NVIDIA Corporation (NVDA) Earnings Call Transcript & Summary

Good afternoon. My name is Mike, and I will be your conference operator today. At this time, I would like to welcome everyone to NVIDIA's GeForce's Launch Financial Analyst Conference Call. [Operator Instructions] Simona Jankowski, you may begin your conference.

Thank you. Hi, everyone, and thank you for joining us today. We're very excited for this opportunity to discuss with you our big launch today of the GeForce RTX Series 30 GPU family. On the call with us, we have NVIDIA's CFO, Colette Kress; and our SVP of Gaming, Jeff Fisher. The format of today's call will be a short presentation followed by Q&A. You can find a copy of the presentation as well as the replay of Jensen's virtual launch event from earlier today on our Investor Relations website. And after the presentation, we'll open it up for Q&A for either Jeff or Colette. Before we get started, I need to read a quick statement. During this call, we may make forward-looking statements based on current expectations. These are subject to a number of significant risks and uncertainties, and our actual results may differ materially. For a discussion of factors that could affect our future financial results and business, please refer to our most recent forms 10-K and 10-Q, and the reports that we may file on Form 8-K with the Securities and Exchange Commission. All our statements are made as of today, September 1, 2020, based on information currently available to us. Except as required by law, we assume no obligation to update any such statements. Okay. So with that, let me turn it over to Jeff to kick us off.

Thanks, Mona. Welcome, everybody. Thanks for joining us today on the call. This has been a very big day. It's been really exciting, working up to our launch today. It is our first virtual GeForce launch. I kind of miss the live events in-house with gamers, but this was certainly a successful event regardless. And the team did an amazing job pulling it all together. Everyone working at home was a new experience but it was an amazing production. The work was made even easier because we had a great story to tell with Ampere. I hope you all enjoyed it. Let's get started. I want to go through some of the highlights of the presentation today. You may have all seen Jensen's keynote. Hopefully, you did, but I want to go through some of the key points in slides, and then we'll get on to question and answer. In the wake of COVID, it's evident that gaming has become a part of everyday life for billions of people. Gaming connects people. It entertains. It tells stories. It's a sport that allows people to create or even become broadcasters of their own content. GeForce PC gaming is large and thriving. It's open. It's scalable. Games like Call of Duty and Fortnite have attracted hundreds of millions of gamers, and are now offering new experiences to participate like the Travis Scott Concert in Fortnite earlier this summer where almost 30 million people were participating and watching in the game. 75% of GeForce gamers play eSports titles. eSports has an audience of over 500 million people, and is a new competitive genre for most of a brand-new generation of gamers. Gaming has also opened up a new world for creators who can modify games, create their own stories, using game assets to share with their friends, a whole new medium for creation. With PC gamers, anyone can broadcast their gameplay with PC -- with a gaming PC, anyone can broadcast their gameplay like a pro to a wide audience or just to a few friends. Today, there's over 20 million streamers globally. So let's talk about the product. NVIDIA RTX is not just a GPU, it is a full stack. It starts with a new GPU architecture -- our new GPU architecture, which is RTX. On top of RTX, the device, it includes new engine tech for games and for creative apps. And a bunch of new rendering algorithms that are invented by NVIDIA research team. All the major 3D APIs have extended to include RTX. RTX is supported by all major 3D-creation tools. RTX tech is incorporated in all major game engines. There are hundreds of games being developed with RTX tech, and there's been thousands of research papers written about it. We started designing the RTX architecture over 10 years ago to address limitations that we saw coming in traditional rasterization technology. The RTX GPU has 3 dedicated processors: the programmable Shader, which we designed over 10 years ago, a revolution in graphics in and of itself; and RT Core to accelerate ray tracing; and the Tensor Core, which is an AI processing pipeline. The Tensor Core powers our graphics AI technology DLSS, which delivers high-performance and image quality in games, and is now viewed as critical to deliver high-performance ray trace gaming. RT and DLSS are in a majority of the new generation of games. Let me start by showing you how the 3 dedicated cores work concurrently. First, Shader. This is an example of a single ray trace frame of the game Wolfenstein: Youngblood, so it's 1 frame. The Shader can do all the work of ray tracing and rasterizing. Some architectures may attempt to do this. Even Pascal was capable of ray tracing, but the perf is bad. This is 51 milliseconds of just using Shader to ray trace a Wolfenstein frame. 20 millisecond -- 51 milliseconds is less than 20 frames per second. The game would be virtually unplayable. Adding RT cores gets you 20 milliseconds per frame. RT Core run -- concurrent runs concurrently with shaders. The 20 milliseconds is still not 60 frames per second. That is where AI comes in. Adding our tensor cores in DLSS, you are now gaming well above 60 frames per second at 12 milliseconds per frame. It's all about 3 dedicated cores capable of concurrent operation to deliver high-performance, ray-traced gaming. Now we get on to Ampere, our second-generation RTX architecture. Ampere has a new Shader. At 30 teraflops, it is 2.7x Turing's performance. It has our second-gen RT core. At 58 teraflops, it is 1.7x the performance of Turing and ray tracing, and includes our third-generation Tensor Core with 238 tensor teraflops for AI and DLSS. In all, the 3 cores deliver 2x perf. 3 cores, 2x perf. This is our triple double. So today, in Jensen's kitchen, we introduced Ampere, our new GPU. It's 28 billion transistors. 28 billion. It's in Samsung's 8N custom process, connected to the world's fastest memory, G6X, and delivers 2x perf gen-to-gen. Ampere powered on in our labs this last May and is now in volume production. It has been an amazing execution by the team. Proud to be a part of it. Proud to watch in times of COVID, just amazing execution. Full stack engineering and craftsmanship is the heart of all of our GPUs, and Ampere is no exception. The goal is to squeeze every drop of perf and power efficiency up and down the stack. Ampere delivers 1.9x more power efficiency than Turing. This extends our ability to operate in low-power systems and scale higher performance with power. So today, we announced our new flagship GPU, the RTX 3080. At $699, it has 10 gigabytes of the world's fastest memory. It's a generational leap. That's super important when launching a new GPU. This inspires developers to push realism on a much larger footprint and, of course, for gamers to upgrade. The 3080 is available September 17. As you know, our first-gen RTX, Turing, while groundbreaking and setting the stage for a revolution in graphics and gaming, did not quite achieve this gen-to-gen perf over Pascal. 3080, our second-gen RTX and design in Samsung's 8N is designed to deliver amazing perf for current and next-generation games and inspire gamers that now is the time to upgrade. This chart shows the 3080 at $699 stacked above 980, 1080 Ti and 2080 SUPER, and delivers roughly 2x the perf of our 2080 that was launched just 2 years ago. The second GPU we announced is the RTX 3070. The 70s are some of our most important products. The 3070 is priced starting at $499 and will be available in October. It has 8 gigabytes of G6 memory. The 3070 at $499 is faster than our 2080 Ti. Incredible value. Once again, the 3070 is designed to inspire developers and gamers alike, delivering great perf over Turing and a clear choice for gamers to upgrade. Finally, we launched a big ferocious GPU what we affectionately call the BFGPU, our 3090. Realizing that our TITAN class products were in high demand, enthusiast, creators and researchers were using it to power their gaming rigs and workstations, we decided to add a 90 class this cycle, our first ever, at the top of the G4 stack and to sell it through our worldwide AIC channels. It is a beast of a GPU, and its priced starting at $1,499. It is the first GPU that will support 8K 60 frames per second and with ray tracing on. The 3090 will be available from partners at the end of the month, starting September 24. So all in all, the keynote covered a lot of material. NVIDIA delivers the greatest-ever generational leap with the 30 series GPUs. Ampere is more than a chip. It's a stack that powers the top game engines and creator apps. Ampere's our second-generation RTX arch, well ahead of the industry, leading the industry, with 3 dedicated cores that is delivering 2x the perf and power efficiency. 3 cores, 2x, once again, our triple double. We also deliver new tools for the infinite ways that gaming is expanding. As noted, gamers are not just gamers anymore. They're creators, they're competitors, they connect socially, all important elements of gaming. And the GeForce platform continues to deliver. This is for the tens of millions of streamers, broadcasters, eSports pros, artists and creators. We did this with our new applications for eSports, NVIDIA Reflex, delivering low-latency gaming for the millions of eSports gamers; NVIDIA Broadcast for the millions of GeForce broadcasters who are streaming their games; and finally, NVIDIA Omniverse Machinima, allowing gamers and creators to take in-game assets, combine their web camera and other digital effects libraries to create cinematic film quality storytelling from game assets. This will be -- we believe Machinima will truly be groundbreaking. So thank you all for your time today. Hopefully, we covered everything -- oh, there's one more point I want to make, sorry. Finally, RTX is on, and this is probably one of the most important things. When we launched RTX, we wanted to inspire developers. RTX real-time ray tracing is now coming to Fortnite. With both Minecraft and Fortnite onboard, the #1 and 2 most played games in the world are RTX ON. Adding the most anticipated titles this fall, Cyberpunk, Call of Duty: Black Ops and Watch Dogs, there is no question that RT is the new standard for games. Okay, with that, let me say thank you all for your attention, and we can open it up for questions. Simona?

Yes. Thank you very much, Jeff. And operator, let's open up the line for any questions.

[Operator Instructions] Your first question comes from Vivek Arya from BofA Securities.

Jeff and Colette, it's Vivek Arya. Congrats on the launch and thanks for letting me ask a question. I actually had 2, one for Jeff and one for Colette. Jeff, how large do you estimate your deployed base of gamers is? And how do you see this upgrade cycle playing out versus what you saw with Turing and Pascal? And do you think this is more of a unit- or ASP-driven cycle? And if I could just also squeeze in a question for Colette at the same time. Colette, you mentioned working with Samsung for this new product launch, and I'm wondering how do you think about just kind of the cost and benefit of working with 2 different foundries for your Ampere set of products?

Should I start?

Yes, go ahead. Yes. Why don't you start?

All right. Yes, thanks, Vivek. And thanks for your comment about the launch. We were pretty excited about Ampere. Yes, we -- I believe that as with typical -- as with prior transitions, and as we've seen, 3 of our business, that we will see both unit and ASP benefits in Ampere. Gamers continue to want to buy up the stack. And I think we've delivered -- we're delivering a pretty compelling upgrade opportunity for gamers. In terms of the installed base, there's roughly 200 million GeForce gamers out there. The -- if I look around installed base of gaming PCs, about 30%, maybe a little less, have actually moved to Turing, so there is still a large upgrade opportunity for Ampere. We're also seeing the user base grow. I don't -- COVID certainly has created more energy around gaming, but it's been the trajectory of gaming over time that more and more gamers are coming to PC gaming. And we don't see that stopping anytime soon.

Yes. So Vivek, thanks for the question regarding Samsung. As we've discussed now for several years that we've been working with a dual-fab approach. And working with a dual-fab approach really talks about creating teams on both sides to work with the overall fab. It's not necessarily about at the time you're ready to launch, to be able to move to production. So these teams, to understand each end processes and all the great things that both of these fabs provide, we're honored to now be working with both of the high end and really strong partnerships that are out there with [ TS ] and Samsung. Our decision on this side to take the gaming overall piece to Samsung is really about their higher performance that they could provide on their 8-nanometer, which has been customized now here to NVIDIA. So we'll continue to work on a dual-foundry strategy. We actually think is probably just best practice now out in the market. And keep in mind, Samsung has also been a part of many of our gaming products in the past, so this is just a continuation on building upon the teams that have been created on both sides.

Your next question comes from Aaron Rakers from Wells Fargo.

Congratulations on the launch. Very impressive. Just going back to kind of the earlier question. I know you just mentioned that approximately 30% or less of the GeForce installed base is on Turning. But pretty clearly, the commentary today was that there's a large installed base of Pascal users. And obviously, strong performance metrics around this. So can you just help us understand of how much -- the age of the installed base, how much is sitting on Pascal or even prior generations of that? And then as a quick second question. Just curious as you kind of changed the strategy at the high end with the 3090 product versus the prior TITAN, how big of a piece of mix has TITAN been in the past? And how much could that possibly change now that you more broadly have the 3090 in the market?

Yes. So if you look at our installed base, I mean, on Steam, for example, I think 20 -- high 20% of Steam gamers are on Turing. So all the rest would be on prior generations Pascal or earlier. I see all that as upgrade opportunity for Ampere. The -- let's see, second question, I think was about 3090. So TITAN, it's -- TITAN was constrained, was sold on nvidia.com and through a couple of system integrators. It's hard to really know what the potential is of TITAN. But if I combine the TITAN and 2080 Ti class of product where 3090 kind of sits in the middle of those 2, it's, I guess, a meaningful part of our business collectively. And we think that, that -- both the combination of 2080 Ti, TITAN buyers and as well as inspiring some new buyers will come into that space.

Your next question comes from Toshiya Hari from Goldman Sachs.

This is Charles on for Toshiya. Congrats. I had one around gross margins. I think historically, you guys have benefited from, obviously, gen-to-gen ASP increases, SKU mix across the different SKUs and then the no transitions, the wafer costs. With ASPs largely flat on the 70 and 80 generation, how should we think about the ceiling in terms of gross margin going forward? Is it more about mix across the SKUs or better wafer pricing? Just wondering if we can get some color on that. And then maybe part of that, with ASP flat, how should we think about the ceiling and overall ASP growth going forward?

Sure. So let me start with an overall discussion about gross margin. We'll see if Jeff wants to add a bit in terms of what we expect in terms of the ASPs. So you are correct that generation to generation, we have continued to improve in terms of gross margin of the company. A lot of that growth, whether that be with overall gaming or some of our other business lines, has really been about the higher-end platforms and the amount of software that is incorporated in them. That really allows us to continue to reach the high end of the markets with the overall customers. In this case, when you think about our Ampere gaming, our Ampere gaming gross margins coming out are quite solid, quite good. And we baked this into our guidance that we gave for Q3. So it's still early to determine what we'll see over the life cycle of providing Ampere to gaming. But we probably expect some of those same things that we have seen with each generation to come through. Number one, you'll see an upsell. You'll see a continued excitement to try and have the best of the best. And we've done this over time where we get an average ASP overall increase as we move forward. Additionally, over the time of the life of our overall architecture, we talk about the increases and improvements that we see in the gross margins. That's everything from the overall value engineering of the component. But just over time, we tend to see an overall increase in those gross margins. So with that, I'll turn it over to Jeff to see if he wants to add more.

Yes. I think -- I believe the value prop of Ampere will drive upsell. I would expect just to see some ASP lift as gamers are upgrading to higher-class GPUs to get the best perf on future games. It's what we've seen over time with prior gens, and I don't -- wouldn't expect that to stop with Ampere. You -- when we move to Turing, we had a brand-new architecture in a similar node. And we struggled, as I noted and has been noted, we struggled to deliver great gen-to-gen perf. Groundbreaking architecture, but gen-to-gen perf on traditional games was not that compelling. Yet Turing delivered -- was a great -- was a successful product and delivered a great deal of growth. Ampere is -- benefits from a new node, our second-gen architecture. I think it's going to be a great upgrade cycle.

Your next question comes from Stacy Rasgon from Bernstein Research.

I have 2. First, I think, similar to what we saw with Turing, with Ampere, you seem to be starting in more the high end of the stack. When do you think we might see more of the mainstream offering come in? I guess the equivalent in the prior gen would have been like the 1600 series. And secondly, given these purchases is supposed to be available in a few weeks, kind of end of September, how do you foresee the supply situation? Do you believe that you will have enough supply in the market to satisfy the demand that you're forecasting?

Thanks, Stacy. It's always good to hear from you. Let's see. In terms of timing of the launch and the rollout of products, I think you'll see -- with Turing, it came out -- we rolled out products slower than we probably anticipated or could have, and that was a unique time for our business. As you know, we take a great deal at trying to take care of our channel, and we wanted to make sure that we had sold through prior product before we rolled out some new generation products at certain price points. With Ampere, the channel's healthy and the business is strong. And we are prepared for a pretty healthy cadence of rolling out new devices. I think you'll see Ampere's cadence of new parts to be faster than what you saw on Turing.

I'm sorry, supply of new ones?