Nvidia’s GTC Conference Takes Center Stage… And Countdown To The Fed Decision 3/18/24

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. NVIDIA in focus. We were watching shares of the AI darling as CEO Jensen Huang take center stage at the company's GTC Developers Conference. We were bringing you all the headlines from the event and the trades in the AI stocks. Plus, bubbling higher. Shares of Pepsi seen their best day since October 2022 after a big upgrade on Wall Street. What analysts think is giving this soda stock a little extra fizz. And later, Tesla shares revving up. Will new price hikes be enough to get the recently slumping stock out of its rut?

0:34We'll debate what is next for this EV maker. I'm Melissa Lee. Coming to you live from Studio B at the NASDAQ on the desk tonight, Bono and Eisen, Karen Feinerman, Dan Nathan, and Guy Adami. First to our big story today, you are looking live at NVIDIA's annual GTC Developer Conference. CEO Jensen Huang speaking right now. He has already announced new partnerships and unveiled a new GPU. Shares of NVIDIA had been up more than 5 % earlier in the day, but closed up less than a percent. They are currently trading down by just a quarter of a percent. Christina Parts-Nobles is at the conference in San Jose, joins us now with the very latest.

1:07Christina. Well, NVIDIA CEO reiterating that we are just at the tipping point for AI, and all that means is that you're going to need bigger GPUs. So NVIDIA revealing its new generation artificial intelligence platform, keyword platform there, to run AI models. They're calling it Blackwell, and it comes with also the Blackwell BH200 GPU. So that's the graphics processing unit, the latest one that we've all been waiting to hear about. It's going to be four times the training performance of the H100, 25 times lower in energy use, and 30 times the inferencing power compared to the previous Hopper architecture or the H100 chip that we've all talked about constantly.

1:48Inferencing, though, is a big driver. The fact that they're mentioning that. We know just in the last earnings call, inferencing contributed 40 % of NVIDIA's data center revenue. So the fact that this chip is going to be even stronger with inferencing allows NVIDIA to maintain that lead. The Blackwell will also be manufactured by TSMC and set to be launched later this year. You have Amazon, Google, Microsoft and Oracle all selling access to this particular chip through their cloud services as well. We haven't received costs on this actual Blackwell chip. Usually at these types of events, NVIDIA doesn't reveal pricing.

2:21But keep in mind, the previous older H100 chip was anywhere between$25 ,000 to$40 ,000 per chip. And you need a lot of them to build your entire AI system. So it could cost anywhere up to$200 ,000. That was the older generation. So you can only imagine that this one will probably be a little bit more than that for a lot of companies if they can get their heads on it. On the partnership front, we know that they've talked about partnerships so far with Ansys, Synopsys. We've seen those stocks pop. on the mention of the companies, Christina. Have there been the cameo sort of star appearances yet from any CEOs or luminaries in the tech world?

3:01Well, I was just told there's some celebs right next to me right now, but I'm not like Ashton Kutcher. Swing that camera around, Christina. There's security everywhere, and that's why my eyes darted while I was going live. But it's just a rumor so far, but they just happen to be right behind the camera. I haven't seen any CEOs just yet, but you mentioned Ansys, Synopsys, Cadence, and then TSMC. TSMC is going to be making that new Blackwell chip, so we could start to hear more. Keep an eye out for Oracle as well and Broadcom and AMD and a few others that may move soon. All right, Christina, thank you.

3:36Christina Parts and Nebulas on the ground in San Jose. It's funny, we were listening to John Ford talking over time about how the similarities between this event and Apple's event. and how when Apple launched a new product, it was sort of like this. And typically it was a sell the news kind of event. What did you make of the stock action in today's session going into this event? I'm glad you said that. So Blackwell is replacing Hopper, right? And as you know, Mel, because we watch it, Hopper is the sheriff in Stranger Things. And the price action has been really strange over the last couple of weeks.

4:09Go back two Fridays ago, that reversal in NVIDIA, where it basically from peak to trough gave up almost a quarter of a trillion dollars. And then today from peak to trough is probably close to$100 billion. I mean, to me, that's not an encouraging sign now. It's a great company. This new chip, four times more efficient. I get it. Price action is everything. And over the last week and a half, when you see moves like that, you absolutely have to take note. You know, it's really interesting when you think about that comparison to those Apple events. And they were fabulous, right? There was always a whiz-bang new thing that a consumer could go out there and buy, and they could have it in their pocket.

4:43And it also happened in and around, I think it was 2012 or so. These events were just kicking off. I know that, you know, Tim Cook had just taken over. The company had started buying back their stock. So consumers were also buying the stock. They were buying the products and buying the stock. One of the biggest differences here is that most consumers don't have a clue what a GPU is. They don't understand what a data, I don't mean that in a condescending way, a data center and all this sort of stuff. This is a product that is powering something that they actually haven't even used that well yet because they're still hallucinating a lot of these things.

5:12So I just think it's kind of interesting. And then I bring it back to the company. I bring it back to their earnings. I bring it back to their sales. In my career, on any sort of scale, I've never seen a company go from$3.50 in earnings two years ago to$12 in earnings to a$24 expected earnings. And then you can do that as far as their sales from$25 billion to$50 billion to an expected$100 billion. It's never happened. Are you saying that you don't think it's true? It did happen. It did. I mean, so what I'm saying is what comes this year, this is the one where it's up again, up 80 percent year over year after we've already seen up.

5:44So what I'm saying is, is that in the last two and a half months, we've seen this company, this stock double. OK, on expectations of all this stuff that we're just hearing about right now, all these names that don't mean anything to most consumers. Right. And we don't know what their demand is going to be like for them when they finally come to market. We don't know what the margins are going to be. We don't know what the competition is going to be. I'm just saying. So I just think we're at a weird spot. A lot of you folks say, well, you said that to me six months ago. You said that to me. Right.

6:10You could have said that a year ago. Okay. And I've also said to you, what's happened here, we've never seen happen before. But if you're going to make that same bet right now, right here at$900, that it's going to continue to go that way, that's a bad bet. But let's play the game that the competition is coming, that AMD is making a product, that Intel is making that product, that they're also offering software solutions, same as NVIDIA. NVIDIA has the lead at this point. You two are NVIDIA holders. So I'm curious, what's your take? At what point do you start getting concerned that that competition is actually real and that share will actually I mean, share is already going away from NVIDIA.

6:44It's just that NVIDIA still has a dominant share at this point. True. But I think this is a situation where the pie is growing tremendously. Right. There's room. I remember early, early on in Amazon's cloud business thinking, well, eventually there'll be competition. OK, so that was years ago now. And the cloud business has just continued to grow and actually accelerated growing at various points. So I don't think that there will be no competition. I think there will. I'm sort of more curious about, all right, so now they have this, all these embedded H100 GPUs, which were anywhere of$25 ,000,$44 ,000, I think for some of them, maybe more.

7:22What are those worth now? OK, so if you are, I'm wondering, if you're a meta and you own$8 billion, dollars. And these are however many. Do you mark those down? What happens to that? I don't know. I'm curious what happens to that before they get to these. Now, these may be wildly more efficient. I'm confused also by that 25 times more efficient. Does that mean they're 4 percent the power usage? Is that what that means? Or are they 25 percent less power? I don't exactly understand. So there's I mean, the other thing that I'm sort of intrigued by here is, aside from being a genius at this business, Jensen Wong is also a genius at being a CEO of a company that is in the middle of a huge wave and how to manage expectations and sort of how to manage the stock price, which is a very difficult thing to do.

8:12And the higher it goes, Dan, the harder that gets, of course, right? So he's done a masterful job so far. I am long. I'm staying long. I do think that we are still in the early innings of the story. There's a lot left to play out. I think that absolutely AMD and long some AMD as well, which is actually more expensive. I do think they will have some share, but I do think the pie is just growing tremendously quickly. Yeah, I would agree that the pie is growing. And Dan mentioned the fact that most consumers don't know what it is, but like a lot of the knock on the stock was that they had that consumer concentration.

8:46And what you're seeing is with these partnerships, them at least bolstering and addressing that issue. Additionally, I mean, I think a lot of the knock on the company going in was that, listen, when it comes to training these large language models, this is really the only show in town. But once we get an opportunity to kind of pivot away from them with inference, we're really going to see cutting into margins and diversification away from them. But you've seen 40 percent of their revenue last was inference. So the fact that they are starting to broaden out their revenue streams and become a bit more of a household name, not to mention the software.

9:16And to me, that's really where the mode is. That's really where the vertical stack is and why this company has expanded from, what was it, 22 times to now 35 times. So I'm with everyone in terms of, yes, expecting them to continue to grow at this rate. No, at some point, at some point, that's not going to happen. The question is, at what price? OK, as I said before, I do expect there to be some consolidation in the name. But there are a lot of unknowns. None of us. We may not know what a GPU is. We also don't know what the overall target adjustable market is or what what this whole AI thing really is.

9:51And until we get more of a grasp on that, saying that we don't understand the opportunity, but that we do understand the risk to me is really kind of like speaking out of one side of our mouth. So I'm willing to grasp the unknowns right now as as things stand. They are the axes, picks and shovels and now inference in this gold rush. And I think you stick there. I mean, this is a developers conference, right? They want people to develop their models and train their models on their hardware, on their software. They want to make it harder to switch. And that's sort of the comparison to Apple. I don't want to draw that out too much.

10:26You didn't like it? No, no. It was fine. It was fine. It was fine. But that's the whole point. So that's their advantage, right? That's their probably three to four, five-year head start with everybody else. But, again, price action is important. and we outlined the price section over the last week and a half. And at some point, you're going to look at peak margins. And maybe this last quarter wasn't it, but 77 % margin. And again, we've tried to do the math. Karen does it very well, but price to earnings, cheaper than AMD without question. Price to revenue, a lot more expensive than AMD without question.

11:01So at some point, they're sort of out-earning, they're out-kicking their coverage in terms of their revenue and in terms of their earnings, which means, in my opinion, that 77 % is going to start to come down. Now, the question is, is that just a natural progression? Will the market give them sort of a pass on that? Or is that sort of the inflection point for the stock? It clearly has been it's been heading the right direction. At some point, that 77 percent stops starts going down. That's when people get concerned. And that's been I mean, that's been true with technology and chips all since chips were started to be made and traded.

11:35For more on what this could mean for AI, let's bring in ARK Invest chief futurist, Brett Winton. Brett, good to have you with us. Have you heard anything out of Jensen Huang so far that, you know, changes your mind about the story? No, but I mean, I think a lot of the panel covered the key issues here. It's clearly an amazing company. I think that there's a lot of demand for AI chips. We think that$14 trillion will be spent on AI software by 2030, and you're going to need trillions of dollars of AI chips per year to power that. And we underwrite that into the stock. And still, it's a difficult valuation case to make over the course of the business cycle relative to other opportunities that are out there in the market.

12:20How do you think about the sort of halo effect that spending on AI chips will have in terms of broader tech spending? Wedbush had an interesting figure saying that for every dollar they estimate is spent on H100 chips. There's$10 to$12 being spent on software and other sort of parts of the tech ecosystem that you need to make the AI chip work. So what is that sort of trajectory in your view? Yeah, both on the kind of like powering ecosystem of tooling to have AI chips work and on the software that needs to be generating revenue on the back end to justify the capital investment in the AI chips.

13:01We think the foundation model layer companies are really profoundly interesting. I think that that's probably a new emerging operating system for computation. And it puts a lot of the tech heavyweights at risk. There was news today that reportedly Google and Apple might team up for Apple to license Google's Gemini model. I think that's a sign that these companies actually don't get that this is a new operating system that's emerging. It's not a feature to layer on top of existing tech platforms. Brett, it's Karen. Thanks for being on today. So it sounds like there are other names that you might like more than NVIDIA.

13:36Can you tell us what those are and why you like them? Sure. On the private side, we think Anthropic and its AI model is supremely interesting and that enterprises are scrambling to figure out how to deploy these things to generate productivity for their knowledge workers. Then on the public side, Actually, Tesla is the most compelling AI story in the market. You're not being honest if you don't think that autonomous driving is not more likely to happen today than it was two years ago because of the advances in AI. And they have an amazing distribution network of all their vehicles and their business model transforms if they can turn kind of their autonomous robo-taxi software into a working commercializable product.

14:21Along those lines, Brad, since you are a futurist, when you're thinking about future ways to play AI, you know, there was an analyst on this morning and he had an interesting take on it. You know, when refrigeration was invented, it wasn't the refrigerator parts makers that became the winners in that all. It was Coca-Cola. So when you sort of apply that metaphor to this situation, what are those kinds of companies that you think down the road will benefit the most? Well, I think there's a clear opportunity on the consumer side, and its entertainment is going to be transformed by having hyper-compelling AI models that captivate our attention.

15:00So think of TikTok as the algorithmic feed, which Facebook or Meta has co-opted. Well, imagine generated media content and how compelling that will be for end users. And we think that Meta's open sourcing of its AI operating systems is actually an interesting angle that they'll be able to play to generate that really compelling consumer experience across their entire portfolio of apps. Wow. Brett, thanks. Good to speak with you. Brett Winton of Arc, the chief futurist there. They have junior futurists, too. Is there a more junior? I don't know. Maybe. I don't know. You look into the near term.

15:36I don't know. So when do we move from picks and shovels to the bigger picture? Because that's really the question. If enterprise is going to continue to spend on it, you want to see that there are gains, whether we produce activity gains, product gains. Well, I mean, listen, I think most folks that we've had on over the last few months or so, I mean, Meta has been a universally liked name, how they are using generative AI to aid their own business, how they're basically monetizing the ads that is their business. And, you know, I don't know if you guys noticed in your Instagram search feed over the last week or two, you're starting to see meta AI as part of the thing there.

16:09So it's working throughout the consumer sort of products there. And then you're also getting served better ads. And for the most part, I know that most people think that Instagram's pretty good on the ad front. So they're going to have better monetization. That's why that stock keeps working. I know it was one of your early picks probably before NVIDIA and how to play Gen AI. And the one that continues to be disappointing was Google. And so ultimately, hopefully someday they're going to get it together. But right now, even with today's gains, it doesn't seem like that's a great validation because Apple has no strategy is looking to them rather than an anthropic or doing a deal with an open AI.

16:44Listen, I think that when we kind of when we kind of dig into this Apple Google possible partnership, the headline and the initial gut reaction is for us to say this probably is not good for Apple. And I had the same type of reaction. And then I thought about when I think about the distribution, when I think about the Apple Watch, when I think about kind of the GLP ones and then and all the focus on fitness, I actually think in the long term, maybe a few generations down, they actually end up winning again because of distribution network and just because of just how ubiquitous that name is. And from an application standpoint, to me, I can see how having a real-time suggestion on what one should do to solve a certain problem that comes up in real time, that seems to come to mind for me.

17:28Quickly in the here and now, though, so much of the broader market is predicated clearly on the NASDAQ, more specifically on semiconductors, more specifically in NVIDIA, which is 27%, I think, of the SMH. And, you know, you can pull up an after-hours chart here in NVIDIA. I think it's trading about 865 or so. So if we came in today waiting for something, and this is at least at 8, 5, 17 at night, a sell the news event, then one has to think, OK, what does this mean for a broader market, which has been lower left to upper right for quite some time? So just keep your eye on the price section here, as we said earlier.

18:00All right. Do not miss the CNBC exclusive interview with NVIDIA CEO Jensen Huang tomorrow, 10, 15 a.m. Eastern Time on Squawk on the Street and then with Jim Cramer on Mad Money at 6 p.m. Eastern. Coming up, a lot more on the AI race, how Apple and Alphabet could be coming together in the most unlikely of alliances, and what it will mean for the iPhone in your hand. We're on that next. Plus crude copper in rates, oh my, all three ticking higher as investors await the Fed's next move. Why our next guest says to prepare yourself ahead of this week's policy meeting. Don't go anywhere. Fast Money is back in two.

18:32This is Fast Money with Melissa Lee, right here on CNBC.

18:43Welcome back to Fast Money. Google could be on the verge of a deal to bring its Gemini AI engine to iPhones. Bloomberg reporting today that Apple and Google are in active talks to license Gemini for new software features coming to iPhones later this year. Sources also saying that Apple had recently held talks with OpenAI and has considered using its model. Google parent Alphabet up more than 4 % for its best day since December, while Apple closed half a percent higher. So it did give up the gains of about one and a half plus percent during the session. I had your same take in terms of, wow, Apple can't get it together to have its own AI offering by the developers conference, which is widely expected.

19:23Yeah, I mean, again, that was my initial initial reaction, so to speak. Right. Like, OK, so if all the the major players have been investing here and allocating resources here, what has Apple been doing? And I think in the shorter intermediate term, that probably is what the stack's up to be. But again, I do think that they have the stack and distribution necessary to be able to take advantage in terms of the presence with consumers in the long run anyway. So I think perhaps it is a short-term headwind, but ultimately I think they're able to play through. I'm sort of wondering why build it if we can buy it, right?

19:54We don't know how much it costs. But they're not buying it. We don't know what they're doing, actually. Right. We don't know. We don't know. We don't know how it's going to be. If Google is paying Apple. So we don't know. So it is quite possible that this is just a more efficient way to do it. Right. And maybe that's the simplest explanation, not the right one. I don't know. But that's sort of what popped into my head. I mean, we always talk a lot about Apple just making whatever it is better, not first. Right. But they do it better. Well, think about this. It's probably an extension of this existing deal that they have for Safari on iOS devices.

20:27Right. So basically, you know, Google pays Apple 20 billion dollars a year, 36 percent of the search traffic that goes on Safari. devices, largely on iOS devices. And you think about that. OK, so right now, if you go to Google and you do a search and you subscribe to Gemini, Gemini is powering that search the way OpenAI is powering the search on Bing or something like that. So it has the potential just to be a stop gap to your point. You know what I mean? And an extension of the existing search deal. And it might be the sort of thing where maybe they're not spending a whole heck of a lot of money.

20:57It's interesting. I don't know if they can pull up a day chart of Apple right now. So this stock gapped up at 175 and it went a bit higher to 177 and a half sold off. It literally flatlined at 175 all afternoon. And at the end of the day, it just dropped. So whoever was buying that stock all day, it was kind of interesting action. So the fact that it closed on the low, open on the high, not like that encouraging for the way at least investors are thinking about this deal. If the model is the current relationship, then this product, whatever it may be, I would think would be an Apple product to any consumer using this.

21:29And that therefore is a stamp of approval for the Gemini product, you know, a vote of confidence that Google can actually get its act together, fix whatever needs fixing and tacit approval. Absolutely. And talk about trading stocks again. So Google, I think, traded up to 152. So 153 in chain. So let's call 154 sort of the line in the sand. That was a prior high a couple of years ago. It was a prior high we just made seemingly a month, month and a half ago. You can do this one of two ways. You can do what Karen does and say, this is a value stock. I'm staying with it through hell or high water.

21:59That's been right, it's hanging in there. Or you could say, you know what? Price action wasn't great. I'm doing one of two things. I'm going to buy it on a breakout above 154, which sounds somewhat counterintuitive, or wait for the pullback, which we've been in the midst of. Because right here, to a certain extent, you're almost flipping a coin. Yeah. I think we want to get back to Christina Partzinevelis with more from NVIDIA's GTC conference. Christina? Well, one of the things we wanted to hear about is how they're going to monetize software. And NVIDIA just announcing that they are launching a new service.

22:29So it's an enterprise software subscription. It's called NIM, which stands for NVIDIA Inferencing Microservice. And what that means is that you can take NVIDIA GPUs that you already own, and they will provide the extra software and allow companies to build large language models on said GPU. So instead of using OpenAI, for example, you would turn to NIM. So you would buy the GPUs for a cost of anywhere between, let's say,$30 ,000 to$50 ,000. And then per GPU, the new cost will be$4 ,500 per GPU per year for a license. So again, this is a way to monetize software, a reoccurring source of revenue for NVIDIA on top of the hardware that they're already selling.