NVIDIA Corporation (NVDA) Earnings Call Transcript & Summary

Why don't we go ahead and get started? I'd like to welcome everyone to this morning's session. It's my distinct pleasure to welcome to the stage, the management team of NVIDIA Corporation. To my left is Colette Kress, Executive Vice President and CFO. Also in attendance in the audience, Simona Jankowski, who runs Investor Relations for NVIDIA. Our format is fairly straightforward. We're going to have a 30-minute fireside chat in this room. If you have any questions, there are mics that are being sort of walked around, there's one stationary over there, please raise your hand and we'll try to get to you the -- but I'll start things out. First, Colette, I really appreciate you. It's great to have people live after a year hiatus, and I really appreciate you guys participating this year. So thank you.

The first question I tend to ask is really a setting the stage question. I think since the last time you were at -- on stage here, the stock is up sixfold. I think your stock is up 150% this year. Help us remind -- or at least remind people in the audience, what's the vision, the mission statement, that Jensen and you have for the company? And specifically, what are you doing to try to fulfill kind of that vision, especially around that recent acquisitions like Mellanox and the proposed acquisition of Arm?

Sure. Great question to start off with. And thank you again for breaking the ice and holding the very first in-person investor conference. We're glad to be here. Our vision of NVIDIA may not have really changed over a long period of time. We set out with the journey of helping in accelerated computing. The very first app that we found that could use accelerated computing was gaming, and it's still in our heart and soul as a company as we continue to improve that. But what we did find was other great use cases and platforms for accelerated computing. One of the great outcomes of accelerated computing has been AI and building up. We look forward and think of us as a full stack data center computing company. When we mean a full stack, that means we start layer by layer. First layer that we think about is those chips, those systems and build up to the second layer, which would be our platform layer that is focused on the engines that are important to so much of that acceleration. Everything from RTX to AI to simulation, many other types of engines are very important to us. And then lastly, on our top part of the stack is software. We will enable software frameworks towards applications for different industries. And that's how we have designed this full stack. So Mellanox has been added to that because we really think about the overall data center computing world as looking at the full data center end-to-end, every step along the way of data and networking is part of that. They've been a great addition to us, and we'll continue to build upon that strategy for future growth as well.

A lot of company-specific questions, but I'd like to start off with kind of more of a bigger picture question about the state of the industry today because clearly, we're in one of the worst supply shortages for semiconductors that the industry has ever seen. If I look at kind of what you guys have done over the last couple of quarters, your purchase obligations were up 45% sequentially last quarter after being up 40% sequentially the quarter before, after being up 35% sequentially the quarter before. How tight is the current environment? And when do you expect sort of supply to actually start to accelerate so that you could actually meet all the unfulfilled demand out there that you see?

Our largest issue that we deal with is demand is greater than supply. Demand is greater than supply. We've been working all through the last 18 months focusing on supply but really understanding that the world has changed in terms of how they need to think about procuring supply. It ranges from everything starting with wafers to substrates, to components, to contract manufacturers. We talk about the distribution, the logistics, all companies really have to think through that. It's also taking the full ecosystem. You have to think about your supply ahead of your products. It is also thinking about the supply of the systems that you may enable downstream. And that's when we started focusing full end-to-end on our ecosystem of doing that. We have also, therefore, started to procure longer term. Most of our discussions with our suppliers has been what do we see going forward. How can we help them and enable them to build the capacity that they will need to support the system going forward, helping them with that and working on long-term procurement. You will see now in our purchase commitments. You will see that in our inventory. How long will it take until this balances out? It's difficult to say. We feel very good about, of course, our guidance that we have provided. We feel good about the beginning of next year and what we've computed. But when we think about the second half of next year, I think we'll be in a great position given so much of the work that we've already done on long-term procurement.

You and I talked about this on the call back after your recent quarter. In an environment where supply is this tight, do you just kind of take seasonality and throw it out the window. Is growth over the next couple of quarters really a function of how much supply you can get, question number one. And the second question around that is, given the concern in gaming around crypto. And you guys have done a much better job with CMP to try to ring-fence that potential risk. But it's kind of silly when you think about how lean channel inventory is today in gaming to think that crypto is this big overhang that could potentially come and be a problem anytime soon.

Yes. So first, when you think about seasonality, we get asked quite often in terms of our quarter-to-quarter seasonality. It's a common question, but you are correct. With the supply constraints, we are really operating on what demand levels do we need to fill? How do we think about the supply behind that? So yes, seasonality has not restored itself and it's not very clear in terms of when it will restore itself in this supply-constrained world. Secondly, when we think about our gaming business, one that has probably been the most impacted from that lack of seasonality, we are still in a case where demand strongly exceeds the supply, and our channel inventories are at a very low level. We work each day trying to improve that situation but there's a lot of different reasons that are causing that. One, RTX is just a phenomenal new capability for gaming. It has really transformed how game-makers are making their games, how creatives are working every single day. So not only are we exceeding the overall gaming market as a whole, we are continuing to expand new use cases beyond just a gamer, the creatives, the broadcasters. Additionally, you've seen the large rise of notebooks. Notebook gaming, high-end notebooks that are still thin and light that you can do your best gaming, you can take your gaming everywhere, has been a great growth driver for us as well. Finding and making sure that supply as well all the way through the OEMs is an important part of our work today. And then lastly, there probably is still some cryptocurrency in our numbers, very difficult for anybody to quantify because we just can't see that piece of it. But with the lean channel inventory and our steps that we have done to create other avenues for the crypto money, we believe we're in a really good position for it to continue to grow.

And then Colette, probably the #1 question I get about NVIDIA, especially as -- and I don't want to jinx things, you're knocking on the door of a $1 trillion valuation, which is just astounding, is what really is the TAM opportunity here? And you and I have had many discussions about this. And I think the latest -- and it's really a data center question, I think, more than a gaming question. I think the latest numbers you guys gave out at the last Analyst Day was $100 billion TAM. Now I look at that number, and I think, okay, you guys are trying to balance something that's big with something that's credible. Not that there isn't a lot of bottoms-up work that goes into that. But we all struggle in this room trying to figure out what that TAM is. We've tried to look at attach rates of GPUs to server shipments. But if you look at fundamentally what NVIDIA is doing is you're fundamentally low in the cost of analytics. And typically, when you look at technology, when 1/2 of the ecosystem figures out ways to do things cheaper, the other half of the ecosystem comes up with brand-new use cases that always grow the market to be significantly larger. You're also augmenting what you're doing in silicon, as you mentioned earlier in your opening remarks, that you're augmenting your silicon with software stacks. So how do we think about the TAM opportunity you have inside of data center over time?

Yes. Really good to break down why the success has been there, why the TAMs are so large. One, we created a capability with acceleration to streamline the entire path in terms of how you process data. And if we think back to the piles and piles of data that was out there, that was just not analyzed, not able to get to. And when you think about a world of prior to the end of Moore's Law, there was never going to be an opportunity to get that done. So that huge transformation and that aha that came forth with a new platform approach really expanded out the TAM. It's not the easiest to look at an attach rate to a server because we're not each server because we're going to change the overall configuration within a data center regarding the number of GPUs versus the number of servers in a whole. So it's not exactly the best way to look at it. But what we did was we analyzed each of the industries, how much they spend, procure on their data centers and getting an understanding of how quickly they would move to this accelerated world. They all will. Our dream, our thoughts, our vision is that they will all move to acceleration. How fast? That depends. Then comes in our software approach. Our software approach is to really help enterprises make that move. Enterprises need the help in stitching it together that system software, that overall application framework to their applications to move to an accelerated work. We've chosen large overall industries, industries that we know would benefit quite quickly from an accelerated and AI model. Thus, has absorbed and created the large TAM that we have. We started with $100 billion, probably the last time we were here on stage, and these white chairs. But we've added quite a few things that are not in there as well. Our choice of adding software separately, adding and licensing software to enterprises separately is not in there. Our favorite topic in terms of Omniverse is also not there in terms of an opportunity. We also haven't included our Grace CPU, our niche CPU that's really focused on high-performance computing and AI attached to that overall vision that we have that it's data center computing as a whole that we'll focus on in every single part of the data center. So the TAM, the opportunity is large. How large will it get? It's very unknown, but there are big markets out there for us.

You brought up software, and that's an area that I get a lot of questions in. I think last time you and I spoke offline, as it gets to be 10% of revenue, I think that's sort of the threshold that you think about breaking it out as a standalone within the financials. One, is that right? Two, how quickly can you get there? And I guess, third question is what exactly are you doing? This VMware announcement on AI enterprise software -- and I'm just a dumb silicon guy. I have a hard time spelling software. Can you help explain to the audience what software suite and tools you're actually providing the enterprise within that JV?

Okay. Our most recent announcements that we've discussed is the amount of software that we have available is quite large. Software is not new to what we've been providing. We've talked about it for many years that our success has been really that system software or some of the key things that we provide free. For example, CUDA, 30 million downloads to 3 million developers that are out there. All free, but it's a development platform for them to also write software. But we assist even beyond overall CUDA and things that we bring. We bring to market AI frameworks, application, different frameworks and the engines that we've talked about. But now there's an opportunity for us to sell the software separately. Now why would we? Because some of our big opportunities in front of us are working with enterprises. Enterprises are very key in terms of the assistance that they need. When I say the assistance, they need support. Our ability for them to have a license that they know they can come to us and seek the support that they need for their application is key. Our work with VMware is very important. Why? Most enterprises use VMware. If we are incorporated in the stack that they use inside of their IT data centers, that they can see and view and model their jobs and their virtualization of their accelerated servers in the same way that they do their CPU servers, that's an initial important step for enterprises well. So it is linking them together for all of the bells and whistles that the enterprises need. Which applications will we offer separately? There are several out there right now. Enterprise AI is an opportunity now in general availability, and we are signing up customers for them to have within their infrastructure as well. We'll talk further in terms of Omniverse as well. That is also something that's available for the creatives for them to do. From time to time, we'll do more, but we have 150 SDKs out there. 65 of them have been revised and/or new for people to look at. But the goal is helping the enterprises move into this world of acceleration.

And it really is a question of which of your businesses are going to grow faster. But is there a benchmark you can provide us with as to when you think software might actually be a reportable segment within the model?

The software is an assistance in terms of helping a bigger piece out there, which is the overall purchasing of the infrastructure. The software has a multiplier to it, which is the infrastructure that is purchased by there. We will provide metrics along the way in terms of what we are signing in terms of software. You will likely be able to see pieces of it as it is still on the balance sheet in those pieces, but we'll talk about it from time to time.

And then if you go back 3 to 4 years ago, I think one of the concerns in the AI market is that you were very well positioned in training, but perhaps not as well positioned in inference. And the view was the overall AI market at its endpoint would be 25% in training, 75% in inference. And there was a fear that maybe your mix would be the exact opposite given your core IP. You've kind of changed that. Can you talk a little bit about the inference side of the market for you? You've talked about that as being a faster-growing portion or the fastest-growing portion of AI for the last several quarters. Where is it as far as the mix in your AI business? Where does it go? And is that a limiter that some people thought it might have been 3 or 4 years ago?

Yes. So inferencing and inferencing with GPUs has been extremely successful. Probably years ago, sitting in this room talking about we are going to go after the world of inferencing that we knew was taken up by the CPU world. But we understood that the complexity of inferencing in front of us was important that they had both the programmability and the flexibility of a platform to take on those larger workloads. So inferencing is now growing faster than our overall data center as a whole. And when we indicate inferencing in this thing, this would be our inferencing processors that we have. We have some processors and platform specific for inferencing. But we also did a different thing with A100. The A100, Ampere 100, was an opportunity to provide both. You have an opportunity not only to do training. You could take a portion of the A100 and divide it for specific inferencing types of jobs. So even only counting the inferencing platform specific, it is still growing faster than our overall data center, not including the overall A100. So our overall approach to not only providing a flexible system and platform, but also all the software that we've enabled so many different inferencing types of frameworks out there, it has been extremely successful.

You mentioned Omniverse earlier. Some people call it metaverse now that Facebook has rebranded themselves. And I want to be careful how I couch this because I would argue that Omniverse has been a background driver of your business forever. If you think about sort of the metaverse as the interface between the physical and the digital world, NVIDIA throughout its history has been trying to improve that interface. You've been doing it in gaming and now you're moving into a broader space. So I'm curious, are we at an inflection point in Omniverse/metaverse? Do we need AR/VR to kind of catch up? And is there -- again, I think this is a very bullish long-term trend for you. I'm trying to figure out if it's been a little bit overhyped here in the near term.

Okay. So Omniverse is a piece of technology and the next wave of technology that we've been talking about for a couple of years. But the great thing about it is it brings together all of the different pieces that we have already from a simulation and do it collectively. Let's talk about that. So first, when you think about RTX. RTX was the ability to simulate light in real time. Additionally, we have done acceleration and huge amounts of simulation for high-performance computing. And then you add the piece of AI and what we have done. Really, what Omniverse is, is simulating physics in a 3D world, 3D collaborative virtual environment. We think that we're in a 3D world. But we're really not yet. If you think about the last 18 months, we've been focused in a 2D world with a little screen in front of us. What if this conference actually wasn't in person? And we had each of the digital twins out there and you and I would be our little digital twin on stage here that they could have chosen them.

I can't wait for the day.

I can't wait for that day as well, okay? So Omniverse really brings together the best of breed of everything that we have developed. Now that the architectures have advanced over these years, Omniverse is not available. It will include VR. It will include AR. It will do everything for collaborative workspaces, it will help in terms of manufacturing and design as well.

Well, I'm curious because everyone -- investors tend to be tactile creatures. If we can see and touch something, it's more meaningful to us. And it seems like where Omniverse is really willing to take off first might be scientific, industrial simulation. What might be the first killer app that we in this room wake up to one day and go, oh my God, we're here, the Omniverse is now present?

Yes. So let's start first, though. But why is it important for manufacturing? Why is it important for architecture? It's an important area because they spend a lot of times collaborating across the world. Their teams are in all different types of regions. And particularly now in a hybrid environment, more and more in terms of working from a different space than the overall office. The world of taking large documents to look at a 2D design is not effective anymore. One, they want to see it real time, not when they finished 3/4 of the way through the design and then say we need to make a change. They want a single source of the truth of the design. Well, wait a minute, what version are you on? Are you on version 17? I'm on version 5. How do they keep that design world working. So Omniverse really helps many of these large manufacturers collectively collaborate on a design before it becomes real, before it is something that they have changed because that change could be rather expensive. Now you and I, in terms of what we will probably see, will be things such as bots, okay, which we'll see call centers, which we will see places, where we will -- customer service be accomplished. Those will be probably 2 things that will be available every single day to us. Additionally, we'll probably have a car that will have our little assistant inside of the car, how to get home, how to get to Judy's house. All those different things will also be available using Omniverse.

I want to go back to gaming and you brought up RTX earlier. And one of the questions I always get is how do I model long-term gaming growth. And I went back and looked at past Analyst Days, and I think the last time you quantified it, and I could be off, so if I am, correct me. It was sort of a mid-teens type CAGR with about half coming from units, half coming from ASPs is kind of how I thought about the business. If you look at where you've been over the last couple of years, 5-year CAGR has been 25%. If you look at the last couple of years, it's been well above that, 40% and 60%. And I think this helps to kind of form in concerns that this is crypto-related and not something else. Can you talk a little bit about kind of how we should think about the long-term growth here? How RTX has really played out units versus ASPs? And I think one of the most interesting data points is just the penetration of RTX and where you are on that curve versus the prior crypto correction with GTX.

Yes. So when we think about gaming, RTX is a very important technology that has been brought to the world. The first architecture that we had was that setting stone for the ecosystem to begin the work in building out the games. Right now, the number of games using RTX, almost all new games, important high-end types of games will be RTX. You've seen the industry as a whole, whether they be working on GPUs, whether they be working on consoles, focused on RTX. That gives us the ability for us to grow not only on desktops, but also with the notebooks and that important piece. We have the ability to upgrade a large installed base. It's not just upgrading the last architecture, but thinking about 2 architectures before that did not have ray tracing at all. It was a very important architecture. It was a very popular architecture in terms of performance. And now when they think about 2 architectures, performance improvement and RTX, it is very popular. We will probably continue to be bigger than the overall gaming industry when we think about our overall growth because we are supporting more than just gamers. We are supporting creators. We are supporting broadcasters, broadcasters in a lot of different universes, whether they're doing AR/VR or working in terms of contractors or enterprises. It's out there and this high-end graphics is an important piece. When we think about what is driving our growth, is it units? Is it ASPs? Yes, it's both. It is both, both that upgrade, bringing on new users, but also people are continuing to turn to higher-end platforms to really make those purchases in the top parts of our stack as well.

Well, one of the things I think it's underappreciated, and you mentioned it a couple of times in your comments, is just the penetration of the notebook market for discrete graphics. Can you help us understand where you think we are on that curve because that's been a great story for NVIDIA over the last several years. And I would argue one of the key drivers as to why the growth rate has probably accelerated versus history.

It has been just an important piece, really working with the OEMs on that design, working together with that ecosystem to help them understand you can put just as great as a GPU into a notebook as you can of a desktop has really transformed. If you think about the world today and the hybrid work environment, the remoteness, that laptop, that workstation, mobile workstation is very important. So the Max-Q technology that enabled that has been just very key and has really assisted both gaming, but also our workstation business.

Do you think we're getting close to full penetration? Or is there more to go in the notebook market?

There's plenty of places to go. I think in the current arena, with the OEMs working with them on the fine-tuning, which ones they can ship right now as we get ready for the overall holidays as a key thing. There's opportunity, additional penetration, not everybody is on a high-end GPU notebook by no means.

Not a business that gets a lot of attention, but it's a good segue with RTX, which is Pro Viz. And I remember when you guys first introduced RTX, you actually brought a bunch of analysts into your headquarters when we were still doing in-person visits. And the pitch was really what RTX could actually do to the Pro Viz business. If you look, that business has been experiencing accelerating growth over the last couple of quarters. And I guess I'm trying to figure out, is this a reopening trade that's helping that business? Or you're actually starting to see RTX to bring down the cost of rendering and creating just new markets and new applications? I think the one that you highlighted when you first introduced RTX was actually could you actually stimulate -- simulate housing, real estate? Instead of driving around and seeing 4 houses in 5 hours,you can see 40 houses in an hour virtually.

Absolutely. So the workstation market has had very, very strong growth over the last several quarters. And you would say, is part of that just due to the environment that we're in? It's really about the future environment that is creating that. The hybrid ability for folks to have workstations both in their home as well into the office is very important. Being able to upgrade to RTX saves them a tremendous amount of time, resources and finishes their work in so many great areas. We've seen this both with the entertainment industry in terms of what they need to do to create overall films. We've helped stitch that together for their overall rendering. You can use an example such as real estate, but there are many examples where creatives working on their own are using RTX to a system and all the creative, whether that be catalogs, pictures, all that type of things are done using RTX. So this is not a trend at the moment but something that will probably take us for many years.

Another end market where your software strategy is going to start paying dividends is autos. With the announcement you made with Mercedes, I guess it's over a year now, for 2024 autonomous driving. That's been a business that's been kind of winding down with your legacy products before it's ramping up with your new core products. How should we think about kind of the growth in that business as we get out into '23 and '24? I think in the past, you've talked about either $5 billion or $8 billion worth of backlog over like a 5-year period.

We have been talking about our autonomous pipeline. And we have analyzed all of the companies that we have signed with in terms of creating autonomous robotaxis, trucking or just a passenger cars such as Daimler. That market size is about $8 billion if we look out to 2027. As we see today, we're still in the early days of this in terms of providing samples, continuing to build out the end-to-end platforms that they will have in the car. So the way that you should think about that is, yes, it may start at the end of '23 move into '24, but hit the $8 billion in total. So it will be quite an important part in terms of how it accelerates in those first couple of years. It's an important part because it doesn't mean just the infrastructure that will be inside of the car, it is also a software opportunity. We have creatively working with Daimler, helping them on that software and will share the software that the company will procure from putting the autonomous vehicles inside those passenger cars.

A couple of modeling questions as we come to the end here. First is just on gross margins. There's a lot of headwinds in the current environment. There's inflation, there's logistics, you're ramping new products. What's the longer-term sort of view that we should have on gross margins, one. And two, as you report the January quarter, one of the helpful things you give us is, as you look out to the next fiscal year, is your OpEx. And as much as you tell guys like us not to look at an OpEx to rev ratio and back into revenue, we inevitably do that. How should we think about kind of the OpEx given the opportunity set that you have in front of you?

Yes. We still get asked quite a bit about gross margin, probably one of those key metrics that people look for at a semiconductor company. We, as we've discussed here, we're different. We're different in terms of thinking about us from a full platform perspective. The best thing that we've been able to do with our gross margins is incorporate software in so many of the platforms that we're doing, meaning the software is not in the gross margin, and that's why our gross margins have been so successful. We look at each of the platforms that we have put out there and the margins have grown, particularly related to data centers and people purchasing more of our high-end cards into the market. As we go forward, probably in the distance if software becomes an important part, whether that be autonomous vehicles, whether it be things that we stand alone, you've got another ability to raise our overall gross margins. When you think about OpEx, you're right that we tend to try and assist thinking about our investment levels. We do not model it versus revenue. We don't. We know the most important part of what we can do as a company is invest correctly, invest appropriately. It's what enables us to build these products, bring them to market. We will be investing. We will provide assistance in terms of what we know we can control. OpEx is some of the things that we can control, and we'll talk about that going forward. We have huge market opportunities in front of us. We have great exciting ideas to do. Now it's our job to think about which ones we can fund, which ones are nearest to the boat, things that we need to fund today. Let things longer term, we also want to. At any point in time, we have things that are providing revenue today, things that are providing revenue a little bit more in the future and certainly work that we're doing long term at all times.

Great. We've come to our end of time in here, but I want to thank Colette, Simona and everyone in the room for joining us. This was great. I really appreciate the time this morning.

Great. Thank you.

Thank you.