Tech Sells Off On DeepSeek Concerns… And Fed Decision On Deck 1/27/25

0:01Live from the Nasdaq market site in the heart of New York City's Times Square this is fast money. Here's what's on tap tonight. AI in retreat. Claims of a cheaper, more powerful chat bought out of China, sending big tech sinking and NVIDIA to its worst day since the start of the pandemic. Are the concerns overdone? Or is the U.S. in a trade and is the U.S. trade, AI trade in real danger and out of energy? The once red hot nuclear energy stocks getting crushed on that same news today. Vistra Energy now negative for the year. Can the sector rebound? We'll debate that, too. Plus, a couple market bright spots in today's Sea of Red.

0:38An activist investor tells U.S. Steel to go it alone. And we count down to the first Fed meeting of the year, how the central bank is digesting all of this data, all the market moves and everything else that we're working our way through. I'm Courtney Reagan in this evening for Melissa Lee. Coming to you from Studio B at the NASDAQ on the desk tonight, we have Tim Seymour, Carter Worth, Steve Grasso and Mike Coe. But we are going to start with the AI apocalypse, maybe, question mark. China's deep seek ravaging names like NVIDIA, Broadcom, AMD, Palantir, Microsoft, after its R1 reasoning model reportedly outperformed OpenAI and cost less than$6 million.

1:17Yes, million with an M to design. Now, NVIDIA losing nearly$600 billion in market cap just today. That's the largest single day loss in dollar terms ever. That's more than the total valuation of names like MasterCard, UnitedHealth and ExxonMobil. The nearly 17 percent drop in flagship AI stock was its worst day since March 2020 and sent it below its 200-day moving average for the first time in over two years. The move in big tech dragging the broader markets down with it. The Nasdaq seeing its worst day since December 18th. The S &P down percent and a half. Remarkably, the Dow managing to end the day in the green.

1:57Our dear Traposa joins us now for more on how a company few people had heard of before today, and I'm still trying to make sure I don't say wrong, could have such a big impact on the market, and how much merit there is to these claims. Dee, we all spent a lot of time this weekend and most of today trying to think it through. Tell us what we've learned. Okay, so AI-pocalypse, not so sure about that. But without a doubt, this is a monumental shift in AI. What's real is its performance. We don't have to trust DeepSeek or the Chinese on that. This is an open source model, so we can see inside of it and compare it using third-party benchmarks.

2:34It is just as impressive as people say it is. Not only was it cheaper and quicker to build, but it's also cheaper and more efficient to run. So developers are adopting it en masse, as well as American consumers pushing it to the top of the App Store charts and dethroning chat GPT. Now, it hit the markets hard today because it raises a ton of questions about the way American AI has developed. OpenAI, Anthropik, Google, et cetera, have raised and spent billions of dollars to create ever bigger and better models. Demand for NVIDIA GPUs was insatiable. But late last year, and this really started before DeepSeek, late last year, model advancements stalled and a new race took off in reasoning or the inference phase.

3:16And that was DeepSeek's opening to essentially jump to the frontier and build off what those American companies already achieved. And this is key, it open sourced it. Now it's not just DeepSeek creating competitive models at a fraction of the price. You've got ByteDance releasing one last week. Berkeley researchers, they did something similar for$450 in 19 hours. So whether DeepSeek did it for$6 million or$60 million, that is now besides the point. It's called distillation. The most advanced models on the market right now, they don't need hundreds of millions of dollars in GPUs and infrastructure.

3:49And that is a massive paradigm shift for the AI trade as we've known it. Why the sell-off today affected everything from chips to energy to mega caps. Now, the saving grace for the bulls, though, and what I hear from my network in Silicon Valley, is that it's highly unlikely that American AI players are going to scale back their spend, those billions and billions of dollars. The race only got that much more important, and the pressure to stay ahead has intensified, and the American companies, they're going to continue to throw money at this because it's existential for them. This is all just so fascinating, Dee.

4:23And at first I was trying to think through it and thinking, is this good or bad? I mean, isn't competition great? If we can do something more efficiently and less expensive, isn't that good? Maybe not so great for America if it was done by this Chinese company when we're worried about all sorts of other ripple effects there. And that's obviously part of a bigger discussion. But altogether, I mean, how should we feel about this news, even if we don't know exactly what it costs? So that is such a great point. And there are many in the technology sphere, founders, CEOs, venture capitalists who say this is a really good development that DeepSeek essentially did what OpenAI was supposed to do all along, create an open source model, democratize AI.

5:05That's what DeepSeek did. It is so ironic that that came from a Chinese company. And when you worry about things like, okay, this is developed in China, So are we going to have the same kind of censorship that the Chinese government imposes on their AI companies? Well, again, it's open source. So developers can just take parts of it. And you've actually seen this happen in real time today. Perplexity CEO posted about this. You can just tweak the model so that you can leave things aside, like the censorship part that won't talk about Tiananmen Square. That is a remarkable development. On the other hand, Courtney, I mean, DeepSeek was top of the app store.

5:39That is the app that is going to consumers that are not filtering, that are not getting an uncensored version. And I mean, when you think about what we were worried about with Tic Tac, that's a social media network worried about influence and propaganda. This is ideological. If American consumers are adopting a Chinese AI GPT, essentially, that has massive geopolitical implications. Wow. That's just a lot to go through. Deirdre, thank you so much for joining us and for following this all day long. For more on DeepSeek and how it's putting the U.S.' place in the AI race at risk, she tried to say, check out our TechCheck digital doc, scan the QR code right here on the screen, or go to CNBC.com slash TCTakes.

6:21Deirdre, thank you again. I mean, wow. We just went through a lot there, Tim. I mean, when you started to digest the news from the weekend and then you saw the reaction, what did you think? Well, first of all, kudos to Debo, who was out on this news well before the market and a few days back, really covering a story that I think caught a lot of people by surprise. But, you know, the reality of what's been going on in the chip wars with China is, you know, what's the expression? I think necessity is the motherhood of invention and restriction might be the motherhood of invention here, too. And in terms of what's real and what's not, in terms of the cost to develop this, I think there's still a lot of questions.

7:00But there's no question, and Debo emphasized this, is that this is a case where what we have here is a large language model that is doing things a lot more efficiently than some people that have been in the market for a long time. So I don't think this necessarily calls into question the 2025 CapEx. And remember, this is a big week for big tech. We're going to be laser focused as we were. We're only a few days from Meta telling us about$65 billion in CapEx for 2025. So think about that. But it really does call into question 26. And that's something that if you look back to some of the chip players and NVIDIA notably, that is where valuations are looking at.

7:35Look, it comes back to also just positioning in markets. And this is here also what we talk about is just what the markets have. And what the markets had was a very crowded trade, a case where you had a dynamic even just three, four days ago. And maybe there's a coincidence to Stardate or not or, you know, Star Trek or where. Stargate. Yeah, exactly. But the bottom line here is I do think this is the kind of news that the market needed. And I do think this is the kind of news that also Carter is going to have a great call on what's broken, what's not on chart land. But there's still so much faith that had been placed in AI and spend in the impacts that today is at least a wake up call.

8:12Yeah, Carter, I wanted to go to you. I feel like I can't go too much further into show without talking about the impact a little deeper on NVIDIA stock today. I mean, do you think that that was overblown? And then what about some of the names that are competing with NVIDIA? I mean, because this opened the door for them. Maybe they don't have to spend as much money. Sure. So obviously, in each case, it's different, NVIDIA versus Invago versus Taiwan. But if you look at ARTI, which is an ETF that covers all AI and it has those names in it, I mean, it's an 8 % drop today. And one could say, I mean, that's a major setback.

8:40And yet the other way to interpret it is you just went back to where it was two weeks ago. Right. So if you think about extreme weakness on any given one day, this kind of thing, typically the precondition for that is one of two things. And it's the exact opposite. Extreme weakness comes from something that's been really bad and really bad. in an established downtrend and then has a collapsing type day, meaning a sick asset that has yet further dropping and collapsing, or it comes from the condition that is happening today, something that's been very, very strong, crowded, complacent, and then out of nowhere you get a drop.

9:11So that is, of course, the circumstance. You use the word crowded or complacent. And so the question is, and this is unknowable, right? I mean, let me just say something. I think I'm as qualified as the planet to determine whether tomorrow is up or down in these stocks, and I have no clue. Okay? So let's get that out there. it's usually right to resist the temptation to buy into a drop in gap. There's an expression called let the dust settle. And so it's tempting to do it, but usually it's wrong. Let a few days pass. And some people, I know you like to refer to a three-day rule or a ten-day rule.

9:43I'm not sure there's a fixed rule, but the concept of let the dust settle, jumping in just because something's down, usually is a bit impetuous. So you don't think it's a buying opportunity if you go by your rules just yet, but maybe it could be in Nvidia? So just to elaborate on that, it doesn't have to be a three-day rule, but it could be when it stops going down. So when you make a higher low in a consecutive day, then you buy it and use that day's low as your exit if it breaks down again. But a couple things that you mentioned. Is it good or bad? I think it's kind of good for some software companies.

10:17Because if the input cost is going to be AI, then shouldn't that in theory be better for your bottom line? The problem is the cat's out of the bag. So if you bar its use of the U.S., we already know that they can do it cheaper. So how much money have these top mega cap names spent and wasted? And going forward, they can't recapture that money. But to Tim's point, I think CapEx is still there. And to Dee's point, will Meta still spend the same amount of billions? Probably. Will they buy a ton more or will they go a different direction? I think the competitive nature that you noted in the beginning of the show, I think, is healthy for the markets.

11:06The problem is the placement of it. So was it Stargate? Was it TikTok? when Gina Raimondo was in China, they unveiled the Huawei phone. Right. So there's a lot of stuff with the friction between China and the United States that they want to show, hey, we have this up on you. We can still do this better than you. So there's a lot of infighting. But what we haven't talked about is, does Trump, where he started with tariffs, forgot about the tariffs, did those come back in now, where now you get the markets focused on something else. So it's the shell game of the market. Let the dust settle. I made a purchase today.

11:47I think that things got overdone. So, again, I think I would – and you asked about NVIDIA, and I didn't give you an answer to that. I mean, there are two ways to interpret it. Is something that's up 300 % and 400 % entitled to pull back 15 % or 20 %? Not to personify the stock, but that's the expression. And the answer is yes. The question is, is it weakness to take advantage of or weakness to stay away from, right? And there are two types. My hunch is that the downside is to 100 NVIDIA. And so that's, if it goes there, that's another 17, 18 percent. But right now, the stock is where it was in June.

12:20So it can't be considered extended, right, up and up and up and full. It hasn't made any price change since June. We're now January. And so support does come into play at some point. And at some point, people will step in and buy. and some will be doing it right now in the aftermarket and tomorrow pre-market. I think the downside is to 100. And again, I would apply that rule. Just stand aside. Nothing to be lost by postponing all new buying in a name like Nvidia until it just starts to stop, go down. Let's wait a minute. How many people said at 140 in Nvidia, if I could get it at 100, I'd go all in again.

12:54And then you're going to get tested because as it falls to Carter's point, is it something you want to buy? Is that weakness ongoing now when you really get your feet to the flame? Well, and one of the stories around NVIDIA that we've been talking about for months is at what point are the hyperscalers who are the biggest customers going to begin to be competitors? Oh, congratulations. Some confetti just fell. And to what extent that's right. I mean, there's going to be already competition within the ranks and that there will be more efficient models coming through. And, again, I think this puts a focus on ROI.

13:27And this puts a focus not only on ROI in terms of the companies that actually are developing models and where they're actually monetizing, but also in terms of which companies really are actually monetizing AI now. And if you look at the broader implications for the market, in today's market reaction and if you look at the indices, we know why the Dow outperformed, because the Dow is a very different group of companies. There's really nothing to say that suddenly that the industrial world or some of these older, more traditional companies are now going to benefit. But I would say there's no question that what you have done on the other side of this, if you are significantly cheaping the cost to broaden AI and raise the game and bring the game in that much faster to that many more companies, it is very positive.

14:06And it's very positive more broadly, possibly for the multiples and the margins of the Market Trades Act. There's so many threads to pull on here, but I do want to take a look and turn to our friend Mike because options traders are also betting that a full U.S. tech meltdown is unlikely. Mike Coe joins us here with a little options action. What do you have for us, Mike? Yeah, well, I mean, it's probably not that surprising that NVIDIA was the busiest single stock option today, net of that huge decline. First of all, it was always going to be one of the top three on any given day. But today's big move prompted a lot more flow into it.

14:37What was interesting is that calls still did manage to outpace puts. And actually, rather than reaching out and grabbing the falling knife, which is how we refer to that sometimes, and waiting for the dust to settle, as Carter was just saying, you know, where people are making their bullish bets, was in short-dated upside calls. So, for example, the most active contract were the weekly 130 strike calls. Buyers were paying about$1.80, a little bit more than that. Almost a couple hundred thousand of those things were trading. And so that was a way for people to risk a little over 1 % of the stock price betting on a rebound.

15:10I don't know that that's all that likely because I think the dust probably will settle for a few more days. But I think this is probably the point that Tim was trying to make. I think if AI is going to be so much more accessible and so much less expensive, we can think about that as a meaningful productivity and a more democratized productivity boost for businesses more generally. And if you take a look at all of the S &P 500 stocks today, two thirds of them were actually higher. Yeah, that is that is obviously very interesting. And I'm also wondering, you know, does this just give any of us the the incentive maybe to just broaden out a little bit?

15:47Right. I mean, look at all the eggs we have in this basket. I think that if I had 469 ETFs have Nvidia among the top 15 holdings. So even if you think that you're not an Nvidia holder, you hold it in some way. And maybe that makes you want to diversify that portfolio a little bit, even though, as Carter put it, a nice perspective. We are back to where we were in June. Yeah. In fact, if you look at semis relative to the market, I mean, they're back to where they were in February of last year. So, I mean, yeah, there's no question that this trade and the excitement around this trade has been kind of dead for a while.

16:17And, in fact, the market looking for leadership from both the mega cap techs, the triple Qs to the S &P or the semis to the Qs have been lacking for that for a long time. I think, Carter, since when? July? I mean, is that really? Yeah. I mean, and that's the stunning thing because you only can, only by looking at a ratio chart or relative, you see, wow, this has felt good, but actually it's negative alpha. That is the case. Yeah. And I just I think we're again when I woke up and read these headlines and I quickly checked my screens and heard from some of the smartest analysts on the street. The sentiment was pretty skeptical, skeptical that, hey, how could Google and Microsoft, how could these companies not have heard anything about this?

16:56How could we just have gotten some some insight? And yeah, I mean, ultimately, this is, you know, four or five days within the announcement of Stargate. This is a concept that I think a lot of people felt that we won't really ever know the costs attached to this. And, hey, by the way, do you really want DeepSeq as your backbone when they potentially can back up to your infrastructure and your entire Gmail account or all the things that we already were concerned about? So, you know, there's there's again, it's layered on top of the geopolitics that were already ripe and red hot in the middle of tariff season.

17:28And by the way, this has implications for our Fed conversation because there's no question. I was going to say, there's so many different, like, webs we can weave here. But I do want to move on to our guests because we have a noted head fund manager who started worrying about DeepSeek around Christmastime. So let's bring in Dan Niles of Niles Investment Management. He was a semiconductor and computer hardware analyst. So, again, tying everything together during the dot-com bubble. I know, Dan, obviously, this is something that you brought up as a concern. So when you started to hear the news, it probably didn't feel as shocking to you as it did maybe to some of the rest of us.

17:57But how are you thinking through the action that you saw today and what you think we should be aware of as investors and traders moving forward this week? Well, I think it's more than just this week. So I've been talking about the fact that I expect an AI digestion phase in 2025 since probably mid-year. And there have been a lot of data points along that way, where if you think about it, Microsoft, which has the 49 percent economic interest in open AI, when they reported their June quarter, they guided revenues below the street for the September quarter. Then they reported the September quarter.

18:30They guided revenues below the street for the December quarter. And then when Microsoft CEO was asked in December, are you supply constrained? He said, well, I'm power constrained, but I'm not chip supply constrained. So, and then you had Ilya Sutskvar, founder of OpenAI, saying back in November, I think, that we've already trained on all the data on the internet. So, training scaling is basically dead. So, you've had a lot of data points leading up to this point. And when I initially saw the DeepSeq data in, you know, it was released on Christmas Day, I was like, yeah, there's no way they trained on it for that little.

19:10And then as I talked to contacts in Silicon Valley, because don't forget, this came out on Christmas Day, right? Everybody's on vacation. We're all enjoying New Year's, et cetera. It became increasingly clear that they had made some fundamental improvements in the way to approach this, whether it's distillation or mixture of experts, partial activation. I'm not going to get into all the technical stuff, but bottom line, they did make some big software improvements. And from what I heard, everybody, because it's open sourced, was looking at it saying, hey, how can we incorporate some of this into our models?

19:42Because don't forget, it's not like OpenAI came out of not having the benefit of Google, which you could argue invented transformers or did a lot of the early work. They just didn't commercialize it, which OpenAI did. So all of this stuff builds one on top of the other. And what it gets down to is I don't think there's any chance that NVIDIA, and I've said this before, does revenues up 50 percent this year because whether it's Microsoft cutting back on CapEx, because don't forget, they said they're going to spend 80 billion this fiscal year. Well, if you do the math, they spent$20 billion in September.

20:17They said that's up in December. Their fiscal year ends in June, which means the first half of 2025, CapEx spending is down from the back half of 2024. And it's been growing at 70 to 80 percent for a year and a half. That's a very big change when the largest spender on AI is going from growing 70 to 80 percent to not growing at all. So I understand that you don't believe in NVIDIA necessarily, or at least you don't believe that the revenue growth they'll have is what they've said they will have. What about some other names, names like Meta, for instance, which was up today with all of this? I understand that's a name that you like.

20:54What are you thinking about that name and how this plays out? Yeah. And to be clear, NVIDIA is never guided for this year. They only guide one quarter in advance, but that's what the analysts have out there. And I just don't think those numbers are achievable. But you're correct. The thing you should look at, and I talked about this on Friday in something I posted, which is, look, you have to look at the stuff this benefits. Because when you lower hardware costs, it really helps with the software that's sitting on top of it. And also guys that use AI. So Meta uses AI to figure out, hey, this is the video that Courtney would like to watch.

21:30And, oh, these are the ads she's willing to engage with. And they use AI to help with that, which is why their numbers went up so much last year. And so that's one of the names that I think will benefit. And funnily enough, a name like you look at the software names, they're also going to benefit a lot. And so those are areas that we've been swapping into in terms of our likes versus dislikes. And if you go further back to the internet bubble, originally you had all the hardware stuff take it up. So Cisco, Sienna, et cetera. And then you went through this 78 % crushing in the NASDAQ. And what were the biggest companies that came out of that?

22:10Well, the ones that used all those lower hardware costs, like an Amazon, for example. And so I think you're going to see the same thing play out over a number of years. And by the way, I think NVIDIA revenues will grow. It's just going to grow slower than a lot of people anticipate. and but I think the value is really going to be in the software guys that go okay these are the changes I can make and how can I ride on all of this investment that's happened underneath it much like dark fiber did not turn out to be what everybody thought back in the internet bubble and you went through a crushing but that crushing was great because all the guys that built on top of all that dark fiber overspend did very very well.

22:50Got it that all makes sense and yeah that AI for the Facebook and the platform is really working because I'm seeing an awful lot of Buckeye championship videos, and I like it. Dan Niles, thank you so much. Well, for more on DeepSeek and how that could impact the future of AI, let's head to the iConnections Global Alts Conference in Miami where the captain of the fast money ship, Melissa Lee, she joins us with Sunny Madras, the COO and president of AI Company. Gosh, how many times am I going to say that today? Grok, Mel, take it away. Thank you so much, Courtney. Great to see you and great to have you here, Sunny, on a day like today.

23:24You were saying that you thought the market reaction was overdone to the news. Why is that? Well, just of the magnitude. You know, really, if we think about what's happened, we now have a world-class open source model that is low cost, easy to deploy, and will probably be deployed in many different places. So you would think that the market would be excited. There'd be more use of AI. And if there'd be some amount of movement downwards, it wouldn't have been to the magnitude that we saw today with NVIDIA. Right. I mean, in terms of broader adoption, the thesis, though, is that it's going to force deflationary pricing within AI.