Closing Bell - Closing Bell: The Deep Sink 1/27/25

Episode Date: January 27, 2025

The deep sink today in AI stocks weighing on the Nasdaq. So, is this really a watershed moment for investing in that space? We discuss with our own Deirdre Bosa and Bernstein’s Stacy Rasgon. Plus, Fundstrat’s Tom Lee tells us the names he would be buying on today’s dip. And, Nvidia shareholder Doug Clinton reacts to the stock’s worst single-day market cap loss in history. 

All right, let's do it. Welcome to Closing Bell. I'm Scott Wobner, live from Post 9 here at the New York Stock Exchange. This make or break hour begins with what else? One of the worst days on record for shares of NVIDIA. Looks to be the largest ever single day market cap loss for that stock. We will follow it into the close. Hammered today on fears about AI competition and cost. Both issues sending shares of many tech companies sharply lower today. Coming up, top chip analyst Stacey Raskin on what all of it means for your investments. In the meantime, take a look at the majors here with 60 to go in regulation. NASDAQ obviously where the action is today. Many of the mega caps are lower. It's down more than three and a half percent is the NASDAQ. It's worth noting,

by the way, that Meta and Apple are higher. And we'll tell you why that might be coming up in just a bit. AI power plays. They're all lower as well as other software stocks tied to the space. So it is a dramatic and widespread sell off. If it's AI, it is largely lower today. Fundstrat's Tom Lee will be here in just a bit with his view on the markets and where they might head from here as well. He'll try and put some of this into context, too. It does take us to our talk of the tape, the deep sink today in AI stocks and whether this is, in fact, a watershed moment for investing in that space. Let's start with our own Deirdre Bosa. She's been out front on this story from the get go. Why are we seeing what we are seeing, Dee? Because the whole story, the whole narrative has changed, and it happened so quickly. I mean, you could argue that it went back earlier than DeepSeek, but really over the last few months,

there's been this acceptance in the AI world that we sort of, scaling lots, we hit a wall in terms of data. And we went from the pre-training phase to the inference phase, and that's what allowed something like DeepSeek to even happen. Companies, startups, AI labs in China, they can get right to the frontier, get all the technological developments and advancement that OpenAI and others have done, and build on top of that. This basically tells us or makes us at least question, do we need the same kind of infrastructure? Do we need the same number of GPUs to create the most advanced models? That's what DeepSeek told us. And not just DeepSeek, by the way. ByteDance as well last week.

Berkeley researchers did something similar as well. So this era of bigger is better, throwing billions of dollars to create the next best model is over and what deep seek told the market told silicon valley told users is that you can do this a lot cheaper and more efficiently and by the way cheaper to run as well i mean i know there's a lot of focus on how much it cost or didn't cost deep seek to build this model but a really important point here for developers too is that this is extremely cheap and efficient to run. And that is also a very big change, coming back to the fact that this is open source. And that also is something that should not be lost today. This is open source,

and I think we're clearly now in a world where open source is going to be the predominant model, which raises a lot of questions about all the companies like OpenAI that developed closed proprietary source models. You're not the only one who's thinking about that, Dee. I know you heard our conversation with Bill Gurley last week. I want you to listen to what he said. He singled out your reporting, by the way, on this whole issue. But listen to what he said about open source and what it may all mean now. The open source performance is right there now. And I don't think you can put the cat back in the back. We may be moving from performance at all costs to the optimization phase of AI. And we will at some point in the future, even if it's not today. And that may mean that there

are different companies that are doing different things. It could be positive for open source. It could be positive for these inference clouds that are working on super high performance inference. It could be positive for non-GPU startups like Grok or Cerebrus or even Google's TPU or Amazon's Tranium. All those things could be better positioned in an optimization phase than maybe they were in the build-out phase. Your reaction to Gurley, Dee, and also the fact it's not lost on anybody today that meta is higher, albeit slightly, but it's a decidedly red tape. Meta is an open source builder, right? So they would benefit from these developments, whereas a lot of others wouldn't. I think it's interesting that Gurley didn't note NVIDIA,

right? There's a lot of questions. NVIDIA had the dominant GPUs, the dominant chips in the pre-training phase. In inference, there's going to be a lot more competition. I mean, the people I talk to say that NVIDIA is going to be okay because they're still the best around, but it does raise questions about its valuation and its market dominance going forward. Another interesting angle of this, Scott, which I watched with such fascination over the weekend, was DeepSeek rising

to the top of the App Store because it's one thing for developers to take a hold of this. But as ChatGBT showed us, when it enters the consumer zeitgeist, that is something really important and has a lot of ramifications. And it does make you wonder what's happening to OpenAI's moat that DeepSeek could sort of surpass it in the app store. And that is, you know, raises a lot of geopolitical concerns, right? Because this is not an open source model that developers are tweaking and they can take out the censorship parts of it. This is consumers just getting whatever sort of was built under Chinese government guidelines. So there's, you know, more knock on effects of this, too. Couldn't help but wonder today if OpenAI were a stock, what would it look like today? We'll never know. Obviously, not today. We won't. Dee,

thanks for your reporting. Appreciate that. For more on Deirdre's reporting, by the way, on DeepSeek, you can go to cnbc.com slash tctakes. She digs into the research paper, its cost, the technological breakthroughs, and so much more. Let's try and get more answers now and bring in top chip analyst Stacey Raskin of Bernstein. It's great to have you on a very important day to speak with you. What do you make of this story? What is the worst drop for NVIDIA, at least, in years? Yeah, you bet. So I spent a lot of time this weekend,

most of my weekend, looking at this. We knew it was going to be a circus today. It was pretty clear, just given everything, all the takes that were there in the Twitterverse. I had sort of three conclusions. It's my own conclusions from this. I call it a semi-informed layperson around this. But number one i think

the big issue was that the headline the initial takes that came on this was oh my god china's duplicated open ai for five million dollars which is categorically not true we can talk about what they did but they did not train this reasoning model for five million dollars that that did not happen we can talk about why but but number two how did they get the efficiencies that that did not happen we can talk about it but number two how did they get the efficiencies that they did these are these are very very good models like i don't want to discount that but the types of things that they did they're not miracles they're a mixture of experts um and and they did some very interesting with multi something called multi-headed attention and they use lower precision calculation these are all things things that are known and all of the labs are all working on,

if not deploying already. So there was nothing here that was like unknown to anybody else like in the AI industry. Like these are not miracles. Finally, like number three, I personally think, we'll talk about it. I think that this is clearly a panic today. I think it's overblown.

The school of thought here is that like, oh my God, these are so much more efficient. Like, we don't need as much investment. So number one, we actually don't know what they're spending to train the reasoning model that's causing the angst. My suspicion is quite a bit more than what they did to train the base model. And number two, you got to remember, like, I'm a semi guy. Like, I actually view cost reduction as a good thing we want adoption like in semiconductors again i'm probably biased but like cost fell by a factor of two every two years for 50 years it was not a bad thing for semiconductor it was a fantastic thing for semiconductor man

unless you think that we were already close to saturation in terms of the amount of compute that was going to be required for ai we need these kinds of of innovations and efficiencies because otherwise like you're not going to be able to get to where they're going anyways. And so I think this was already happening. I think it needed to happen. And I can make, I think, a very clear argument that the more efficient things get, the more demand we will have.

Not so I actually think over the long term, this is a good thing, not a bad thing. But clearly the market disagrees today. So that is what it is. I mean, NVIDIA hasn't been afforded the multiple it has, nor the stock price, for that matter, on the idea of a commodity. Hasn't it been rewarded the way it has on the idea of the best of a moat on some level of exclusivity, at least now? Well, the issue here is not that. Where is the commoditization? It's not the chips.

It's the models, right? So again, if your business model is we're developing AI models that we're going to sell to the general public, maybe you worry. I don't know. It doesn't mean that the folks

that are doing this still don't need to buy a lot of chips. I think they're going to still be able to buy a lot of chips. But what did we see last week as all this was happening? We saw Facebook or Meta, I guess, massively take up their CapEx guide. We saw all the Stargate stuff. We even had China actually announced a $1 trillion RMB, so like roughly $140 billion US program for AI investment in China.

This was all last week, right as all the DeepSeek stuff was happening. DeepSeek's not new either, by the way. The V3 model, the base model, was actually released on December 26th. Like, none of this is new. None of this is surprising to any of the large U.S. guys that are spending all this money.

They're spending it anyways. Yeah, but I mean, NVIDIA's future in some respect is leveraged to the development, the production and the success of Blackwell. If the company is telling you it's I'm going to read a comment from a spokesperson at NVIDIA, which I think speaks to this. And we can discuss on the other side where they say DeepSeek's work shows how new models can be created, leveraging widely available models and compute that is fully export control compliant. Now, you could read that and say, well, is the company admitting that you don't need the latest, greatest, most expensive chip to do what deep seek appears to have done? No, no, I don't think that's what they're.

So they're making a statement on China here because that came out of China. But look, you're going to have like... If you're getting a massive proliferation of models, frankly, of all shapes and colors and sizes, you're getting adoption. This is over time what we want, right? This is not... Again, this is not a bad thing.

You're going to hear this a lot today, probably already, but there's a concept known as Jervin's paradox. The idea is as efficiencies and costs come down, like demand more than offsets. I keep going back, you know, I went to the Hot Chips conference a couple years ago and Bill Daly, who's NVIDIA's chief scientist, was giving the keynote. He was actually talking about performance improvements in GPUs. And he talked about like over the prior 10 years, NVIDIA had improved GPU performance by like a thousand X. It was a bunch of things, process tech and architectures and sparsity and lower numerical precision, all kinds of things.

But what he said in that keynote was like over the next 10 years, we want to improve the GPU performance by another million X. Like, again, unless you think that we are at the limit of what the compute that we need, I think this is a good thing. And if you look historically, we've never, ever had enough compute. Like I've got the data going back even before AI was like taking off, just looking at like sort of installed compute in the data centers, like in MIPS and millions of instructions. I have this data going back like 30 years. It was going at 50% a year forever.

It never stops. So I'm personally not of the belief that you can have enough compute. I think if we're getting efficiency gains, which I think we need, if we're going to meet the compute demands that AI and accelerated computer placing on us, I think that we're going to get absorbed. I think we need this.

I'm not of the belief that there's a limit on compute needs. Maybe not, but there might be a limit to which some are now willing to pay for these chips, which are very expensive. So is there a price issue potentially to think about? And then the obvious margin issue. I mean, the reason they're making money hand over fist because these chips are expensive and everybody thought they needed as many of them as they could

get their hands on. I still think they're going to need lots of chips. And in terms of like their ability to price, it comes back to exactly that same point. How much can they improve performance? So if you, I'm going to make up the numbers, but you kind of get it. You look as Nvidia went from, say, Ampere to Hopper. And the price, I don't know, kind of doubled, whatever it was. But the performance improvement was something like 10x. So that was a massive NPV positive investment for their customers.

And they sold them hand over fist. And if you lookwell and again they don't really sell gpus they sell servers but if you wanted to sort of suggest the gpu price is going up you know 50 or whatever it is you know if you look at the performance metrics for blackwell versus hopper they're going up anywhere from 3 to 30x depending on what you're doing so the same thing a massive improvement and these are efficiency gains too that like nobody nobody seems to be complaining about. My own belief for NVIDIA pricing and margins is as long as they can continue to improve the performance, that is sort of the cap on what they can raise pricing on. So as long as they can keep performance going up at a good clip, they can keep pricing going up and they can keep margins. I worry about like this is why I

don't really worry about competition and everything in terms of driving margins down. If we're ever reaching a point where they can no longer improve performance, then I really start to worry about their ability to price and their ability to drive margins. But it comes down to the same kind of trend. I think we need compute. I don't think we can have enough compute. And again, in that context, I think the sell-off today, given these kinds of dynamics, this idea that this is just over-efficiency is going to completely swamp the need for the data, I don't think that's true. I don't think it's ever been true.

We'll see how it all develops. I appreciate so very much your insights, Stace. Thank you very much. Stacey Raskin, this day wouldn't be the same without hearing from you, so I appreciate you. Today's DeepSeek development, by the way, raises all sorts of questions about what the hyperscalers like Meta and Google are spending on AI and whether it's necessary given this apparent breakthrough. Steve Kovach is here at Post9 with us today. I mean, that's the natural sort of follow-on for this conversation.

Yeah, and it's not just this conversation. We're getting into big tech earnings just in a couple days. Wednesday, Microsoft is going to report. And, Scott, I want to give you a little history lesson here. You don't have to go too far back. Just six months ago, all these hyperscalers were getting shellacked because they were saying we're going to spend, we're going to spend, we're going to spend on CapEx. We don't know how much we need to spend, but trust us, we need to spend. I think there's a really demonstrative quote here. I want to play you from Alphabet CEO Sundar Pichai in earnings call from July last year. Take a listen to this. One way I think about it is when you go through a curve like this, the risk of underinvesting is dramatically greater than the risk of overinvesting for us here, even in scenarios where if it turns out that we are over-investing, we clearly, these are infrastructure which are widely useful for us. So basically saying we're going to spend, we don't know how much we need to spend,

but trust us, we need to spend it. Then you have Satya Nadella over at Microsoft saying kind of the same thing, but also saying we just see so much enormous demand for all this stuff. Now that narrative is just getting thrown out the window because what DeepSeek seems to have proven was you don't necessarily need these latest and greatest chips. You don't need to be spending $80 billion every fiscal year in order to do what OpenAI has already done. And that is the lesson that they seem to be absorbing today. And it's going to be the dominant theme, I guarantee you, on Microsoft's call on Wednesday. That's all that people are going to be talking about. Don't you think that's a good point? I mean, this has now theoretically

co-opted the earnings season for these big companies. That the majority of the questions that are going to be asked are going to be all related to this. And you're going to have different analysts trying to model in different scenarios as a result. And how good is Mark Zuckerberg feeling today? Because just a couple of days ago, sure, he upped the CapEx for Metis up to like $65 billion for this year. That's great. But also, to Deirdre's point earlier in the segment, saying they are the open source king right now. They have this Lama model that they deployed relatively quickly, and they were able to show, hey, we can just do what OpenAI does. We can do it free. We can do it open source. And by the way, the deep cheeks of the world can go ahead and build off of this. Jan LeCun, who is the head

AI scientist over there at Meta, basically saying we don't need open sources, what this is being built on. This is a proving ground for open source models that this could be just as successful as these closed source models that we keep talking so much about at OpenAI and Google. So this is going to be the theme of the calls. Meta in a great position right now. Microsoft and Google,

they're going to have to really defend their spending more than they ever have. And again, six months ago, that was the theme. It's going to come right back around here again this earnings season. Has changed the conversation. That is for certain. Thanks for being a part of it, Steve Kovach, covering that angle for us today. Let's bring in Trivariate Research's Adam Parker and Payne Capital Management's Courtney Garcia, both are CNBC contributors. Courtney, I'll begin with you. You're watching all this unfold. Has this changed

the way we should think now about these mega cap stocks and what earning season could bring? Yeah, and I think what happens today is there's really more questions than there are answers, right? I think when it comes to these companies, you're saying, OK, is artificial intelligence going to be much cheaper to produce, take a lot less computing power, or is this kind of a false head where this wasn't actually as cheap to produce? Because you're getting a lot of questions of, is this kind of too good to be true true so i think if this isn't as cheap and it's kind of is that too good to be true then some of this sell-off is going to be overdone or on the flip side maybe it can be that much more efficient which in the long run is going to be a really good

thing for the other 493 companies in the spf 100 or your small or your mid-cap companies who are going to be able to cut costs and become much more productive in the SPF 100 or your small or your mid-cap companies who are going to be able to cut costs and become much more productive in the long run with artificial intelligence. So I think, if anything, you want to try to look at the opportunities here. So AP, you preceded Stacy in that role. You are a former chip analyst. You've come on this show recently and really leaned into the chip trade over software. What does this make you think? Have you started modeling this out in your own mind? I think two things. Look, first of all, Stacey's saying, I think he said, I'm a semi-informed layperson. That was probably the

biggest sandbag of the day on your show. If he's semi-informed layperson, I think the rest of us are in trouble. I think what he said makes a lot of sense, that these things always drive more compute. I think two things can be true at once, Scott. One, the sell-off could be merited. And two, semi and compute could still be okay in the long term. And that's kind of my judgment. I think these stocks got very high beta, and I think the growth themes got really correlated. One of my observations today is if you're a growth PM, you say, what do I like?

I like semis. I like power. I like electric creation industrials. And then look at the stocks that are all getting killed today. They're eating. They're VST. They're CEG. They're NVDA.

It's all themes. So I think this, to me, speaks to the real importance of risk management when you run a fund or a portfolio. You have to focus on do you own some high beta or not, and you have to position size correctly. And so to me, we do a lot of that stuff at Trivariate. I know, but do I need Adam to rethink the whole AI trade? And I don't know how, as an investor, I'm supposed to think about what the correct valuation might be in what now appears to be a fast-changing and resetting world. It's definitely going to be volatile. It's definitely going to be fast-changing.

As you know, I've been a little bit more worried about the market in the first half of this year for the first time in a while. I didn't expect this, but I thought with these expectations we could get some volatility. I don't think semis or compute are going to be destroyed in any way. I think to Stacey's points, I think they're all spot on. These things tend to drive more compute. And do I want to say today's the greatest entry point for NVIDIA on itself? No, there's sentiment that can overcome things for a while. But if you look out 6, 12, 18, 24 months, you're going to need more chips. You're going to need more high- end chips. You're going to need them from the best companies. So I just think it's not a game changer for the demand for compute and high margin compute.

It's never been true that there I think what they said is right. We've never had enough high end compute. Sure. But but if anything, I wonder, Court, if it's a game changing moment for the market itself in that, you know, these stocks had once again taken a bit of a leadership role. And here they are coming into earnings season with a bar that was probably higher than not, just given what the stocks had done. Now we're talking about tariffs again. We had this spat with Columbia. Where that goes or what else is next remains to be seen. But we're worried about tariffs front and center. And now we have these earnings reports on stocks that are absolutely obliterated today. Yeah. And I think even before this news, this was a question with your mag seven, where the amount of capital expenditures they have, the question is, is that justifiable for them?

And so this is where we've really been talking to clients to say, you know, I do still want to own these companies. I'm not willing to throw in the towel even with the news today. But most people, if you have not made changes, are really overexposed to these big tech companies. And this is where you can have days like today and you just don't want to have too large of an allocation there.

So we've been adding to those other areas of the market. So even today, there's plenty of sectors that are doing well. This is where if you're well diversified, it's not as big of a hit for you. And even with this hit today, still, you might be overexposed. It is worth saying, do I have too much in there? I still want to own it. I just don't want to own too much of it. There's still a lot of other areas of the market that are cheaper, have less valuations and probably still can benefit from artificial intelligence in the long run. AP, financials, you talk about benefits of artificial intelligence. Healthcare seems to be ground zero for that. Yeah, we love healthcare. That's our biggest overweight. It's been a painful trade up until very recently, but I think they'll be the primary

initial large beneficiary as they get better at predicting customer and employee behavior. I think financials will benefit too, but I think it could take a little bit longer because most big banks have to pay up front to run systems in parallel before they shut down the old system. So you may take a little longer to see some of the productivity from some of the big banks than you will from healthcare. That's my take.

Back to just one thing you said before, Scott. I was just thinking about NVIDIA just so that people could understand. If NVIDIA is 6% or 7% of the S&P, but it's a 2.5 beta stock, its effective position size is 15%, 16%, 17%. It's just enormous when you have high beta stocks like this. So I think this is all about risk and risk management, not the destruction of the business in the medium to long term. OK, we'll leave it there. I appreciate both of you. Thanks so much. That's Adam Parker, Courtney Garcia. Thank you very much. We'll see

you back here soon. We're just getting started here on Closing Bell. Coming up next, Fundstrat's Tom Lee. He reacts to today's sell off and reveals the names that he would in fact be buying on this dip. He's at post nine after the break. All right, welcome back. Today's sell-off in AI stocks happening just ahead of some critical earnings this week, including Apple, Meta, and Microsoft, not to mention the Fed meeting. Here to share what he is watching is Fundstrat's head of research, Tom Lee.

Welcome back. It's good to see you. Great to see you, Scott. Especially on this day. What do you make of this? Well, you know, markets don't like uncertainty. They like visibility. They like moats. So markets are a little scared today. To me, it's an overreaction.

I mean, NVIDIA decline is the worst since March 2020. And we know that that ended up being a huge opportunity for investors. So I think it's, you know, it's not a fun day, but I'd be looking at this as an opportunity. I've just, I've been thinking about, you know, the people who have been coming on today of all sort of professions, whether it's a strategist or investor or whatever, analyst, defending all this or saying it's overblown. How do we know? What if it's not? And if it isn't, what does that mean? We don't know if it's overblown. You're right. But this is the nature of technology. I'd be personally surprised if NVIDIA became Betamax in the past week. I mean, that really would be the kind of change that would be required to really justify selling NVIDIA here.

And given, I think, the need for AI because of global labor shortage and NVIDIA chip dominance is still strong. Unless a new model emerges that doesn't require GPUs entirely, I'd say that markets just generally do fire-ready aim. I mean, I think it's hard to know, though. I won't disagree with you on that. I mean, that's how the market tends to do that. But a lot of AI was sort of sold to investors on a hope and a dream, right? We haven't seen, in many cases, the ability yet to monetize all of this, or at least to have a sizable return on the investment that a lot of these hyperscalers make tens of billions of

dollars that they're pledging to spend. Doesn't that matter? I mean, what if some of that hope and dream just doesn't come to the degree that we thought? Well, we decide that it wasn't worth the money that we were willing to pay for it. I mean, these are fair questions. If NVIDIA was at 100 PE, I'd probably say that there's too much good news priced in. But, you know, the PE is around 30 times. I'm not sure that that's very demanding for investors and if you look a couple years out it's in the 20s i'm i don't think nvidia's price for protect for perfection is the market itself i don't think so i mean the 10-year now is is backing off to four or five so you're still paying over 20 times for a 10-year bond the median pe and the s p is around 18 times yeah but mean, because it's skewed,

though, because of the megacaps, we're still like 22-ish times. So, you know, earnings have to be really good from here on out, right? Don't tell me about multiple expansion again driving this train. You got to put up or shut up when it comes to earnings this period. That's right. But, you know, I've been researching companies since the early 90s, so over 34 years. I never found a PE for a high quality company to be at 22 times, let's say, for MAG7 to be a demanding multiple. Again, I think if you're talking about price for perfection, it really has to be stocks where you have to back into the NPV of 100% earnings growth forever. What happens if, at bare minimum, we are questioning NASDAQ stocks for the foreseeable future?

This doesn't feel like, oh, tomorrow everything's going to be great kind of deal. So as long as we're doing that, is there enough momentum behind some of these other areas of the market that this market as a whole can be OK? Well, today, market breadth is actually pretty good. Yeah, financials, health care, staples, they're green. Dow's up almost 200. That's right. Even in NASDAQ, we have stocks that are up. Yeah, Apple, Uber, Meta. And I think that year to date, we have three weeks in and year to date,

Bitcoin's outperforming, small caps outperforming financials. So I think the market does look pretty healthy. I would say it's also very encouraging that if we can close on the S&P above 58.81 by the end of this week, it really solidifies that 2025 is going to be probably a double-digit gain year. What's your favorite part of the market right now, outside of tech? I think financials, to me, represent a pretty good fundamental case of change this year because we have a new administration, a Fed that is dovish, yields that aren't painful for banks, and at a time when it could lead to upside for capital markets activity and multiples are low. So I think financials remain our number one S&P sector idea.

I mean, because I ask you, because your narrative is so levered, so to speak, to big cap tech. Your top five ideas in the market are NVIDIA, Amazon, Meta, Google, and JP Morgan. Yeah. Are you not rethinking any of that? No, at the moment, no.

And even though this one day is painful, I do think markets tend to overreact. And again, unless we're on the cusp of a recession, this pullback in NVIDIA is going to prove to be a buying opportunity as well. Any Fed risk this week at all? Do you have any worries about what they could say? They're not going to do anything, we don't think, but they always say a lot. Yeah. There's a lot of uncertainty going into January FOMC because last month it really caused markets to rethink the probability of a hike. But the probability of a hike for 2025 now stands at 27 percent. I think that's an extraordinarily high probability.

I think there's a chance the Fed sounds more dovish than market expects. Tom, we'll talk to you soon. It's good to see you as always, especially here at Post 9. It's Tom Lee, Fundstrat. Up next, big tech under pressure. As you know today, those companies are gearing up to report their quarterly results in the days ahead, too. Tell you what to watch for, what's really at stake. We'll talk to Intelligent Alpha's Doug Clinton just after this break. All right, welcome back. Shares of NVIDIA sinking by now. You know that story. China's deep seek rattling the market big time. The company shedding six hundred ten billion dollars in market cap, the worst single day market cap loss in its history. Joining me now

to discuss Intelligent Alpha's Doug Clinton, NVIDIA, his top holding. It's good to have you back. Appreciate you coming on, especially today. Not pretty when you look at the screen or your balance, I suppose, in the holding. What are you doing with it? It is definitely painful. And I know one of the things that we've talked about before, Scott, is, you know, I still believe, too, that we are in this two to four year period where it's still an AI bull market. Today obviously makes you question that reality, but I still think things are intact. Obviously, you've talked a lot about DeepSeq and all the ins and outs, the puts and takes of it here today, but the bottom line for me is this. I think about it as the hyperscalers are not competing with

DeepSeq. They're competing with each other they have large balance sheets that they're willing to deploy if those investments in infrastructure in nvidia chips generate better models and i still think they will and so i don't think the ai trade is dead i don't think the chip trade is dead but it may take a little while before investors get more confident with it just to come back at you on your comment that we're still in an AI bull market. Well, of course we are. And you could make the case that today's story only underscores that.

It's just a matter of who the real winners are going to end up being, how much we should be really willing to pay for these stocks if this technology in general is being commoditized right before our eyes and the way that chips are being sold and to whom and for what price is changing? It is. And I think if you look at NVIDIA PE, it was high 20s to your point about price, just a week or two ago. And now it's under 22 right now

for FY26. I don't actually think the $240 billion the street is expecting for fiscal 26 has changed all that much, despite this narrative and despite what you just outlined, right? That's the question is, can we achieve model improvements with less in terms of GPUs? And I think it's a very important distinction to make that DeepSeq has built a model that does inference cheaper right now at current performance levels. I think that's the important takeaway. But it hasn't proven that they can build a model that exceeds the performance of the cutting edge models that we have now in the United States that have been built with NVIDIA chips. So that might not feel like good constellation today, but I think that

is the narrative over the next few weeks. What do these companies say about how they're continuing to invest and deploy large infrastructure to build these cutting edge models that will be better than DeepSeek over time, I believe? I mean, as we've already discussed on this program, the narrative around open source versus not seems to be where the action is and where this story is ultimately potentially going. You look at Meta today, for example, which is up one percent. The only of the hyperscalers, I don't really include Apple. It's just a different kind of AI story. But all of the others are lower and maybe for good reason. Do you think about those who are going to be at the top of the heap as it relates to open source versus all of the

others and how that narrative and how those goalposts might be changing today? Yes and no. And I'll give you kind of the bull and bear component of that. I mean, Meta is a holding for us in our Intelligent Alpha Livermore ETF. So our AI models who pick our stocks, who populate our funds, they are a believer in Meta. And the reason is simple. They have been one of the biggest beneficiaries from AI actually deploying it to their business so far. And obviously they were, I would say until now, probably the leaders in terms of open source AI development. I don't think this is a death knell for them, certainly, and the market seems to agree that I think they can take some of the things they're probably learning now from DeepSeek, incorporating it into their models, and probably jumping back in the

lead in terms of open source in the next few months. All those things are good for meta. But I don't think that means it's the end of the story for the closed source models yet. Because again, I go back to the comment of DeepSeek has shown us that they can perhaps deliver current edge performance for less, but they haven't shown us that they can develop and show cutting edge performance. performance and you look at OpenAI 03 which is their current Frontier model they're going to be releasing that 03 mini to the public soon I think that model is probably still a significant step above deep seek it's still a significant step above any other open source model out there and that really will be where the battleground is is how much advanced reasoning can you build into these closed source models keep that gap ahead of the open source stuff. I still think that will exist going forward. You guys have GE Vrnova, Eaton. I mean, do you at all worry about the power side of this conversation, that maybe that was the most fragile going in? And now we have really legitimate questions about power coming out. It's definitely concerning. But again,

I think a lot of this, there's so much trading going on with AI and a lot of investors that are in these stocks and maybe exiting them today probably haven't done a whole lot of thinking about what is the long-term potential. It's more about the momentum. And clearly, these stocks have lost their momentum and they're going to need to recapture that to get sideline, let's say, investors back involved. You think about GE, Vernova, and the energy needs. I don't actually think DeepSeq really changes that all that much. They've reduced the cost of inference. They've reduced perhaps the need for electricity to power doing advanced reasoning inference. But I actually think it might create a much higher demand for that inference. And I'll give you an

example. I mean, this has been a great development for anybody building an AI-powered company like we are at Intelligent Alpha. We had workloads where we'd have to spend over $10,000 to use OpenAI's O1, and we can run those workloads for about $800 on DeepSeek, and we've been testing that. So if you're an AI company, I actually think it's going to encourage you to use a lot more AI, and that I don't think will cause any decrease in the needs for electricity. No, that seems to be a part of the story that some are looking to highlight today as well. Not necessarily less, but perhaps even more. It's all about cost.

Doug, we'll talk to you soon. Thank you, man. I appreciate you. That's Doug Clinton, Intelligent Alpha, joining us. Up next, we track the biggest movers as we head into the close. Pippa Stevens is standing by for us today. With that, hi, Pippa. Hey, Scott.

In a sea of red, one telecom name is bucking the trend and is one of the S&P's leaders. The name to watch coming up next. All right, we're less than 15 from the closing bell. Back to Pippa now for the stocks that she is watching. Tell us what you see. Well, AT&T is jumping to its highest level in nearly four years after the company's Q4 earnings came in ahead of expectations. The telecom giant said it saw solid momentum in gaining

and retaining profitable 5G and fiber subscribers. Shares are up 6%, and it's the second best stock in the S&P 500. And Estee Lauder popping just now as it reportedly prepares to review its portfolio of brands. According to Bloomberg, the review comes as the makeup giant shakes up its leadership amid a sales slump. That review could include the sale of some of its brands, the report said, and those shares are up 1.4 percent. Scott? Thank you, Pippa Stevens. Still ahead, some key energy players are getting slammed. We just referenced a couple of those. Look at those. Those are all AI plays getting hammered today. We'll drill down on that next. All right. It's closing about market zone.

CNBC Senior Markets commentator Mike Santoli here to break down the crucial moments of this trading day. Plus, some AI losses in the energy space to say the least. Pippa Stevens going to tell us the stocks that are down the most. And Seema Modi on two bright spots. Yes, bright spots in tech today. You heard me correct. There aren't many, though, in tech. No.

What do you make of today? Some software. I mean, what you can say is that it was a rotation more than a liquidation. So if you're going to have a day when you have an absolute washout trade in semis, in the AI hardware trade, where you have one of the strongest themes and most closely held to investors' hearts in theis, in the AI hardware trade, where you have one of the strongest themes and most closely held to investors' hearts in the market get completely pulverized, you could have had everybody just say risk off or out, or it could have gone elsewhere. And what you did see is rotation, people maintaining exposures, more stocks up than down. I mean, some of it was very mechanical, meaning reaching for

the most neglected, under-owned themes. Campbell Soup's up 2.5% today. It's not because it's cold outside. So you just do have this kind of reflex asset allocation type move. So for a day, you're able to absorb it. What I find remarkable is the dispersion, which has allowed volatility, measured volatility to come in. The VIX peaked above 22. It's at 18 now, a four-point drop intraday. which has allowed volatility, measured volatility to come in. The VIX peaked above 22.

It's at 18 now, a four-point drop intraday. I know of some trading systems that say if it goes down three from a peak, it's a buy signal for the market. I don't know if that applies today, but it does show you that mostly it's been kind of skimming the winnings from one part of the market, reallocating it somewhere else. So far, so good. On a day when yields are down, that absolutely helps. Down in yields means up in market breadth. That's been the pattern recently. But we'll see. You

can't take too many more days of this where we're in a comprehensive way, rethinking one of the strongest kind of earnings momentum stories of the last two years. Still at 6,000 on the S&P. Thank you, Apple, Meta, and now Amazon, which is green as well. Now, Pippa Stevens, the AI story is really hitting the energy stocks hard today. You can give us more details on that. Yeah, Scott. So the energy and power stocks that have ridden the AI wave higher are getting hammered, as DeepSeek's promise of a more efficient model is raising concerns about how much power will actually be needed. Now, nuclear power named Vistra, the biggest loser in the S&P, down 30%, with Constellation down 20%.

The stocks are among the top performers in the last year, more than doubling on that AI hope. Now, small modular reactor makers NuScale and Oklo also tumbling, while uranium stocks are down more than 10%. The sell-off is also hitting traditional energy stocks, too, with gas-focused drillers the biggest decliners since they are seen as a play on the AI trade. EQT dropping 10 percent with Comstock Resources, Antero, and Xpand Energy all in the red. Finally, GE Vernova, which makes gas turbines for data centers, getting crushed, dropping some 20 percent. Scott? All right. Thank you, Pippa Stevens.

We all heard you, however, Mike. It's all good. I'll tell you what. I mean, if nothing else, what timing for today's conversation, given the mega cap earnings begin in earnest tomorrow? Yeah, well, there's one element of it that's a little bit tough, which is NVIDIA on the fiscal year doesn't report till March 6th. But in terms of CapEx budgets, why we think it still makes sense to spend tens of billions of dollars this year from Microsoft and Meta, of course, Meta, the open source kind of customer touchpoint of AI, I think that's in a more privileged position. So it is sort of fascinating.

I know this debate is going to just slosh around quite a bit, I think, over the next few weeks. And we're going to get some good things to work with in the interim from Microsoft, at least. All right. Now they have maybe a little bit more to live up to. But as you said, there are a few more sectors that have now gone green. I mean, as Amazon has gone green, discretionary is green. You talked about Staples, which Campbell's Soup is just one example.

That's just a tag end to the market. But also banks, I think. Banks, materials, health care. All day today. So obviously it's the market's way of saying this isn't really an economic event. There hasn't been an inflection point in terms of growth or expected growth. It's much more about whether we just have to kind of rethink the multi-year story about power demand and how many of these boxes we need to build in the desert for data

centers. Well, let's find out, too. If these stocks are in some sort of prolonged falter period, can the rest of the market pick them up? We are going to find out, potentially. There's the bell. So the Dow is going to be near 300 green. S&P, as I said, negative. Now, as you know, that's for Nvidia. The worst day in years. And O.C.