from BofA Semiconductor Semicap Equipment team. Absolutely delighted and honored to have Jeff Fisher, the Senior Vice President of Gaming at NVIDIA. I call gaming the original gangster of acceleration a t NVIDIA. Really happy to have Jeff. What we'll do is, we'll go through a few slides that Jeff has prepared, but then we'll get into a fireside Q&A. With that, very warm welcome, Jeff. Up to you.
Thanks, Vivek, and thanks for having me here. Just to repeat, you know, I am talking about gaming. You guys may have thought this was gonna be about data center and all that, but as the original gangster of NVIDIA, thank you for that. I'm here to talk about, and I'm excited to talk about gaming to all of you. I don't know if most of you are gamers, but I expect all of your kids are, and hopefully, they're all GeForce gamers. Maybe you can carry back some of this back to your kids. First of all, I think you're all familiar with our forward-looking statements. You can refer to our filings online to understand and read about the risks and uncertainties of our business, but that PSA is now over.
Let's talk about gaming. PC gaming is strong and is getting stronger. It's not the first time you guys have heard me say that, and it's still true, and I still believe it. If you look at from 2019 to 2022, in spite of the tailwinds of pandemic and work-from-home residing, we have gained 100 million gamers from 2019 to 2022, according to Newzoo. From that new base, we expect to see continued growth as more people come into gaming. Gaming is arguably the most important entertainment medium in the world today. It provides competitive, a competitive environment, provides a social environment, provides immersive storytelling, and ultimately will evolve into a complete interactive entertainment experience. For these reasons, we see a continued growth in folks coming into gaming. This is also evident in Steam concurrent gamers.
They saw a surge of concurrent gamers through the pandemic and work from home, but they continue to see growth, setting new records, are now 33 million concurrent gamers on Steam. If you compare GeForce gaming growth to the consumer PC market, the consumer PC market was largely flat ending 2022 from pre-pandemic in 2019. While GeForce gaming sell-through units increased a whopping 1.4x over that period of time. If you look at GeForce gaming revenue from the same period of time, 2019 through FY 2023 or 2022, we've seen a 20% CAGR annual growth rate over that period of time. This reflects also a clearing of the channel inventory through the pandemic, the post-pandemic phase.
This growth is driven by new gamers and creators coming to the PC, unit lift, as well as ASP lift. We can see gamers valuing more and more the GPU in their purchase and upgrade to PC. Of those gamers and creators that we can see have upgraded their PC, on average, everyone is upgrading to a higher class of GPU. Our ASP mix is becoming richer as gamers choose to buy up to a higher class of GPU. Both ASP and units have driven this growth, and we expect to see a similar trend going forward. If you look at our installed base, there's a huge opportunity to continue to upgrade our installed base. Only 44% of GeForce installed base is on RTX. The balance is yet to upgrade.
If you look at our latest announcement, the 60 class of GPUs, this is starting at our $299 class, the core gamers. Core gamers have a much longer replacement cycle. They have prior generation architectures, Ampere, mostly Turing or earlier, and lower performance GPUs. This class of gamer is a huge opportunity for us to upgrade. Only 18% of our installed base enjoys a 3060 class of performance or higher. The balance all needs to upgrade. I'm super excited about the 4060 launch and what this class of gamer, what this segment provides for us in terms of opportunity. Last, I wanted to share with you a bit about our Ada ramp. As you guys know, we have, we launch our products, our new generations from the top down, starting at the top.
For reasons you're also aware of, Ada has taken us a little longer to move down the stack than typical generations. For this reason, we've got a fair amount of history on the $699 up class of Ada. This is our 4070 Ti and up. Ada, in terms of the revenue generated, is ramping much faster, 3x that of Turing and about 40% faster than Ampere. We're super pleased with the Ada ramp to date. Some time ago, it became evident that Moore's Law was slowing and was not gonna allow raw transistors to continue to drive the gaming market. It was important for us to find an alternate path, and in 2018, we announced RTX.
RTX featured ray tracing cores so developers could create movie-class cinematic games, and Tensor Cores for AI. The first technology we delivered, AI technology we delivered for gaming was DLSS, and DLSS provided a massive performance boost. DLSS 3, our latest innovation, is a generative AI approach to rendering games. DLSS 3 uses generative AI to create seven pixels for every pixel rendered. You guys now, I think most folks know how gen AI works, but it has become a very powerful performance boost and quality, image quality boost to games and gamers. We now have over 400 games and apps that are supporting DLSS in our ecosystem. From the start of our RTX launch, we've continued to innovate both in ray tracing performance, DLSS, and AI features inside of our chip.
The result is a 16 times improvement from the beginning of RTX to our current generation Ada, in operations per pixel. This is effectively a function of image quality and performance. We've delivered 16x image quality and performance, while our transistor count has only gone up by 4x. The result is that RTX has transformed gaming and built an ecosystem around a new level of performance and image quality. Accelerated AI, a game changer for all apps, especially gaming. Microsoft gave a huge boost to AI on PC, working to make the PC a first-class client for AI at the Build developer conference a few weeks back. At Build, Microsoft announced a framework for AI model plugins for all Windows apps.
They released a set of tools for ISVs and developers to build and provide models for all the Windows apps. NVIDIA and Microsoft announced a collaboration to accelerate all the AI models coming to Windows on NVIDIA GPUs. If you consider AI for PC and AI for all apps as the next killer app for PCs, which I believe could easily happen, you will need a range of hardware to accelerate, like gaming and like graphics, to accelerate all these AI models. If you look at our stack from our entry laptop, all the way up to our highest-end desktop GPU, RTX 4090, we provide a huge range. You compare that to what I'll call DLAs, or VPUs, or whatever you call them.
There's a.. These are AI accelerators that will be built into future CPUs and SoCs for accelerating a number of these, what I'll call small models. AI accelerators will be perfect for the type of mainstream AI apps like noise reduction, background blur, keeping your eye gazed at the screen. That will become a broad base for work for developer, for AI developer models. For heavier weight models, things that developers really want to change user experience, like generative AI, all the way up to full character animation, again, like gaming, you're gonna need an entire range of performance, and we have a family of GPUs to deliver that range of performance, just like on gaming. At Computex, was it last week? I forget when I got back from Taiwan. I think it was last week.
We announced our first AI platform for gaming, ACE for Games, and it's really a great example of what the future of interactive entertainment and gaming will be. ACE for Games is one of our NVIDIA AI foundries, where game developers can bring all of their game lore or their game, even their game assets, to NVIDIA to train a character to become knowledgeable specifically on the domain of the game. If you talk to a character inside of a game, you could ask it anything about the history or lore of the game. If you ask it something about the 2024 presidential election or about Vivek, you don't want it to answer. You want it to have no clue what's going on. It'll take you out of context.
This is what our AI, what our AI foundry does, is effectively, you bring your assets, you train them on our models, and then you have a domain-specific, a character inside of a game. ACE for Games is effectively, is three components: Riva, which is our speech processing. It processes the input speech. You're talking to a character. It has NeMo, which is our large language model, gen AI chat model, which takes your input speech, and ultimately, Riva converts it into text, and NeMo will then create a response. Riva then will take that response and convert it into speech back from the character. We have another technology called Audio2Face, which will take that audio stream and convert it into an animated character, that's a component of Omniverse.
Between Riva, NeMo, and Omniverse, these are the three main components that make up ACE for Games. Developers can either take them in total or take any piece of it and provide, put them into their workflow. I want to show you a demo of ACE for Games, and I also want to point out that while this was recorded in real time, and while the chat and conversation is specific to the game, it isn't pre-scripted. Let's take a look at ACE for Games. This was developed with a company called Convai.
We are thrilled to showcase an immersive and dynamic interaction between a player, Kai, and a non-playable character named Jin. This captivating ramen shop scene was created by NVIDIA using the latest RTX rendering technologies. NVIDIA collaborated with Convai, an NVIDIA Inception partner, to showcase how developers will soon be able to use NVIDIA Riva, NeMo, and Audio2Face for AI-powered speech, conversation, and animation. These models were integrated into the Convai services platform and fed into Unreal Engine 5 and MetaHuman to bring Jin to life. Please take a look.
Hey, Jin. How are you?
Unfortunately, not so good.
How come?
I am worried about the crime around here. It's gotten bad lately. My ramen shop got caught in the crossfire.
Can I help?
If you want to do something about this, I have heard rumors that the powerful crime lord, Kumon Aoki, is causing all sorts of chaos in the city. He may be the root of this violence.
I'll talk to him. Where can I find him?
I have heard he hangs out in the underground fight clubs on the city's east side. Try there.
Okay, I'll go.
Be careful, Kai.
A peak into the future. Just to close, gaming is strong, and it's growing. RTX transformed gaming. We've broken loose of Moore's Law. AI will be a game changer for all apps and specifically for gaming. We're really looking forward to what that opportunity brings. Thank you for listening to my opening.
Excellent.
Vivek?
I think we could watch that game all day long and not have to go through Q&A.
Can we?
Maybe just to start, you showed that, you know, that very impressive chart of, you know, how there's a big upgrade opportunity. How large is the deployed base for NVIDIA gamers, and how has that deployed base grown over time?
Yeah, we've got the installed base of NVIDIA over 200 million NVIDIA gamers out there, and GeForce-based PCs. How has it grown over time? I mean, our units have grown at about 10% annualized. I think that's probably a fair estimate.
Got it. One interesting statistic I saw there is that only 18% has upgraded to over a RTX 3060 class. Why use the RTX 3060 as the metric? Like, what's so special about the RTX 3060 to use as the metric, right, that should be used to compare.
Yeah.
The base model?
It 's taken us a while to get the Ada series down to this, down to that 60 class.
Right.
There's a lot of talk about how much, you know, opportunity and what do we think of the 60 series. From my perspective, it's, it is a perfect upgrade for that huge installed base of gamers. Remember, we're bringing 60 series all the way down to $299, which is below the price of our, of our RTX 3060. It, you know, it's to share with the team, I mean, some of my optimism about what's the market we have to go pursue with that device.
Got it. What have you seen so far in terms of the Ada adoption trends? Because I think initially you only launched the high-end SKUs last year, but now I think, as you mentioned, you launched the 60 SKU as well. How has been the adoption of Ada, and who's the typical, yeah, is it somebody who is two generations old, three generations old? Who is the typical gamer, you know, upgrading to the Ada series?
Yeah, that's a good question, and it kind of depends on where in the stack we're talking about. I mean, I shared with you how the RTX 4070 Ti is doing relative to prior gens. In fact, the early results from lower down Ada are also, you know, very strong. I mean, I'm really pleased with the Ada ramp. You talk about, you know, who is upgrading? At the highest end, those gamers, a RTX 4090 gamer wants to play with everything turned on. They wanna play the AAA games.
Right.
They wanna play with everything turned on. From a perspective of ray tracing and DLSS, they will see a 2x improvement gen to gen. I mean, if they're upgrading from Ampere, they are gonna see a really great improvement. If they're upgrading from prior gen, it's kinda, it's a lot more than that. As you go down the stack, especially in the 60 series, those gamers have a much longer replacement cycle.
Right.
They'll go back two generations or longer. Their comparison is relative to a Turing or a lower-end device or something even pre-Turing. They're easily in that class, they're even in basic rasterization gaming, they're gonna see a 2x performance as well.
Got it. What's the typical upgrade frequency? If I look at a conventional PC, you know, people might upgrade it like in the four, five, six year, right? More likely five or six-year time frame.
Yeah.
What's the typical upgrade frequency for a enthusiast gamer?
Yeah, it's I'll tell you my expectation, and that is that at the highest end, those gamers will upgrade every generation. I mean, they like the best, they likely have the wherewithal to afford the high end, each gen, and that's about every two and a half years, right? As you get to the bottom of the stack, I think we're stretching out to probably three and a half to four and a half years for replacement cycle. It's probably we see more upgrades than we do new PCs sold, so it'd be shorter than the overall PC cycle.
I see.
That's about the frequency. Yeah.
Averaging somewhere around three-
Three and a half.
Yeah.
Three and a half years, something like that.
Three, three and a half years.
Yeah. Yeah.
Got it. Do you collect data on when they upgrade? Do they typically, I assume they would upgrade to a higher-end SKU, right, than before?
Yeah. Yeah, that's a good question. I mean, they, we have some of our gamers who are using our client, I think you, we've talked about this support, GeForce NOW, which will help keep your PC upgraded with the latest drivers automatically.
Right
A nd helps you and will automatically set the best settings for every game, which is, these days, super important for your hardware. They also share with us their history, among those users, we can see when they've upgraded a graphics card inside of a PC. We have seen generation to generation, gamers buying up more each time. The lift from what they had to what they're buying continues to increase, which is a great testimony to the value they place on the GPU relative to other components in their PC, and what they're willing to invest in a great entertainment experience. So to answer your question, yes, we do see a increasing investment from gamers who are upgrading within our base of GFE users.
Is there an easy way, Jeff, to look at how much, You know, you mentioned you've seen double-digit kind of unit, growth, right, in your install base? What about the content expansion for NVIDIA? Has that also been 5%, 10%? What, what is the right way to think about what that content growth has been over time? I ask that question because I think there have been two instances where, you know, I hesitate to use that word again, crypto, because it caused a lot of confusion. You know, that created, you know, an abnormal lift on the pricing side? If you kind of normalize for that, what do you think has been the average content, right, or mix-up in your install base?
When you say... Oh, mix-up, that's what you mean.
Yeah
B y content? Okay.
Yeah.
Well, it's a good question. First of all, when we look at, when I look at mix-up, or upsell, remember, even in the, even in the days of, I'll call it product shortages, throughout the pandemic and work from home and crypto, as you mentioned, we didn't really change our GPU pricing. I mean, the way I measured upsell was from my core price, not what it ultimately got inflated to in the end market. You know, what we've seen today for Ada, in this class, $699 and up, is a $300-400 upgrade in mix, in the high end.
Gamers who are buying our high end of Ada are buying from a much lower class of product, which is super exciting for us. Ampere, it was I'll speculate, was in the $200-300 upgrade range, and Turing was probably a bit below that. We've seen an increasing mix of upgrades over time.
Just to get it right, what you're saying is that people who upgraded to Ampere, you know, on an average, bought a card that was $200 more expensive, and now those who are upgrading to Ada are buying something that is $300 or 400 more expensive?
Right, relative to what they had in their PC.
Relative to what they had?
Yeah.
Okay.
Relative to what they had in their PC.
Right.
That's right.
If you were to do a similar, you know, mix-up math over long periods of time, is that also kind of a double-digit average across your base, or is there a simple. We like simplicity, you know, if you give us units and mix-up, it'll help us.
Well, you know, I'm not sure if I'm answering your question exactly, but historically, we've seen about a 10% in lift in units and a 10% lift in ASP over time, and I don't see any reason that necessarily, that trend would change.
I see. The one other interesting thing that, you know, also alluded to is kind of the synergy in NVIDIA between gaming and data center. From your perspective, right, how do you benefit from having, you know, data center in the same company, and then-
It's good-
Vice versa, right? How do they benefit?
Yeah, that's a great question. I thought you were gonna ask me, you know, why are we, you know, well, are we gonna throw all of our money at data center? But it, I like it, the way you phrased it better. The effectively, you know, I've been at NVIDIA a long time, and it's easy to see how this company has grown. I mean, we have a unified architecture across all of our all of our devices. Unlike, I guess, many companies, certainly our scale, we have a single engineer, hardware engineering team and a single architecture team. The core of our products is, if you took it down to the atomic level, I guess there's CUDA cores.
CUDA cores were used to build our gaming franchise, but they also are powering our AI hardware. The investment in CUDA cores will continue to drive roadmaps both on the data center AI side, as well as the gaming side over time. Also, our I think Jensen often describes our company as a giant kitchen, and he likes to be involved in the cooking, as you could imagine. But inside the kitchen are these really incredible technologies that are shared across our businesses. While we've been investing and driving our AI platforms for data center for quite some time, we're now starting to bring them over to gaming.
I mean, you could see ACE for Games is taking Riva, NeMo, Omniverse, all things you've seen in our other businesses, and we're now bringing them in, into formatting them, bringing them into the game development community. It's a very powerful. The company and the way we've organized and our unified architecture is very powerful for all of our businesses. You can see we try not to stray too far outside of them because it's hard to leverage that investment.
Got it. Just one near-term question as you look in the back half of the year. You know, there's a lot of concern about consumer spending, right? You know, the non-rebound in China and so forth. How do you think about the growth of gaming? Do you think you are expecting kind of a typical seasonal lift, or do you think that right, things are gonna be tougher in the second half, given all the consumer macro concerns?
Yeah. You know, we are Q1 came in as expected, and it's seasonal. I expect the year you know, we'll see. I expect the year to play out with a seasonably larger second half. I think the, and that, I'm also assuming some stability relative to where we stand today in the world. I mean, anything can change. China has got some pockets of strength and also macro weaknesses. Europe, I would say, has a similar scenario as does the U.S., but generally, the base of business, the foundations of our business, relative to people want to game, and they wanna play the latest games. I don't know if you caught it, but Diablo IV, huge game, is now out in open beta, soon gonna be released.
We expect that and Starfield and Call of Duty: Next to really be drivers for folks coming out and upgrading their hardware for games. In spite of the, all of the macro, we are optimistic that we'll see a regular seasonality this year.
Got it. Jeff, if you look out over the next, you know, five, 10 years, do you think this remains a hardware business, or does it become a software and subscription business over time? What would be the pros and cons if that were to happen?
Well, I think it's gonna be a hardware business for the foreseeable future. I mean, I think that as games have got a long way to go to deliver what I always refer to as real life in real time. It's gonna take a continued hardware investment, and the most economical way to deliver that is really on the client. Gamer buys a PC, he can have the best experience local. It may continue to move up the stack. As you guys know, the volume of the mainstream has largely disappeared, that entry-class GPU, but the gaming segment remains quite strong.
I don't know if you have a question about the cloud there, but I ultimately, gaming will expand, and we'll have these massive universes. I hesitate to use the word metaverse because it may be a dirty word in the room, but I do think that these large, interactive, shared experiences are gonna become super important and not just important, but the way people are gonna wanna play games over time. While local PCs are gonna be used for rendering and some AI modeling, the cloud will play a more important role over time. That, the cloud part of that will likely be subscription-based.
I think what you'll be either buying or subscribing to is hardware, local or in the cloud.
All right. Maybe one last one. NVIDIA has enjoyed, you know, very strong, 70%, 80% market share, right, among enthusiast gamers. You know, your competitor is also bringing out a number of new products, and they can also buy transistors from the same foundries.
Yeah.
That you can. What gives you the confidence that you can maintain this kind of very high, right, premium market share?
Yeah, well, they buy the same transistors, so that's great. I mean, at the core level, we're at a level playing field, and we can then innovate on top of that. I mean, I wanna point out that the, you know, those of you who read the headlines about ASPs and pricing, we offer a full stack. I mean, we now offer our RTX 4090 is actually priced below the RTX 3090 Ti. I don't know if anybody realizes that, but the RTX 4060 is priced below our RTX 3060. We have a range of prices. If, you know, a competitor comes out with certain pricing, that's fine. A gamer can pick between them or GeForce.
What we've built really is a promise of compatibility and performance to our gamers. I mean, we release a new driver with every major game that comes out. GFE will update your PC, or you can go get it from the web and download it. That's a cadence about once a month. The new drivers aren't, you know, fixing bugs 'cause the stability, compatibility is awesome for GeForce. They are improving the performance of new games that have come out, and they've been fully qualified on new games that come out. Gamers know that. We've also built a broad channel. Our channel knows that if they stock GeForce, that it will sell.
They can confidently take in our new product and make it available globally, worldwide in every country, within about three weeks of us launching a brand-new product. Availability is huge, and we also have a ecosystem. We have a very large, massive team of hardware and software engineers that call on game developers. The purpose of that is to make sure that games and our technology really shines, and game developers know how to take advantage of some brand-new things like AI and ACE. As a result, games coming out, they run best on GeForce. 400+ DLSS games and apps is a real testimony to that.
Excellent. Thank you so much, Jeff.
Okay.
Really appreciate your time.
Thanks, Vivek.