Chit Chat Stocks - Nvidia (NVDA) with Luke Hallard

Episode Date: July 21, 2022

Nvidia provides graphics, compute, and networking solutions. The company provides graphics for gaming and is pioneering data solutions for world problems. Listen as Brett and Ryan ask Luke questions about the company, its business model, and valuation. Enjoy the show! ***************************** This episode is sponsored by Stream by AlphaSense, the highest quality expert network library. Sign-up here and get a 14-day free trial:  https://streamrg.co/CCM ****************************** Access our “Not So Deep Dive” episodes by signing up for CCM+. Sign-up directly through Spotify or Apple Podcasts. If you listen on another podcast player, use this link and create a private RSS feed: https://anchor.fm/chitchatmoney/subscribe Need more information? Check-out our launch newsletter: Here ****************************** Want updates on future shows and projects? Follow us on Twitter: https://twitter.com/chitchatmoney Interested to see more of Luke's work? Find him on Twitter here: https://twitter.com/7LukeHallard?s=20&t=7mp9swgMRyapVhG_99ypYg Contact us: [e-mail] Timestamps Nvidia | (4:53) Problems | (14:35) Disclosure: Chit Chat Money hosts and guests are not financial advisors, and nothing they say on this show is formal advice or a recommendation. Brett Schafer and Ryan Henderson are general partners and portfolio managers at Arch Capital. Arch Capital and its partners may hold securities discussed on this show. Learn more about your ad choices. Visit megaphone.fm/adchoices

This episode is brought to you by Stream by AlphaSense. Stream is an expert interview transcript library with more than 10,000 interviews spanning across all industries, including tech, media, consumer goods, and plenty more. Not to mention 70% of these experts can be found only exclusively on stream. Thanks to many of the interviews that I've read on stream, I feel like I've gained a much more intimate understanding of the companies that I cover. And at this point, it has become an integral piece of my research process. So if you want to check out some of their transcripts for yourself, you can go to streamrg.co

slash CCM and sign up for a free 14-day trial using the promo code CCM. Again, that's streamrg.co slash CCM, S-T-R-E-A-M-R-G dot C-O slash CCM. Welcome to Chit Chat Money. This is our Thursday deep dive episode where we interview an analyst on a single stock. Today, we're talking about NVIDIA with Luke Hallard, lead advisor from our friends at 7investing. What were your highlights from the interview? Yeah, and we should mention that they are releasing their top 20 buys portfolio, and

this is one of their teaser reports with that, and for good reason. NVIDIA is one of the largest chip makers in the world. They have a lock, or not a lock, they're the leader in GPUs, which are graphics processing units, or I actually don't know what the G stands for. I think it's graphic processing unit, but they're key to a few big things. The gaming market, the data center market, the AI industry, potentially the autonomous driving industry, and many other things down the road. And Luke talks about them all, including some of their software stuff, CPUs, all that good

stuff. Complicated company, but a fun one to talk about. All right. Well, we don't need to go any longer. Without further ado, here's the interview. Welcome to Chit Chat Money. On this show, hosts Ryan Henderson and Brett Schaefer interview industry experts and riff on the world of investing. As a quick reminder, Chit Chat Money is a CCM Media Group podcast. Ryan and Brett are also general partners at Arch Capital,

and Arch Capital may have positions in the securities discussed in this podcast. Anything discussed on Chit Chat Money by Ryan or Brett or any other podcast guests is not formal advice or recommendation. Now, please enjoy this episode. welcome in today we are joined by luke haller first time guest uh on the podcast we've chatted with him a little bit before over uh twitter messages but he's a lead advisor at seven investing um and there's a recent announcement from seven investing called the strong buy portfolio i'm gonna let you talk about that but first of all welcome to the show thanks guys a pleasure to be here i'm a big fan of chit chat money actually so it's uh it's a bit

of a thrill to be on and joining you both nice there we go there we go well to have everyone uh joining the show from time to time how's the uh so so can you go through what the strong buy portfolio is yeah i'd love to so i guess i joined seven investing as a lead advisor at the start of this year in january um but the firm's been going for about two and a bit years now and one of the criticisms we get from our subscribers pretty consistently is we've got like a ton of recommendations out there it's like 150 different recommendations and they felt our scorecard was getting a little bloated and so subscribers wanted a way to identify and kind of hone in on what our favorite ideas were right now as a lead advisor team so a couple of months ago

we started this and we started assigning conviction ratings to our recommendations and company updates kind of conviction ratings like strong buy the stuff we really really believe in down to kind of potential sell it's kind of on the chopping block to help our subscribers understand kind of our thoughts on the company from month to month and then in june we took all of our strong buy stocks our highest conviction picks and we methodically forced rank them as a lead advisor team so kind of seven investing so on seven seven we uh we launched our strong buy portfolio so that's our 20 highest conviction ideas from all of those strong buy recs and really you know it's a great place for new seven investing subscribers to start if they're building their

own portfolio but also for seasoned investors who are maybe thinking about concentrating the portfolio a little bit um and i suppose the reason you guys have invited me here today is to chat about one of the companies I'm really close to and a real fan of, NVIDIA. So NVIDIA, it's actually a recent addition to my own personal portfolio, but it's one of the stocks that I really pushed to get into the StrongBuy portfolio. I really feel it deserves a place in Seven Investing's top 20. How did you even, I guess, when did you first come across NVIDIA? Maybe you'd heard about it before but as an investment yeah so i guess i you're right like i'm a bit of a gamer so i'm kind of familiar with nvidia i thought i knew what they did and um and it's only when i really

started looking at a bit more detail when the valuation started to get a little bit more sensible in the last five or six months that actually i realized i had a really wild misconception about the company so why don't we go into that a little bit you know kind of who are they and what do we do and they'll we'll clear out that misconception perhaps for anyone who's listening so do you want to go into i guess briefly what they do yeah so um i suppose you might think of and i i did think of nvidia as being like this firm that makes graphics cards for computer games and like you're not wrong historically that's been their biggest revenue segment it's where the company started and they are number one in pc gaming they've got like three times the revenue

of their nearest similar competitor AMD but gaming is really not where the company's future lies so really my whole thesis in investing in NVIDIA is really around the NVIDIA powered next generation data centers which are becoming you might call the AI factories of the future accelerating machine learning accelerating deep learning and training and refining AI models This is really, really quite exciting. So if we really say, what do NVIDIA really do? They're not about gaming. They're about pioneering accelerated computing to help solve the world's most challenging computational problems. Right. And that is a great elevator pitch. I know for anyone that's not well-versed in the semiconductor industry, you're probably confused listening to this. We're going to get into all the specific details later. But first, I want to talk about and just give an overview on the

unit economics because they're a design-only chipmaker. And I think this is one of the most important parts for understanding the basics of NVIDIA's business. So what are NVIDIA's major costs? And what are their general margins gross operating cash flow? Yeah, okay, let's get into it. But maybe just one 30 quick second sidebar, because I think we should give a little bit more frame. So let's just talk briefly about where the revenue comes from. And that'll give us a bit of context when we talk about kind of where they spend their money. So really, they get their revenue in four segments. And we'll go into much more detail later in the discussion, I'm sure. but about half kind of 44 percent of their revenue comes from that gaming division

essentially selling graphics cards but you do some interesting other stuff with those as well we'll get into 45 percent of their revenue now so it's just overtaken gaming is that data center segment and so these are these ai factories but then there are other two other interesting little segments and i'm sure we'll talk about those too one is something called professional visualization So NVIDIA have got something they call the Omniverse. It's basically kind of a bunch of tools and an environment where other companies can build digital twins and kind of train their own robots, kind of simulate their factories. And NVIDIA, quite excitingly, have got quite an early step into automotive. So actually, they're embedded with almost basically every autonomous car manufacturer that's not kind of Tesla or Waymo. Almost everybody else is working with NVIDIA to build their autonomous platform.

So we've got a sense there of kind of what they do, relatively diverse. I mean, they're kind of embedded in every industry. So now I think we can probably dive in a little bit to kind of where they spend their money. So there's unit economics. So maybe let's start with kind of big picture. How much money do they make? So in the last full year of reporting, they delivered revenues of $27 billion, which is up 61% year over year. And let's just think about that in context of kind of how big the company is.

I think today they're like the world's 12th biggest company, market cap of about $400 billion. So to be up at that kind of strata of size and delivering 61% year-over-year revenue growth, that's kind of special. But where do they spend their money? So cost of revenue is about $9.4 billion in that last full year. And that was up 54% year-over-year. So maybe the first kind of reflection is we're starting to see cost efficiencies in the business model. Now the company, you know, it's a 20-something year old company. They're pretty mature now.

They've really got a nice way of operating. And so they're growing revenues at 61% year over year, but cost of revenue is only going up 54% year over year. So we're starting to see like the first glimmers of those cost efficiencies. And I think we should understand, you hinted at it just now, we should understand kind of why they can be efficient. And it really is, it's their manufacturing model. So you look at the sector, semiconductors in the NASDAQ, where NVIDIA sit, and typically in that sector, you've got companies like Taiwan Semi and Intel, who are actually, you know, they've got factories, kind of fabrication plants, and they're making the chips.

well there's also two other companies in that sector nvidia and one of their key competitors amd and they off they operate what's called a fabless business model so they basically do the design of their chips and then they outsource the fabrication to those other guys who do the manufacturing so that's quite a nice model right because they don't have this huge capex to kind of set up the factories so they have to kind of forecast demand because they've got a plan like 12 months ahead to get orders in and have kind of their um their goods ready to sell into the market um but they can be very flexible and they can change the kind of product mix from year to year without having to like massively retool because that cost is all borne by the manufacturers

um so let's sort of drill through the numbers a little bit then so kind of because of that model what it means is nvidia have got a really high gross margin compared to you know almost every other company in the same segment so their most recent full year gap gross margin was 65.5 percent it's actually quite a big number i'm comparing that to those two other competitors amd and intel intel 54 it's quite impressive actually because they do their own manufacturing amd 49 so kind of a you know a much more efficient model than amd for manufacturing um and where do they spend that money i mean essentially it's kind of you know ordering the inventory um manufacturing support you know they've got some of their own kind of labor and

overheads and there's also kind of costs associated with managing an inventory of um of cards actually um i know there's a kind of supply chain concerns right now around this space but in inventory provisions aren't killing them i think we could see some problems on the horizon we'll come to that a bit later in the discussion but right now inventory provisions are around 1.3 percent of revenues so relatively modest cost okay and maybe if we can we uh what is inventory provisions uh Can you give a definition there? So they don't break it out in their 10K. I think these are the costs associated with kind of building a pipeline

and maybe they've got like chips sat in their own distribution chain ready to go out. But this is the kind of the cost of goods sold to kind of manage that inventory. Gotcha. All right. Continue. Sorry, go down the operating or the income statement.

Yeah, absolutely. Yeah. So I guess, as you said, operating expenses next. So total operating expenses for the last fiscal year was $7.4 billion, up 27% year over year. So again, we're still seeing, you know, increasingly the efficiency as they scale up. And then because they don't have, you know, all of these factories and they don't have all of the kind of hardcore manufacturing themselves. So for them, their Oprex is primarily R&D costs, which is about 20% of revenue, and then SG&A costs, which tend to be kind of 8% to 10% of revenues. And when I looked at AMD and Intel, that's pretty typical.

But really, because of that efficiency in the model, they're able to achieve a much higher operating margin than their competitors. So NVIDIA, in the most recent earnings report, operating margin of just over 38% compared to AMD's 20% and Intel's 25%. So we're really seeing quite a significant delta there. Actually, when I dug into the company, I was almost reminded of Apple in some ways. So if we think about NVIDIA's consumer facing segment of graphics cards, it's kind of the same model as Apple with their iPhones, right? They design something, somebody else manufactures it, and then it gets into consumers' hands. Well, you know, we think of Apple as being this highly efficient machine. But actually, you know, comparing Apple's operating margin to NVIDIA, NVIDIA 38%, Apple just under 31%. So, you know, they've really got this model ironed out quite well, I think. If you're listening to this ad right now, we know you're already a listener to our show. But for our avid listeners, we've also started a paid membership service called Chitchat Money Plus that extends beyond just our podcast. Every Tuesday, subscribers get access to one not-so-deep-dive research episode that covers everything you need to know about a company.

You also get an email newsletter with our written show notes, important charts, a transcript of each show, and access to our Chitchat Money research files. Chitchat Money Plus costs $5 a month. You can subscribe directly through Spotify or Apple Podcasts, or if you listen on another platform, click the link in the show notes to go through the simple steps of signing up. If you're a regular listener to the show, we think the membership will provide tons of additional value. On top of the stock research episodes, members will get one Arch Capital Fund episode a month

where we outline why we bought, sold, or continue to hold a stock in the Arch Capital Investment Fund, along with shows on our broader investment strategy. Sign up and become a Chit Chat Money Plus subscriber today. We can't wait for you to join our community. I think we're going to see some problems in the immediate term. So one thing NVIDIA tried to do a couple of years ago, I think 2020, they tried to acquire Arm, who are a British company and really a leader in building CPUs, which is kind of a different type of processor to a GPU, which is what NVIDIA specialize in.

And unfortunately, that acquisition was prevented by regulators, seen as kind of anti-competitive. And so they've kind of thrown in the towel on that. They've announced that formally fairly recently, and they're taking a $1.35 billion kind of write down as a result of that failed acquisition. So we're going to see that in the next quarter's earnings, which I think get reported in August. So I think we expect to see like a short term headwind on some of those margins. but um but that's kind of a one-time cost okay and then you alluded to two of the big segments which are gaming and data centers what has what's allowed them to i guess win customers in those

markets and then you mentioned the graphics cards and this is kind of a basic question could you describe what those what a graphics card is and uh why it's valuable to gamers yeah absolutely so um so i guess nvidia invented gpus graphics processing units and they launched the first consumer card in 1999 so i've mentioned kind of cpus and gpus let's kind of explain the difference so um so a cpu is kind of the engine room of your computer and cpus are typically highly optimized to do an incredible rate but kind of do kind of processing in series like do step a step b step d but you know billions and billions of these things very very quickly um gpus are a bit different gpus have what's called that they have lots and lots of cores kind of

mini processors so instead of doing a then b then c they can do a b and c all at the same time so individually each step is a little bit slower than the equivalent step on a cpu which is highly optimized to run that stuff in sequence but it's doing like tens of thousands of these things in parallel um on the latest architectures you know far more than that so what that means is if you've got certain types of processing gpus are like way way more efficient and so when we talk about gaming um the key thing is you know you i don't know if you guys play computer games or maybe you've watched you know movies and special effects um stuff is becoming so much more lifelike and realistic and almost now on latest versions of like unreal engine with the latest gpus almost

indiscernible from like a real scene it looks like reality and the reason for that is is when you're kind of rendering when you're sort of building the picture to go on the screen um that's highly parallelizable you can do like each pixel you can kind of do all at the same time so gpu is perfect for that kind of processing but it's also perfect for lots of other types of processing which is why excuse me why the ceo jensen huang describes the company as a kind of universal accelerator because it can take certain kinds of processing and just do it way quicker than a cpu i mean that makes sense i appreciate the explainer and how have they been able to succeed and win over customers and who are their customers i guess yeah um so if we stick let's go

through each of the four segments but we'll start with gaming so their customers are gamers um also cryptocurrency miners i guess we'll come on to that a little bit later in the discussion so if you want to run a computer game on your pc um then if you're initially running something that's really quite low power maybe you know five year old ten year old game you're going to want what's called a dedicated gpu you literally buy this chunk of hardware and you stick it in the back of your pc and that's that's the graphics processing unit that does all the hard the heavy lifting to give you this kind of fantastic visual experience um so really there are two manufacturers in this space which are which are the sort of amd and nvidia they've actually kind of jostled and

they've changed places over the years they've both got a long history um um but currently amd are prevailing and there's a there's a good reason for that you know amd have got something like 80 of the market of kind of these dedicated gpus and the real reason for that is really it's that 20 that they spend on r and d like they've led with innovation and i'll probably pick on two things i won't go super deep into what they are but let's just talk at a high level and give you a good example of why they're kind of winning here so you know not only were nvidia the first guys to launch a gpu back in the 90s but the biggest kind of revolution in this whole space was in 2018 so nvidia released hardware on a kind of two-year cycle and we're coming up this year to release

what they call the 40 series processors but back in 2018 it was the 20 series and so on the 20 series processors what they enable with two uh really major innovations and that that got all the gamers kind of wanting to use their cards because um they were just just delivering better results and those two innovations one was something called real-time ray tracing so that's basically um the graphics card is effectively kind of simulating the real physics of light so you get like perfect reflections and refractions and shadows and stuff just looks real and prior to real-time race racing if you wanted to do this stuff in a game like in real time you'd have to fake it um with something called rasterization it's it took amd a long time to catch up with that

really they didn't have an equivalent solution available till about 2020 so um um so you know they got there much quicker um and then the other thing that nvidia pioneered on and actually amd don't really have an answer to even today um back in 2018 they launched something called deep learning super sampling so kind of a bit of a mouthful what does it mean um if you maybe you got your tv in the lounge it might be like a 4k tv you know by laptops a 4k laptop well that's a lot of pixels you have to kind of work out what they look like to render the picture for the screen so you kind of have one frame of the game well in deep learning super sampling what's happening is the game itself can render at say 1080 like a lower resolution and then the ai can

look at what's happened in the last couple of frames and because it's like it's been an ai it's a machine learning algorithm that's trained to kind of render the next frame it can kind of fill in the gaps and so it can kind of guess what they all the other pixels should look like so it moves a lot of the workload away from generating the original picture because it kind of gets upscaled using ai and what amd do today is they have a kind of similar thing that they call um fsr but again they're kind of faking it they're not really using ai they're not looking at the last couple of frames they're basically just kind of upscaling it and trying to like tighten the edges and make things look a bit sharper so today i think serious gamers who are really you know into trying to get like

you know the best quality experience are really sticking with nvidia okay and two follow-ups on that does this software hardware combination is that kind of where you think the long-term competitive advantage with nvidia lies and two does this uh what is it deep super sampling i forget the acronym uh does that make them more attractive for say a playstation xbox or nintendo hardware makers or even a who i guess their steam any anyone making hardware for games does that make nvidia more attractive than say an amd uh it is i'm i don't i don't recall now they're in one of the consoles might be nintendo or one of the others but um but yeah i mean they're they certainly got the best hardware i suppose amd have tried to compete more in the gaming segment

on price delivering they're kind of seen as a little bit like the budget option so um so that that's really the kind of differentiator but um you know i i don't think though that gaming i mean it's that it has been their biggest segment as a company for a long time i really don't think that's the future of the company. I think this is now starting to be eclipsed by the other biggest revenue segment, which is data center. And that's really, data center is really the foundation of my investment thesis. Right. And who are the customers in the data center? I'm assuming it's like an Amazon and a Microsoft, stuff like that? Yeah. Well, almost all of the cloud providers buy NVIDIA kit. Some of them are now making their own silicon, but NVIDIA are also launching like

their own data centers their own data center on a chip architecture and one of the things i looked at when i was researching the company was um there's a there's a list of the world's top 500 supercomputers and um a number of those are accelerated using gpus so technology like the technology nvidia produces well 168 of them are of those top 500 are powered by gpus out of those 168 157 of those are nvidia um and um probably if we pick on two sort of notable installations that are coming up uh meta facebook are launching the research supercomputer based on nvidia's current um architecture and that's so they can build the metaverse that's going to be a beast of an installation but nvidia themselves are anticipated to launch something called eos which

is expected to be the world's fastest ai supercomputer when it comes online expected so towards the end of this year so you know not only are they selling kit to all of the clouds you know anyone who wants to build their own data center but they're also building their own hardware at the same time so then when they build their supercomputer eos uh i know they have the strange names for everything but yeah is it say like a a research lab or uh an individual company can license that time from them and use the super fast computer to do research all that good stuff yeah exactly you can run your workloads on their supercomputer installation that'll be things like the omniverse maybe um but also actually game streaming you know we talked there about that

gaming segment maybe you've got a bit of a weedy pc or you're running on a laptop you haven't got the hardware well actually you can still stream a game and um there's kind of an nvidia uh game streaming option so i imagine they'll be using things like eos to run those streaming workloads so kind of you know my game is being rendered in the cloud rather than locally on my laptop right and i guess we want to talk we want to get past gaming but i have one more follow-up on the gaming segment and it is the transition to streaming gaming or i guess for anyone that maybe wants an analogy here the netflixification of gaming we know microsoft is going into that nvidia is trying as well does is that a growth driver for them or i guess does that revenue end

up in the data center part of the segment but it's also powering a lot of gaming companies does that make sense at all yeah it does i think that falls under the gaming segment okay okay but ultimately i think you know many of these different verticals may be leveraging the same hardware and the same installations behind the scenes i know one of the gaming streaming probably the most prominent gaming streaming company or who's been the most clear about their intentions i think is microsoft with their xbox streaming solution i know they just announced that there's going to be a console-less uh experience it's kind of in beta mode right now they're rolling yeah is that would that be a customer or would that be more of a competitor in that sense uh so i don't think

microsoft make their own silicon um and certainly they are one of nvidia's customers at a kind of data center level so you know a lot of azure microsoft's cloud is built on nvidia hardware so i can't say with certainty to be honest but um you know they've certainly got a chunk of nvidia um processing that they could be streaming that xbox experience from okay and we hit the big things you know gaming and data centers that's a big chunk of revenue right now but there are some other stuff we've been talking about uh we're messaging on before the show and i want to hit on these three things we already talked about the supercomputers but there's omniverse automotive which is very small but again we all know the the long-term potential of autonomous driving

and then the uh cpu i forget all the names of this stuff they have so many different products but the cpu stuff as well so first let's hit omniverse what's what is that like i don't know i i tried to look what it is i have no idea what it is what is it who are they trying to displace what market are they trying to go after yeah so uh so yeah just remember those numbers right um data center and uh gaming are the two you know they're the really material drivers of revenue right now but visualization is kind of the next chunk it's much smaller in comparison like um so what is omniverse um it's kind of a virtual world i guess everyone's talking about the metaverse right so um the metaverse could be lots of different things you could think of the

omniverse as kind of being a metaverse for industry so maybe let's give a real example um so amazon have partnered with nvidia to use the nvidia omniverse to optimize their warehouse design and train their robots so actually what they've built is what's called a digital twin of some of their warehouses and then probably sort of two things that they would use the omniverse for um one would be say they're thinking oh how do we bring in like a new product line or optimize the warehouse and the flow of people and robots well they could redesign it in the omniverse and then kind of you know hit go and run it and watch these like virtual robots run around and watch the kind of virtual orders coming in and getting packaged and going out and then they

can iterate virtually and they don't have to spend a fortune kind of building something that's suboptimal um but they could also use it to train the robots themselves so what the omniverse really is it's like a bunch of tools with you know fantastic sort of visualization but all the real world physics so you know if you build a kind of uh the ai you're trying to train your robot to live in and operate in the real world maybe it's like a little amazon like little floor drone thing running around picking up and dropping down shelves where you can train that in the omniverse and the tool set has just done a lot of the heavy lifting for you um so it makes it much easier for companies to build kind of industrial applications that's kind of what it's all about

i don't know if that sort of cut through the complexity is it a software subscription or are they doing like i guess maybe amazon's probably an enterprise partnership but are they is it offered as a software subscription uh yes so you can so anyone can download and install the omniverse and actually i don't know the commercial model so i imagine there's yeah there's some kind of subscription element and it will be based on like processing volume something like that when was it released like i know it's really new so they're still kind of building this out yeah they are it's uh it's relatively recent but it's uh it's been appearing in their numbers for at least the last uh eight quarters i think okay gotcha only eight quarters all right let's move

automotive unless you have anything else on the omniverse um that is a it's i i don't have the numbers in front of me but when i look i say you look at the numbers it's way way smaller than data centers right now but we know the potential of autonomous vehicles what exactly are they doing in automotive because i know they're not making the cars is it software and hardware what's their goal here yeah exactly it's both it's software and hardware so um you know if you're uh if you're tesla and you vertically integrated the whole thing and you're building the cars and you're actually you're building your own silicon and you're building like the uh the machine learning like the ai that's going to do the driving um or you can you can kind of afford to do all that

because you've got that kind of culture of innovation you got that insight in-house if you're like a land rover or a mercedes it's a bit of a stretch to go from being like a car manufacturer to being an ai expert and building your autonomous driving platform so what almost literally almost every other manufacturer apart from tesla and google's waymo have done is engaged with nvidia um to partner with them and nvidia are actually building the autonomous driving platform so the hardware the sensors to plug into it but then the most complex bit the software and then that will talk to the car and tell it you know speed up slow down turn left turn right and if we just think about some stats here so evidently uh in the most recent earnings release

they shared some stats they said nvidia is currently in 20 of the top 30 passenger electric vehicles 7 of the top 10 autonomous trucks 8 of 10 robo taxis and all 30 data centers that are doing autonomous driving training because i guess they you know they're including tesla there right now because perhaps they haven't got their own silicon across the line just yet so as you as you said it's like a tiny fraction of revenue it's like 1.7 percent of revenue right now but and this has got to be a super long-term bet right at least i've got a tesla on the driveway like it's not very smart it's the it's the pinnacle of ai on the road but it's pretty dumb when you're not on a highway on like a big road um i think realistically i got my spacex t-shirt on right

now i'm a musk fan but realistically we're probably a long way from autonomous vehicles being a reliable reality where you could take the steering wheel away um so you know i don't think this is going to be material to nvidia for quite a long time perhaps not before the 2030s but i mean if it gets there and it could come to nothing well this could be like a trillion dollar industry in itself if you imagine that's like nvidia not just uh providing the kind of hardware but i suppose the recurring revenue of managing every single car on the road that's not a tesla or a waymo car gotcha and there's seems like they're trying to set themselves up for the future there have they mentioned how much the current r&b is going into like automotive

of research because that can maybe be masking, you know, the true profitability of the data center business? Yes, that's a good question. Actually, they don't share that breakout, but they have said, at least on the revenue side, like it's tiny today, but they've got something like an $11 billion design win pipeline over the next six years. So they definitely see the revenue in this area accelerating, but no, I don't know how much of that is the kind of R&D spend.

but i suppose one thing to think about is you know these aren't all um totally separate there's definitely like overlaps between all these different elements of r&d so it's not it's not you've got these four completely disconnected teams right and there's like the the omniverse ai training you know leads into the autonomous stuff as well all right the last segment uh or different product line they just and again i forget all the names of their products but they launched a CPU, I believe. I don't know if it's commercially available yet, but this is a big leap outside of the GPU. And as we know from Intel's market cap, the CPU market is still quite large. What's their goal there? Do you think they can win a lot of customers in the CPU market?

Yes, it's interesting. So I suppose a little while ago, we mentioned the failed ARM acquisition. So ARM were, you know, this industry leading guys with CPUs. So they took the billion dollar write down but they have kept a strong partnership with arm so they've got a 20-year license now to use arms um ip to build really a range of cpus so the main cpu that they're building right now is part of their data center segment it's something called grace and so they're building don't go too deep down the kind of rabbit hole of the hardware but their latest um hardware platform for data centers is something called grace hopper so kind of the grace cpu from arm and then their hopper architecture for kind of putting it all together and you know when when

that's when that stuff comes online fairly soon it's going to be i think a 30 times improvement in performance compared to their current iteration of hardware and fear um but because they've so they've you know they've already extracted a lot of value out of that arm license in the data center segment but they've got the license so you know who's to say that they won't build um a complete range of cpus you know maybe for laptops for pcs maybe even for mobile phones so that is quite interesting um it's particularly interesting actually when we think about their manufacturing model because um at the moment they're using taiwan semi and samsung for the manufacturing Well, they've got a conversation open, evidently, with Intel right now,

potentially to partner with Intel to build their hardware as well, which is going to be really, you know, it's going to be great in terms of de-risking some of the concerns that we'll come on to in a bit. But quite interesting that they've got this ARM license. You know, maybe they're going to partner with Intel, but at the same time, perhaps start to compete with them for CPUs. So it'll be interesting to see how that plays out. Right, and Intel's moving to the foundry business,

So they might be OK with, you know, offsetting some of that demand as well. Yeah. And I suppose actually even right today, I'm probably not bang up to the minute on the news, but I think there's this chips bill being debated right now in the US. So something like a $52 billion subsidy package to try and build like a strengthened US manufacturing of semiconductors. Well, like a big chunk of that $52 billion looks like it's going to go to Intel. so that's going to help intel kind of scale out um so you know great partnership for nvidia if they get that operational because it's going to allow them to kind of de-risk some of their exposure to taiwan semi what is their relationship you just mentioned it what is their relationship like with taiwan semi do you see that as i guess what are the positives and negatives of it yeah

so um so i think it's good and it's strong we have seen some um some some sort of insights into perhaps what's happening behind the scenes right now so i mean first thing i suppose though is you know the partnership with taiwan semi is fundamental to the way nvidia operates because you know that is their fabulous business model which are all those financial benefits we talked about at the top of the discussion but at the same time it's this massive risk right now for nvidia so i dug through their recent 10k so that was only a few months ago they said our manufacturing is performed by taiwan semi and samsung and then we have a bunch of other companies to do assembly um well actually i saw a i saw a kind of in the rumor mill in the kind of tech news a few

days ago that there may have been a falling out between nvidia and samsung so it's possible actually there's this one particular person believes that now nearly a hundred percent of their manufacturing is happening through taiwan semi foundries so that's a you know It's kind of a key supplier risk there. There's obviously, you know, the geopolitical situation between the US and China, you know, bad things could happen if suddenly NVIDIA lose access to TSMC. Well, that could be a major problem for the company, clearly.

So I do like the attempts to kind of broaden out with Intel. Yeah, how difficult would it be for them to switch to another foundry? Well, I don't think Intel can do the 5 nanometer and 7 nanometer. well can they do you probably know if they can do seven nanometer but i don't even know if they can make it right now right yeah perhaps not and it's going to take them a long time to kind of scale up and let's not forget it's it's kind of i think it's kind of a 12 month lead time to go from like design to actually having the chips being manufactured i guess the you know the the foundry has to kind of tool up to to deliver that new design so like if they i

doubt it would really happen right because probably i mean if if if bad things happen between the US and China and TSMC, you know, if Taiwan was annexed, for example, there's probably bigger things for investors to worry about than kind of what's happening with NVIDIA. But yeah, it's going to put like a major supply chain problem for them that could kind of, you know, put the company on its knees for a good six to 12 months

while they ramp up production with an alternative. But, you know, on the good side, at least they haven't got this CapEx invested that's potentially at risk. but it's just a case of kind of moving their supply to another manufacturer right could be some short-term risk but eventually they'll figure it out because they do have the best products all right here's a maybe easy i don't know if this is an easier one to a risk to talk about but nvidia has a exposure to the cryptocurrency market can you explain why everyone in the crypto market loves

nvidia chips or gpus excuse me and how do you try to look at that because it can be so cyclical can can be so unpredictable yep okay so maybe let's start with kind of why are we even talking about these two things in the same sentence um so if i kind of wind back we talked about gpus were built for gaming but they do this highly paralyzable processing kind of a b and c all at the same time so the way cryptocurrencies originally worked if we think about the original cryptocurrency bitcoin and then the second biggest cryptocurrency ethereum well without without doing like a whole episode like an hour on crypto basically you've got this distributed ledger and part of not all of the complexity part of the complexity is getting everybody who's involved in the network

to have consensus basically agree with a single version of the truth so the way that was created with Bitcoin and originally with Ethereum was an algorithm called proof of work. So what proof of work is, it requires the miner, the guy who's trying to sort of create cryptocurrency, and there's millions of these people around the world, to do really a vast quantity of processing to solve a mathematical puzzle. That's what cryptocurrency mining is. Basically, you're doing like a load of maths to try and figure out the answer to a question, and then you'll kind of win inverted commas the next uh block in the chain like the next bitcoin um so these mathematical puzzles it turns out are highly parallelizable so gpus are perfect for running this kind of

processing so what it meant was cryptocurrency miners kind of went crazy and started buying up all of those graphics cards from amd and nvidia that were initially targeted for gamers and the crypto miners went out and bought them because they they had like a massive arbitrage you know they buy the gaming card and then they could generate much more value from that card than the cost of the card plus the kind of electricity to power it so that's so we had this sort of you know hump as kind of the value of crypto mining goes up and down um where a lot of cards have been owned by crypto miners now right now the value of cryptocurrency is down along with grey stocks um but also something interesting is happening with one of them ethereum it's kind

of the second biggest cryptocurrency what ethereum is doing is moving on kind of evolving from proof of work and they're going to adopt something called proof of stake and the difference there is um you don't have to do all that crazy math there's a kind of different algorithm different way of working out consensus getting agreement and so suddenly if it and this is due to happen it's thought sort of later in 2022 but i think i think there's no firm date just yet there's still a bunch of complexity to be ironed out but when um when this does take place when the merge happens and ethereum moves off of proof of work to proof of stake kind of all these guys that are doing ethereum mining their hardware's wasted so what's happening already is all these ethereum miners

are seeing this coming and they're starting to dump their graphics their gpu cards on you know ebay and newegg and other markets they're just trying to sell them because they're going to be valueless to them at some point in the future and so this creates a big problem for nvidia and i guess for amd too because um you know not only have they got this sudden boost in the secondhand market but also kind of retail prices are being impacted because you've got all these used cards coming onto the market so um so that's i see that as a short-term headwind um but but definitely you know that's something that that you have to think about when you're sort of thinking about the valuation of nvidia perhaps over the short to medium term what percentage of the revenue is

or do you know i guess they don't give it out but have you seen any data what percentage of revenue comes from crypto um so that they wouldn't but it would come from the gaming segment because it's like someone's bought the gaming gpu but they're using it for mining rather than playing games so it's in that kind of you know big chunk of revenue all right all right that makes sense what what threat is there from uh companies making the chips in-house like amazon do you see that as a big risk or is it or is nvidia competitively advantaged enough that it doesn't seem practical for everyone to do that. I mean, it's certainly not practical

for everybody to kind of do a Tesla and go and manufacture their own hardware. It's quite complex. And, you know, either you need like a fabulous model and you engage with the same, you know, TSMCs and Samsungs, or you need to build your own foundries.

So, you know, NVIDIA are in every data center. Perhaps they're going to lose some of that processing. I suppose some examples that we do know about um amazon uh have got their own line of silicon they're working on right now i think something called tranium um meta evidently are also having some success with their own silicon um but i mean to me this is such a big market and it's going to grow at such a rate you know i think

um 61 growth in the data center segment in the most recent quarter um no that may be short i think it's 83 growth in the most recent quarter like this doesn't have to be a winner like single winner scenario there's room for everybody to profit and you know amd and nvidia i think are going to do very well and i guess yeah i was thinking about that because you see the headlines for amazon you're like whoa this could be a big risk here but that happened and i know it wasn't specifically invidious customers but apple went in-house um a few years back they're the largest customer for taiwan semi but yet nvidia still found the market out there so i think you know that that fear might be overblown this is just such a such a big market and speaking of that

like you mentioned before nvidia is one of the largest companies in the world when i wrote it down it said the market cap was 377 billion but i think stocks have kind of bumped higher so we're closer to 400 billion now uh how at this price how do you think about the valuation kind of looking you know three five maybe even longer years out yeah so it's a hard one especially in the the there's so much uncertainty in the current market but let's try and put a few stakes in the ground but i'm not going to try and put hard numbers to any of this really um but maybe let's start with you know their revenue today 27 billion dollars a year um and the data center revenue was up 83 percent year over year and that data center revenue is kind of like a cagger of 66 percent

over the last five years so that's not like an immediate a recent bump like it's consistently growing at a crazy rate um so and you know what's the terminal rate of this stuff as well like to my mind there's probably decades of growth at this rate in data centers before nvidia starts to reach kind of saturation there is such a demand for this kind of processing um so um you know so i so it's a data center has just become the biggest revenue segment i think it's going to become increasingly so um and maybe another little sort of short to medium term stake in the ground so um like we know a few of the negative effect negative impacts coming over the next couple of quarters so we've already had that 1.35 billion dollar write down because of the acquisition

um but also a kind of tail end of the pandemic plus exiting their russia custom like russia business so nvidia are forecasting a 500 million dollar headwind because of russia and specifically china covid lockdown because um um china is quite a big market for the company um and so uh so i think we're going to see we're going to see a decline over the next couple of quarters while that has to wash through the same as this kind of big unknown of proof of stake you know all of these used graphics cards coming onto the market, probably, you know, continuing to have a kind of headwind on the gaming segment revenue. So I would imagine, you know, that the sort of downward impacts on gaming are not going to be offset by the upward impacts on data center over

the next couple of quarters. It's probably a little bit of pain to work through. But when we think about really the long term of this stuff, you know, I suppose we used to say every company is a tech company well it's very real to say really every company is now an ai company and with their data center segment nvidia have put themselves at the forefront of artificial intelligence like cutting-edge hardware and software that's powering innovation across almost every segment of industry in our personal lives so you know i'm i'm taking a sort of hand wavy let's look 10 years out um but but i certainly see you know significant growth opportunity, particularly in that data center segment. What do you think of the management

team? I've heard a lot of really good things about the CEO. I think it's Jensen Huang. Yes. He loves his leather jacket. That's his black turtleneck. That's his Steve. He has to have the specific clothing item, but no, that's not the important part. Yeah. What do you think of him and the management team at NVIDIA? He's great. He's a bit of a rock star. He founded the company it's been his vision um i've read an interview with him recently i can't remember the exact words but actually he never envisaged nvidia as being a gaming company i think he always had grander aspirations in some ways they kind of stumbled into um the fact that their hardware was optimized for ai and they got super lucky i suppose that suddenly you know ai is going to be

the engine room of all commerce so you know they've just found themselves in this fantastic sweet spot um i do like that um you know he's he's loved by his staff i think he's got like a fantastic glass door and comparably approval rating something like a 98 ceo approval rating um and i think he owns something like three and a half percent of the company today so you know it's a it's a big company that's quite a lot of money you know this is certainly his life's work um there's actually an interesting relationship with amd i think uh i think jensen might be um the cousin of amd's ceo lisa sue so you know kind of all these guys are sort of tied together in some way it's quite the family yeah he does seem yeah he is at least from the outside seems

maniacally focused on improving nvidia and he has been for decades so it's kind of one of those where I don't even know if he had all, like, I don't know what the stake is worth, $20 billion. But even without that, it seems like he just wants to win in these markets really bad whenever you hear him talk.

Yeah, totally agree. Yeah. All right. Last question. It's one we try to ask all our interviewees. It's the premortem. So what, how could this investment go poorly? Yeah, I guess we've touched on

a bunch of the big risks already. I mean, probably the biggest one is just the fact they have this 12 month development lifecycle. And maybe I hinted at it earlier. But they've had a bit of a misstep with Taiwan semi recently, it seems this is all kind of reading between the lines a little bit. So, you know, they've, they've probably overestimated the demand for their current series of hardware. And they've got too much supply. And I understand, i don't think this has been verified for sure understand they tried to cancel a bunch of orders with taiwan semi there may have been a bit of a sort of back and forth negotiation and they've

agreed that they'll postpone instead of canceling some of that kind of pipeline so you know clearly they've made a bit of an error there and they're you know possibly that's going to kind of weigh down on inventory in the future so you know even for a company like nvidia with the maturity they can still get it wrong with a kind of, you know, demand supply equation. Probably the second risk we've touched on as well is competition. You know, there is this growing ecosystem of homegrown chip makers. I've said, I've taken the position that there's kind of room for everybody to win here. But, you know, if Microsoft, Alphabet, Amazon, Tesla, you know, if all the big clouds, the hyperscalers, do develop their own AI chips, well, that's a big chunk of that total addressable

market that could get eaten up. I happen to think that there is room for everybody to win, but that's certainly a risk we should bear in mind. Maybe a slightly more interesting one, which is relatively unique to a company like NVIDIA, is kind of AI ethics. We're really early in the sort of life cycle of AI, and we don't really understand the impacts on society. And so, you know, NVIDIA are creating this capability, this capacity. If people are running maybe, you know,

unethical AI on NVIDIA hardware, well, that could be a reputational impact and, you know, could result perhaps in, you know, more stringent regulations, which could, you know, limit the market some way. um and then you know perhaps the biggest risk or you know maybe not that likely but massive impact is um this relationship with tsmc you know if it's true that taiwan semi are doing almost 100 of the manufacturing for nvidia if things go bad you know they're going to be really bad

um but what we tried to do at seven investing is we assign a risk rating and so you know when i looked at all of that my view of the risk rating with this was moderate risk as opposed to uh low high or very high and i do think um i do think there are risks but i think the evaluation today is quite attractive it's still an expensive company clearly um but really you know with with every company becoming an ai company i truly believe that um nvidia are positioned to to profit from that plus you've got this fantastic call option with automotive which could become you know potentially a multi-hundred billion dollar market in itself right and their track record of top line growth has been absolutely fantastic since what when was the gpu came out

in 1999 i mean it's been amazing yeah absolutely yep i'm very very consistent yep any more questions okay well for uh any any listeners that want to keep up with you what's the best place to do that other than uh seven investing which by the way i think we still have our our code which is it ccm well yeah uh i guess someone will hear an advertisement at some point but code ccm get a hundred dollars off yeah that's still running all right well feel free to use that but uh for for listeners that want to keep up with you where's the best place to do that uh best place to find me on twitter so my handle is at seven number seven luke hallard and what i'll do is once you guys tweet out this episode i'll also tweet a link to our strong buy portfolio so you've heard all about

nvidia today but there's actually two other companies we're kind of giving away the uh the insight on so simon our founders released one of those already and i'll tweet a link to that and we've got the third one coming up from my co-advisor dana in the next week or so all right perfect. Nice little tease to end there. Well, that's going to do it. We want to remind our listeners that Brett and I are not financial advisors. Anything we say or discuss here on Chit Chat Money is not formal advice or recommendation. We are, however, general partners of Arch Capital, so clients may have positions in the securities discussed in this podcast. Thank you all for listening. Thank you, Luke, for coming on the show. We'll see you next

time. Bye.