean, the scary thing, uh, I'm, uh, I mean, in typical Meta fashion, their LLM-

23. KSKara Swisher6:07

    Mm-hmm.

24. SGScott Galloway6:07

    ... you can download a v- a version of LLaMA with absolutely no guardrails, and you can-

25. KSKara Swisher6:11

    Mm-hmm.

26. SGScott Galloway6:11

    ... you can request information-

27. KSKara Swisher6:13

    That's correct.

28. SGScott Galloway6:13

    ... on anything. (laughs)

29. KSKara Swisher6:15

    Mm-hmm.

30. SGScott Galloway6:15

    You know, the most politically correct I find of, of them is-

Episode duration: 7:51