In this podcast, Motley Fool analyst Tim Beyers and host Deidre Woollard discuss:
- Nvidia's strong quarter and the cyclicality of chip demand.
- Why Nvidia isn't the only game in town for artificial intelligence (AI) chips.
- How Snowflake's data warehouse solutions might grow over time.
To catch full episodes of all The Motley Fool's free podcasts, check out our podcast center. To get started investing, check out our quick-start guide to investing in stocks. A full transcript follows the video.
This video was recorded on August 24, 2023
Deidre Woollard: Nvidia's stratospheric run isn't done yet. Motley Fool money starts now. Welcome to Motley Fool Money. I'm Deidre Woollard here with Motley Fool analyst Tim Beyers. How are you doing today, Tim?
Tim Beyers: Fully caffeinated. Ready to go Deidre.
Deidre Woollard: Well, I think you have to be fully caffeinated because we've got a little rocket ship thing happening with Nvidia. I was one of those weirdos last night, like wait till like 04:00, 4:00 or 5:00, refreshing the Investor Relations page to see what happened. I think I was not disappointed. Revenue up 101% year over year. Data center, sales growth, just crazy. Gross margin was up 26.6 points. This is so good it makes me nervous. Tim, should I be nervous?
Tim Beyers: No, I don't think you should be nervous. But I will say that the market seems to either have priced in most of the growth we're already seeing, or there's just general unsettledness around the market, and so the stock isn't rallying the way you would normally expect it. As we're recording this, Deidre, Nvidia is up only about 3.5%. For results like this, you would expect, maybe even three months ago we would have seen like a 20% or even 30% move. But we're not seeing that. The market's either priced and a lot of the growth we're seeing, or there's just some general unsettledness. But overall, you really can't do much better than this. The data center business, as you pointed out, this is primarily where Nvidia is getting its growth from. Revenue in the data center business was up 171% from a year ago. It was up 141% sequentially, that is bananas. The second-quarter revenue from gaming at $2.49 billion. Again, these two primarily account for 90 plus percent of all the revenue that Nvidia generates. It was up 11% sequentially, 22% year over year. Primarily, Deidre, almost all of this growth is coming from the idea that the world needs generative AI, and Nvidia is going to provide that generative AI. In this latest quarter, boy, did they provide quite a lot of it.
Deidre Woollard: But is there a concern that they're getting over their skis on how much of it they're providing? Or is it just that we're going to see demand coming from places that maybe we hadn't even considered before?
Tim Beyers: Well, I would say it is likely they will get over their skis at some point because it's very hard to, absolutely, particularly in this business, where hardware orders come in and you take a while to build out the inventory to meet that demand. If you're pre-building inventory and then suddenly demand slackens, which does happen in this business, you see things like inventory write-downs, you see sharp reductions in margins, things like that. That's happened to Nvidia before. It could very easily happen again. Nobody knows exactly when that is going to happen. For the moment, it does appear that demand is going to continue at a very brisk pace for quite some time. What we're seeing is that, to the point you just made, it may be from unusual areas of the market. But overall, there's just such an insatiable appetite for feeding things like large language models, and we should be, without getting too far down a tech rabbit hole here, Deidre. But large language models are hungry. You can think of them as the Cookie Monster of data center workloads. They just want more. Usually, it's for things like memory, very high throughput processing equipment where there's tons of memory attached to it. I remember on one show for This Week in Tech with Tim White, and he was talking about how he had loaded a local large language model onto one of his machines, and just the amount of memory it was taking, just the heat that it was causing inside of his machine is like one large language model almost forcing a meltdown of a software developer's machine and an experienced software at that in Tim. These things are very hungry. Because they are so hungry, we can expect demand to continue at pace for quite some time. We just don't know when the tipping point comes. The tipping point is when we start thinking about, we've done all kinds of training, we've been consuming just data set after data set. At what point do we start sharpening our focus to maybe smaller sets of data or different opportunities. The volume is not so vast. When that tipping point comes, Deidre, that's when you do see the risk of Nvidia getting over its skis.
Deidre Woollard: I think it's interesting too, because you think about the total addressable market also seems to be rippling out though as well.
Tim Beyers: I think what we're seeing is that there is a general belief. I think this is partially naive, but I'll say it anyway. That large language models are the key to unlocking big value in a lot of different industries. I think that's partially naive because a large language model in and of itself really doesn't do anything. If you feed a large language model, a lot of data, the value that you get back is really going to be, in some ways, equivalent to just how good that large language model infers useful conclusions. As we know, we've seen a lot of instances where large language models, ChatGPT being chief among them, decide to hallucinate some things. Those imperfections in inference are going to continue to be something that we contend with. There will be probably a quite a lot of work out about how to make data sets smarter. But in the meantime, we are going to just chew through a whole bunch of data and then decide later where we get the greatest value from chewing through data. As we start to winnow that down and figure out where the greatest value is created, then demand will start to slacken. But I would not be surprised if over the next year, demand continues at a very brisk pace. A triple-digit growth rates? I'm not so sure. But for a while, yes, I can see it, Deidre.
Deidre Woollard: It's interesting because I think there is what you pointed out there. There's a gap between excitement and utility. Right now we're in the excitement phase. We're going to get to the utility phase eventually.
Tim Beyers: When we get there, demand might fall off a cliff. Let's be clear. But it's more like figuring out where to put dollars to work most effectively. Because you're exactly right, we're in this very hype driven phase. We don't know where we're going to get the most value for the dollar committed. Right now is we are just committing dollars. We need more cookie. That's what we need. Honestly, I swear large language models are like software versions of the Cookie Monster. With all apologies to Sesame Street, because I love Sesame Street.
Deidre Woollard: Who doesn't love the Cookie Monster? Well, you know, it's interesting that the market hasn't gone crazy over this, because last quarter, they went crazy over the guidance. This quarter, Nvidia came through with forecast of $16 billion in revenue for the next quarter, even bigger. Why did the market not respond? Is it just that this quarter is a little weird about guidance? Because that's what it feels like to me.
Tim Beyers: No. I think the expectations are baked in. In my opinion, you'd see the stock moving even with the market down today as we're recording this. I think you would still see Nvidia moving a lot more if the expectations had been more muted coming into this quarter, but I think the expectations were absolutely enormous. There is just isn't a lot of institutional money that is piling on. Instead, what the action suggests is that there are certainly some institutions that are buying, but there's probably some institutions too that are taking some profits and saying, wow, this is amazing. I know that this is an incredible growth story here, but I've got some big profits on this. How about I take some right here because there isn't such an overwhelming belief, or maybe there is, but we're not seeing it in the stock price action right now at overwhelming belief that growth will continue at this pace for an extended period of time. I think we would have seen a much more bullish reaction if this really shocked people, and they left convinced that the growth story is much longer and much more durable than expected. I think we're still feeling this out and wondering just how durable it can be. To be clear, some of that is justified. We have been here before with Nvidia. We're not that far removed from the crypto days when everybody needed an Nvidia GPU to mine crypto. Aren't you mining crypto? Why aren't you mining crypto? That was a big deal. Then that demand went away, and it really did leave Nvidia longing for growth that had gone missing. There may be some people who remember those days and are being cautious. We don't yet know what we don't know about the growth story here. But I would say it looks positive. It just doesn't feel like the market has decided that, this is an unrelenting growth story and this is going to go on for years.
Deidre Woollard: You make such a good point about the crypto and it makes me think about the CEO, Jensen Huang. I listened to those calls during the crypto era, and he always generates this cool confidence.
Deidre Woollard: He's so believable and there is this quality to him that makes me want to trust him. He's never super super hype, but he's very confident. He's sounds now the same way he did then. He's talked about generational shifts and this is just the beginning. What do you feel about the way that he presents these types of opportunities?
Tim Beyers: I think he's right in that there is the potential for a generational shift here. But if there is a generational shift, we're pretty early in the shift. Treat all predictions with a giant cautionary sign, above them blaring and at least yellow lights. Proceed with some amount of caution, but also proceed with some optimism here because there is reason to believe that AI demand will continue for some time. I think that demand will sharpen and Nvidia will not be the only beneficiary here. I do think there are other ways to look at this, like will they be the only AI company where we'll be only be doing generative AI compute with GPUs and there's no other option for doing that. I don't think so. What you really need if you're talking about generative AI compute, you're talking about machines that can process in parallel at very high rates, lots of throughput. You're going to need lots of memory. It's not going to be one silver bullet that solves it. If it's a generational shift, Deidre, that means it's an industrywide shift that Nvidia will participate in and maybe even lead, but they will not be the only ones. What does that mean for companies like AMD and Intel? It's way too early to tell. I would put more money on AMD than I would on Intel, only because I think Intel is a little bit earlier in its transition to a company that's also building out foundries as well as updating its roadmap of data center processors whereas AMD is not anchored with that.
They are very much building highly advanced, very fast datacenter chip sets, and they've been doing it at scale for quite some time now and they have integrated their Xilinx acquisition from a few years ago, and figuring out how to take the core AMD chipset, put a field programmable gate array with that and that essentially all that means Deidre is that those chips can be purpose built. They can be designed straight and programed straight into the silicon to do something very, very specific, which makes them super useful for datacenter compute. Sure, you could see real tailwinds here for Nvidia. But are they the only ones? No, not by a long shot. You could see AMD profiting from this as well. We just don't know when the spend starts to sharpen. But I think if we assume that this is going to go on forever, the spending at the present rates, then I think we're deluding ourselves. It will sharpen at some point, maybe 12-24 months from now. But I'm not going to get into the predictions game here. I would just say, if I am holding AMD shares and I am responsible for a portfolio that is holding AMD shares, I'm not looking to sell right now, but I am looking at how the market treats that position. If it becomes just irrationally priced, as if growth will go forever, then I might have to take some off the table, but I'm not ready to do that yet.
Deidre Woollard: With Nvidia having a bit of the jump ahead, is some of the partnerships that it's already building and does that create any moat against those other competitors?
Tim Beyers: Is it a moat? It certainly provides some protection. Sure, the ecosystem around Nvidia is pretty interesting. They also have lots of software that is highly useful. The software that goes with their AI systems is particularly useful and that does give them a bit of an advantage. They also have partnerships with lots of different big players. Of course, they want Nvidia GPUs and systems in the large public clouds or partnered like VMware is a recent partner here and they want you to be able to maybe rent out or get access to a virtualized CloudFormation powered instance that is backed by a set of or instead of Nvidia systems or processors and you can rent that out and that is very efficient. That partnership makes some sense that you could go get access to the compute power you want on the terms that you want it. They want to do those deals, I'm certain, with all of the major public clouds as well. Nvidia is in the process of seeding the market and trying to become the standard for generative AI compute, at a moment when the industry has a real hunger for generative AI compute. In some ways, yes, those ecosystem partners give them a lead in developing that market position of being we are the go-to. Look at all of our partnerships, we are where you go when you want generative AI compute. Certainly there's something to that, but they won't be the only ones.
Deidre Woollard: Make sense. I want to pivot to Snowflake because they reported yesterday afternoon as well, getting eclipsed a bit by the monster that is Nvidia but they had a really solid quarter, solid results. What Snowflake does, as I understand it and you can correct me, but they're provider of data warehouse solutions and so they have their own role in this AI space because as you talked about Cookie Monster, you've got all this data to wrangle. But they had 37% year-over-year growth, strong remaining performance obligation. They're clearly connecting with customers. What should we know about Snowflake?
Tim Beyers: They had a very good quarter, but it's certainly not as blowout of a quarter as I would say Nvidia had. It was much more moderated, and the stock is down a little bit because the market has sort of given back some gains today. That's not too surprising and actually that's completely fine, but overall product revenue up 37%. But I'll tell you that things that really stood out to me, Deidre than I thought were super-important. The first is that this is a company that's defined by getting customers to increase their consumption and their reliance on Snowflake over time for storing, analyzing, processing, and using data. So one way to think about what Snowflake does is a very advanced data warehouse. A data warehouse is where you put information, and it's very well organized. Think of a physical warehouse and everything is marked and categorized, and you know exactly where it is on the exact right shelf, and all that thing that's a data warehouse. Very well orchestrated. Snowflake has this and it's in the cloud. They have been very successful at convincing large companies to say go with us, we are independent. We'll let you operate in any cloud that you want and store your data here, and we'll let you do things with the data right inside your Snowflake environment. You don't have to export it anywhere, which leads to more usage of the Snowflake platform, and we have seen that really compound growth inside the Snowflake platform.
As of this latest quarter 402 customers spending $1 million or more annually. That was up from 246 over the same quarter last year. That's up 63%, Deidre. That is what I want to see. When you have those very large customers, growing at that rate far faster than overall revenue, meanwhile, the net retention rate is 142%, that's down from where it's been. It's been in the 170s, but that's still massive. That's 42% more than these existing customers are spending on Snowflake year over year. And they've been doing this for a while now, like really compounding their spending. The thesis is they're going to get more of those really large customers, and so to see that is highly encouraging. Their overall customer growth was up 25%, like I said, the 142% net retention rate. Then a couple of other quick things, marketplace listings, which is Snowflake customers who have data that they are making available to the market through Snowflake, those listings were up to 2,149. That was up 39% year over year. The Snowflake platform is allowing customers who have data to do more with it. That is also a very good sign, and one more, 26% of customers now have at least one stable edge and a stable edge is when two Snowflake environments come together and there's data-sharing in-between those two. You think of them as an edge and another edge and they come together, and they share data and there's no integration that has to happen throughout the transformation, nothing like that is just two Snowflake environments cooperating.
When we see more Snowflake customers cooperating, working together, it makes the entire platform more valuable. Seeing these performance indicators, Deidre, gives me a lot of hope that we're going to see high growth for a much longer period of time. It's a little bit of the opposite of Nvidia. Nvidia, I think is going to be really high-growth for a period of time. Then there's going to come a point where that growth is probably going to go off a cliff, and really slow down significantly because the demand curve just drops off. We've seen that before, it wouldn't surprise me if that happened again. With Snowflake as they build these relationships, more data being used by more customers, and more Snowflake environments operating together, more stable edges, more marketplace listings, all of this type of stuff, more consumption at the highest possible levels. That convinces me that Snowflake can grow at a very high rate for a much longer period of time than the market expects. If you could visualize Nvidia just climbing like a roller coaster and then dropping off. Not because that's bad, it's just because you would think that's natural. It's a cyclical business.
Deidre Woollard: Yeah.
Tim Beyers: For Snowflake, I would say slow and steady, but a pretty high growth rate up and to the right and a slowing and maybe even slightly increasing or maybe slightly decreasing curve, but very high growth, and that line goes out much further than we might even expect.
Deidre Woollard: Interesting. And it's interesting seeing these results. We're still in this, what they call the elongated sales cycle.
Tim Beyers: Sure.
Deidre Woollard: Where their customers maybe aren't making those sales decisions as fast as they used to. As as that changes because everything always changes, is that when Snowflake starts growing again, because it seems like these two companies, they have a connection, but maybe the sales demand or the sales cycle is a little bit different.
Tim Beyers: It's dis-aggregated. It's going to be a little further out for Snowflake. In this part of the market, which will probably call enterprise software or enterprise cloud spend, there are customers that are deciding how much do they want to spend and where do they want to spend it, and so they're optimizing what are the tools where they get the biggest bang for their buck. They're slowing some spending on different tools. We saw this for example, with Datadog. We're seeing it a little bit with Snowflake and so the growth is slowing a little bit, but remember this is still up 37% year over year. That is heady growth. May not be triple-digit anymore, but that is still very heady growth. There's a bit of optimization happening right now where that is not happening at all for Nvidia. But there will come a point, Deidre, where the reverse becomes true. Where we have had a just a flood of spending to feed these large language models and then we get some optimization or like, OK, how much hardware do we really need? That part of the market gets optimized. Whereas then at that point was Snowflake, like OK, we've optimized things. Now, what are we really going to do? With Snowflake, and then you see maybe a little bit of reacceleration. Let's start to get too crazy because again, 37% growth is really high-growth, even if it doesn't reaccelerate, if it stays relatively steady for a long period of time, that is a really good outcome for investors, Deidre.
Deidre Woollard: Yeah, absolutely. I got to wind us up on us a silly question, which is, so Snowflake, they want to get their leadership really excited about the DataCloud. They're doing this thing that I think is interesting. They've got this DataCloud World Tour. It sounds like a concert, 26 cities, three regions. What is this? Is this just to get people excited about Snowflake? I thought it was an interesting marketing spin.
Tim Beyers: Yeah, I think it is. We've seen this before. It's probably a little too high peak, but also when you go on these tours and you are meeting customers where they are, that can be a very good thing. Like essentially it's a giant rotating sales conference. You're just bringing customers in and talking them through what their workloads might be, what their needs are, and trying to get face-time with them. I don't have a big problem with it. It does feel like unnecessarily rock and roll. It feels a little bit strange and it feels like the 80 year olds are getting out on Toric and the bands back together. It feels a little bit like that. But honestly, it is getting in front of customers that really does matter. I don't have a big problem with it, [MUSIC] but I do see how it feels like a little, wow, OK, this isn't, you're not really rock and roll. We know who you are. We know you sell data warehousing software. Let's just calm down.
Deidre Woollard: It is not Taylor Swift.
Tim Beyers: It is not. This is not the Aerostar.
Deidre Woollard: Nope. Thank you for your time today Tim. Great to see you.
Tim Beyers: Thanks, Deidre.
Deidre Woollard: As always, people on the program may have interest in the stocks they talk about, and the Motley Fool may have formal recommendations for or against, so don't buy or sell stocks based solely on what you hear. I'm Deidre Woollard. Thanks for listening. We'll see you tomorrow