Nvidia Part II: The Machine Learning Company (2006-2022)
02:11:02 Share

Still got switch mafia greyhound in my head from the pump up.

Nice, nice.

IT is party how all like GPU companies. Like I was watching a bunch in video keynotes and AMD keynotes ready for this. And everyone is so like techno ion lighting, like it's like cypher before cypher o who.

Easy, you wait, you wait, you who got to? Easy you, easy you with you, sit me down straight.

Welcome to season ten. Episode des six have acquired the podcast about great technology companies and the stories and playbooks behind them. I'm been gillibrand and I am the cofounder and managing director of seattle based pioneer square labs and venture fund, psl ventures .

and David rosel. And I am an intel investor based in separate cisco.

And we are your hosts. When I was a kid, David, I used to stare into backyard bonfires and wonder if that fire flickering was doing so in a random way, or if I knew about every input in the world, all the air, exactly the physical construction of the wood, all the variables in the environment, if IT was actually predictable.

And I don't think I knew the term at the time, but model able, if I could know what the flame could look like, if I knew all those inputs. And we now know, of course, IT is indeed predictable, but the data in computer required to actually know that is extremely difficult. But that is what in video is doing today.

Then I love that. inter. I still, where is been going on with this?

And this was occurring me as I was watching jenson sharing the on universe vision for NVIDIA and realizing and video has really built all the building blocks, the hardware, the software for developers to use that hardware, all the user facing software now, and services to simulate everything in our physical world with their unbelievably efficient and powerful GPU architecture. And these building blocks, listeners aren't just for gamers anymore.

They are making IT possible to recreate the real world in a digital twin, to do things like predict airflow over a wing or similar cell interaction, to quickly discover new drugs without ever once touching a petri dish, or even model and predict how climate change will play out precisely. And there is so much to unpack here, especially in how and video went from making commodity graphics cards, now owning the whole stack in industries, from gaming to enterprise data centers to scientific computing, and now even basically off the shelf self driving car architecture, from manufacturer and at the scale that they're Operating at. These improvements that they're making are literally unfathomable to the human mind. And just to illustrate, if you are training one single speech recognition machine learning model these days, one does one model. The number of math Operations like ads or multiples to accomplish IT is actually greater than the number of grains of sand on the earth.

I know exactly what part of the research you ve got that from, because I read the same thing and I was like, you gotta be freaking and kidding me.

isn't that not? I mean, there's just nothing Better in all of the research that you and I both said. I don't think Better illustrate just the unbelievable scale of data and computer required to accomplish the stuff that they accomplished and how unfathomably small all of this is the fact that that happens on one .

graphics card. Yes, so great.

okay. Listeners, now is a great time to tell you about long time friend of the show service now.

Yes, as you know, service now is the A I platform for business transformation, and they have some new news to share. Service now is introducing A I agents. So only the service now platform puts A I agents to work across every corner of your business yeah.

And as you know, from listening to us all year, service now is pretty remarkable about embracing the latest AI developments and building them into products for their customers. A I agents are the next phase of this.

So what are A I agents? A I agents can think, learn, solve problems and make decisions autonomously. They work on behalf of your teams, elevating their productivity and potential. And while you get incredible productivity enhancements, you also get to stay in full control.

Yep, with service now, AI agents proactively solve chAllenges from I, T, H, R. Customer service software development. You name IT. These agents collaborate, they learn from each other, and they continuously improve handling the busy work across your business so that your teams can actually focus on what truly .

matters ultimately, service. Now, an agenda I is the way to deploy A I across every corner of your enterprise. They boost productivity for employees and rich customer experiences and make work Better for everyone.

Yep, so learn how you can put A I agents to work for your people by clicking the link in the shower notes or going to service. Now document slash A I dash agents, and after you finish the episode, come join the slack acquired data m slash slack and talk about IT with us are, I did, but without further to do, take us in. And as always, listeners, this is not investment advice. David, I may hold positions in security discussed and please do your own research.

God, I was going to make sure that you said that this time, because we are gona talk a lot about investing in investors in invidia stock over the years. That is very wild, wild journey.

So last we left our lucky heroes, Jason wang and NVIDIA in the end of our NVIDIA, the GPU company years, and in kenna, roughly know two thousand and four, two thousand and five, two thousand and six, they had cheated death, not 黄子， but twice, the first time in the super overcrowded graphics card market when their first king started. And then once they sort of, you know, jumped out of the frying pan into the fire of intel, now getting for them, coming to commodities, them, like all the other P. C.

I chips that pledged into the intel motherboard back in the day, and they bravely fend them off, they team up with microsoft. They make the GPU programmer. This is amazing to come out with programmable shadings with the g four three.

They power the x box. They create the cg programing language with microsoft. And so here we are.

It's not two thousand, four or two thousand and five and is a pretty impressive company. Public company building stock is high line after the tech bubble crash conquered the graphics card market. Of course, there's ati out there as well also come up again.

But the three pretty important things that I think the company built in the first ten years. So one we talked about this last, last time, these six months chip cycles for their chips. We talk about that, but we actually say the rate at which they ship these things.

I actually read down like a little less so. In the fall of nineteen ninety nine, the ship the first g force card g force two fifty six in the spring of two thousand g force two in the fall of two thousand g four two ultra spring of two thousand one. G four three. That's the big one, with the program able shapers. Then six months later, the g force three T I five hundred.

I mean, the Normal cycle, I think we said was two years, maybe eighteen months for most other competitors who just got largely .

left in the dust inking. He is gone at this point. But i'm thinking about intel, how often did intel ship new products, let alone n fundamental new architecture there? Two ty and then 3 ty， ty of twenty, about five.

Whatever did I feel like the intel product cycle is approximately the same as a new body style of cars?

Yes, exactly.

Every five, six years, there seems to be a meaningful new architecture change.

And intel is the driver of lords law, right? Like these guys ship and bring out new architecture at warp speed. And they continued that through to today, to one thing that we missed last time that is super important and becomes a big foundation of everything in video, becomes today that we're going to talk about. They wrote their own drivers for their graphics cards. And we are a big thank you for this and many other things to a greek listener, a very kind listener name, jeremy, who reached out to us in slack and point IT us to a whole bunch of stuff, including the agent omelet youtube channel.

So good. I've pride watched like twenty five agent omeo EV OS this week.

So, so good. Huge chat out to them. But all the other graphics cards companies at the time, and most preferable companies, they let the further downstream partners right.

The drivers for what they were doing in video is the first one that said, no, we want to control this. We want to make sure consumers to use in video cards have a good experience on whatever systems they're on. That meant a, that they couldn't sure quality, but b, they started to build up in the company. This like base of really needy, greedy, low levels software developers in this chip comment. And there not a lot of other chip companies that have capabilities like this.

no. And what they're doing here is taking out of bigger fix cost space means very expensive to employ all the people who are writing the drivers for all the different Operating systems, all the different o ms, all the different board that has to be compatible with. But they viewed IT as it's kind of an apple ask view of the world. We want the control, or as much control as we can get over, making sure that people using our products have a great user experience. So they were sort of willing to take the short pain of that expense for the long term benefit of that improved user experience with .

their products that their users high and gamers that want the best experience. They going to go out there going to spend the time three, four, five hundred dollars on an NVIDIA topical line graphics card. They're going to drop IT into the P, C.

That they build, know they wanted to work. I remember ing best around the drivers back in the day and thing that not work. Lake, this is superpower.

So all this is focusing. Of course, they have the third advantage in the company is programmable shapers, you know which A T I copies as well. Bit like they innovative, like they've never done all this. So all of this at this time. It's all in service of the gaming .

market and one seed to plant here, David, when you say the programmable shade ers developers, the notion of a NVIDIA developer did not exist until this moment. IT was people who wrote software that would run on the Operating system. And then from there, maybe IT would get that compute load, would get off loaded to whatever the graphic card was, but IT wasn't like you were developing for the GPU for the graphics card with the language in a library that was specific to that cards. So for the very first time now, they start to build a real direct relationship with developers so that they can actually start saying, look, if you develop for our specific hardware, their advantages for you and .

really are specific gaming card. It's like everything we're talking about, these developers, their game developers, all of this stuff is all in service to the gaming market. So you know, again, their public company, they have this great deal with microsoft.

They're bring out cg together. They're powering the x box three ism. They go from sub a billion dollar market cap company after the tech crash.

Up to five to six billion dollars can buy two thousand and four. Two thousand and five stock keeps going on a chair. By mid two thousand and seven, the stack reaches just under twenty billion dollar market cap. Now this is great, and this all the stories like this is peer play gaming. These guys have built such a great advantage in a developer ecosystem, in a large growing market, clearly, which is video games.

which on its own, that would be a great way of to serve. I mean, I think what's the gaming market today? Hundred eighty billion or something.

And when we talk to trip hawkins, who sort of like help the invented or a no bushi IT was zero then. And so in video is sort of like on a wave that's add an amazing inflection point. They can totally just ride this gaming thing and be an important .

it's not running out of steam. I mean, like how could you not be not to satisfied, but like more than satisfied with this? As a fact, yes, I am the leading company in this major market. This huge wave that I don't see ending anytime soon. You know, ninety nine point nine percent of founder who are themselves as a class like, you know, very ambitious.

are gonna satisfied with that.

But not just, but not gents. So while all this is happy, he started thinking about, well, what's the next chapter damning this market? I want to keep going. I don't want video to be just a giving up to.

So we ended last time with the little, you know, almost a surely a powerful story of a stanford researcher sends the email ed agency, and like, you know, thanks to you, my son told me to go buy off the shelf, you know, g force cars at the local fries, electronics. And I stuffed into my PC at work. And, you know, I ran my models on on this here.

I think he was a quantum chemistry researcher. Supposedly I was ten times faster than the supercomputer I was easing in in the lab. And thank you. I can get my life's work done in my lifetime.

And Johnson loves that code comes out at every G. T, C.

So that story, if you're a skeptical listener, my big two questions, first is a practical one. Know you just said everything about gaming here. And here's like a researcher, like a scientific researcher doing, you know, chemistry modeling using divorce cards for that. What's you writing?

This is, well, turns out program shatters, right?

They were shoe warning cg, which was built for graphics. They were translating everything that they were doing into graphical terms, even if IT was not a graphical problem they were trying to solve and writing in n cg, this is not for the faint of heart, so to speak.

right? So everything is sort of metaphorical. He's a quantum chemistry researcher, and his basically telling the hardware OK.

So imagine the data that i'm giving you is actually a triangle l and imagine that this way to that I want to transform the data is actually like applying a little bit of lighting to the triangle. And then I want you to output something that you think is the right color pixel. And then I will translate IT back into the result that I need for my quantum chemistry. Like you can see why that sub optimal. yep.

So he thinks this is an interesting market. He wants in video to serve IT. If you really want to do that, right? IT is a massive undertaking. IT was ten plus years to get to the company to this point.

You know what cg was is like a small sliver of the stack of what you would need to build for developers to use GPU in a general purpose way like we're talking about. You know, it's kind of like them they worked with microsoft to make cg. It's like the difference between working on cg and like microsoft building the whole that net framework for developing on window or today even Better.

Apple, right? Like everything apple gives to IOS and mac developers, right, to develop fun. Mac.

right? Yeah, the analogy is not perfect, but it's like instead of just apple saying, okay, objective sea is the way that you write code for our platforms, good luck. They're like, okay, well, will you need U I frameworks? So how about APP kit and coco touch? And how about all these other S, D, K, S. And frameworks, like A R kit and like store kit, and like home kit, is basically you need the whole sort of abstraction stack on top of the programing language to actually make IT very accessible to rights software for domains and disciplines that you are gonna really popular using that hardware.

exactly. So when jenson commit himself in the company to pursuing this, he's biting off a lot. Now we talked about theyve been writing in there and drivers. So they have actually a lot of very low level.

I an low level like bad, mean low level like infrastructural close, very difficult systems oriented programing talent within the company, so that can enables them to start here. But like so this is thick. So then the second question, if you're a discerning investor, particularly in the video that you want to ask at this point time is like OK ensor, you're committing the company to we big undertaking. What's the business case for that showed me the market. I would down valentine at this point would be sitting there listening to genson been like show me the market.

And not only is that show me the market, but it's how long will the market take to get here and it's how long is gonna take us and how many dollars and resources is gonna take us to actually get to something that's useful for that market when IT materializes because, well, CUDA development began in two thousand and six, that was not a useful usable platform for six plus years. Add video. Yep, this is closer .

to on the order of the microsoft development environmental, the national development environment. Then what in video was doing before, which was like, hey, we made to make the eyes and worked with microsoft so that you can program for my thing.

right? I want to flash way for register illustrate the insane undertaking of this. I search linked in for people who work at NVIDIA today and have the word kuta in their title. There are eleven hundred employees dedicated specifically to .

the good of platform. I'm surprised it's .

not even thousand. yeah.

okay. So is the market for this? Yes, you asked the third question, which is okay, the intersection of what does this take to do this and when is the market can get there in time and cost on that. But even just put that aside, is there a market for this is the first order question. And the answer to that is probably no at this point in time.

And what they're aiming at is scientific computing, right? Its researchers, who are in science specific domains who right now need supercomputers or access to a supercomputer to run some calculation that they think is going to take weeks or months and wouldn't be nice if they could do IT cheaper or faster. Is that kind of the market there looking out?

Yeah, they're attacking like the create market, like cay supercomputers, like that kind stuff, you know great company, right? But like snow in video today, they were dominated ating the market, you know yeah, it's scientific research, computing and its drug. It's probably a lot of this work they're thinking or maybe we can get into more professional like hollywood and architecture and other professional graphics still means yeah, yeah yes, sir.

But you know you some all that stuff up and like maybe you get to a couple billion dollar market, maybe like total market and not enough to justify the time in the cost of what you're going to have to build out to go after this to any rational person. So you know, here we come dense in the video, like they're doing this. He is committed.

He's drunk. Cooled two thousand six, two thousand seven, two thousand and eight. They're pouring a lot of resources into building what will become CUDA that will get you in a second. My R D is good at this point time.

And I think Johnson psychology here is sort of too fold. One is he is enormous red with this market. He loves the idea that they can develop hardware to accelerate specific use cases in computing that he finds sort of fanciful.

And he likes the idea of making IT more possible to do more things for a humanity with computers. But the other part of IT is certainly a business model realization, where he has spent the last gotcha this point thirteen, fourteen years being commoditized in all these different ways. And I think he sees a path here to durable different ideation where he's like woo.

to own the platform.

You know, it's kind of the apple thing again to own the platform and to build hardware that's differentiated by not only software but relationships with developers that use that custom software. Like then I can build a really sort of like A A company that can throw its weight around in the industry.

one hundred percent. Jenson, I don't know if he used to at the time because he probably would gone in purred, but maybe he did. I don't think he cared.

He's certainly has used IT since the way he thought about this or as a IT wasn't just like if we build IT, they will come, which is what was going on. The phrase users is if you don't build IT and they can't come. So it's that you would like i'm pretty sure if we build IT, they will come.

It's one step removed from that. It's like, well, if we don't build IT, they can even possibly come. I don't know they will come, but they can't come if we build.

So wall street is mostly willing to ignore this in two thousand and six, two thousand and seven, two dozen, eight. The company is still growing really nicely that this great market cap run leading up to before the finances crisis. But then, you know, you maintain last time, I think IT gets announced in two thousand and six, maybe in closes in two thousand and seven. A M D quires A T I O N A T I was a very legit competence, the only standing legit competitor to, in video there were told life. But now I M D, according tly, think they acquired for what, six, seven billion dollars.

Something like that, something like that.

There is a lot of money. And then they put a lot of resources like they weren't just acquiring ing this you know get some talent like there isn't a rack for behind this.

We haven't done the research in the AMD the way we haven't to n video, but the A M D radionet, which used to be the eighty I ready online. That is how you think about A M, D as a company, is that they make these G, P, S, mostly for the gaming use case.

Yep, before the acquisition, I think the first P C I built in like in a high school begin ecole, I think I had already on card in IT. I think I was probably in the minority. I think the video was bigger.

But for whatever reason, I like dt. I at that point time. So really legit. Well, so here's in video now focusing on the holiday thing.

And you're still in the gaming market, which liquid is like a massive rising tide. Your competitor now has all these resources. And A M, D, that's fully dedicated to going after IT mid two thousand eight in video. Wafs on learnings like this is naturally they took their right off the ball, of course they did, and the stock gets hammed because and anything .

that kuda empowers is not yet a revenue driver and they've totally taken the ya off of gaming.

yes. So we said the high was around the twenty billion dollar market cap IT drops eighty percent eight zero o this isn't just the financial crisis. It's almost going, I think, you know, for me, thinking back on the financial crisis now, and like people freaking out the door, S P dropping five percent of the third eight .

days IT is literally the thursday that we are recording.

Yes, for a company stock to drop eighty percent, technology company is stock even daring the first anal crisis? They are not just in the penalty box. They're like getting kick to the curb.

right? Are they done? The headlines at this point are and videos run over.

If you're most C, E, O. At this point time, you probably call them up goldmine or you know Allen company or franquet rone. And you happen this thing because how are you .

gonna recover? But not jenson.

But not jenson, obviously. So instead he goes and build cutters and continues to build cuti. And this is the second text is like we get excited to get a lot of stuff on acquired. But I think kuda is like one of the greatest business stories of the last ten years, twenty years more. I don't know.

What do you think then? I mean, i'd say it's one of the boldest bets we've ever covered, but so were programmable shatters and so was nvidia's original attempt to make a more efficient, quite lateral focus. Yes, graphics.

Those were big bets. I think this this is a bet on another scale, this is a bet that we don't cover .

that often on acquire. Those were big bets relative to the company's size at the time. But this bet is like an iphone sized.

That's exactly what this is. It's an iphone on size bet.

IT is a bet company when you are already a several billion dollar company.

yes, an attempt to create something that if they are successful and this market materializes.

this will be a generation company.

yeah. So what is cute? IT is in videos compute unified device architecture. IT is, as we've referred to, you know, that's far throughout the episode, a full and I mean full development framework for doing any kind of computation that you would want on G, P. U.

Yeah in particular, it's interesting because i've heard Johnson referenced as a programing language i've heard him referencing as a computing platform. IT is all of these things. It's an API.

IT is an extension of c or c plus plus. So there's a way that is sort of a language. But importantly, it's got all these frameworks and libraries that live on top of IT.

And IT enables super high level application development, you know really high abstraction layer development for hundreds of industries at this point to communicate down to kota, which communicates down to the G, P. U. And everything else that they have done .

at this point. So right after we released right the same day that we released part one, yeah the first invidia eis. So we did a couple weeks.

O ben thomson had this amazing interview with Jenny on statements. And h Jason in this interview, I think, puts what CUDA a is and how important is the Better than I seen anywhere else. This is Jason speaking to ban.

We've been advancing quota in the ecosystem for fifteen years and counting. We optimize across the full stack iterating between GPU acceleration libraries, systems and applications continuous, all while expanding the reach of our platform by adding new application domains that we accelerate. We start with amazing chips, but for each field of science, industry and application, we create a full stack.

We have over one hundred and fifty sd case that serve industries from gaming and design to life on earth sciences, quantum computing, A I cybersecurity, 5g and robotics。 And any talks about what I took to make this, this is like the point we give. We tried delic camera home here.

He says, you have to internalize that. This is a brand new programme model. And everything that's associated with being a programme processor company or a computing platform company had to be created.

So we had to create a compiler team. We had to think about S D case, we to think about libraries. We had to reach out to developers in the vandal as our architecture and help people realize the benefits of IT. And we even had to help them market this vision so that there would be demand for their software that they write on our platform and on and on and on.

it's crazy. It's amazing. And when he says that it's a whole new programing evictions, maybe paradigm or a way of programing IT is literally true because most programing languages up to this point and most computing platforms primarily contemplated serial execution of programs.

And what could I did was IT said, you know what, the way that r GPS work and the way that they are gona work going forward is tones and tons, of course, all executing things at the same time. Parallel programing, parallel architecture. Today, there's over ten thousand cores on their most recent consumer graphics card. So insanely, there is a embarrassingly parallel. And quota is designed for parallel execution from the very beginning that the catch phrase .

in the industry of embarrassin ly parallel and it's actually .

kind of a technical term.

I don't know why it's embarrassing.

It's basically the notion that this software is so parallelism, which means that all of the computation that need to be run are independent. They don't depend on a previous result in order to start executing. It's sort of like IT would be embarrassing for you to execute these instructions in order instead of finding a way to do a parallel.

It's not that it's parallel that's embarrass. It's embarrassing if you were to do at the old way on CPU serial.

I think that's the implication. This is so obvious, it's embarrassingly paralel.

Okay, now makes sense. Now, here is the cut graph. Where is spent a few minutes talking about how brilliant this was, everything we just described, this whole undertaking, the legs like building the pyramid ds of egypt or something.

Here, IT is entirely free in video to this day. Now this may be changing. We will talk about this at the end of the episode.

Never charged a dollar for CUDA, but anyone can then learn to learn to use IT know boba. All this work extend on the shoulders of everything in video is done. But then what is the but IT is .

close source and proprietary exclusively to invidia as hardware.

That's right. You do any of this work. You cannot deploy IT on anything but in video chips and that's not even just like, oh, in video put in the like terms of service that you can't to play this on and you know md trip service .

whatever like doesn't look deck.

It's like if you were to develop an IOS, a APP and then trying to play IT on windows like IT will not work. IT is integrated with the hardware, so open C.

L is sort of the main competitor at this point. And they do actually let open C, L, applications run on their chips. But nothing in CUDA is available to find elsewhere.

It's so great. okay. So now you can see this is just like apple and its apple business model. Apple gives us a way, all of this amazing platform ecosystem that they built to developers, and then they make money by selling they are hardware for very, very healthy gross margins.

But this is why jensen is so brilliant, because back when they started down this journey in two thousand and six, even before that, when they started, and then all through IT, there was no I O S, there was no iphone. And like IT, wasn't obvious that this was a great model. In fact, most people thought this was a dumb model that like apple lost and the mac is stupid and nit and like windows and tell is what one the open eco system well.

but windows and intel did have proprietary development environments and you know full stack dev tools.

Oh yeah, there's a lot of a new on here. It's not like they were like open source per say. But I could .

run on any hardware. Well, I accept that I couldn't I could only run on the intel IBM microsoft alliance world that wasn't running on power PC IT wasn't running on anything apple made. That's true.

It's funny in some ways. And video is like apple in other ways, they are like the microsoft intel IBM alliance except fully integrated with each other instead of being three separate companies. Yeah.

that's maybe a good way to put IT eight is sort of somewhere between there is nuance here remembering clay Christensen was bashing on apple in the early days of the iphone being like, yeah, yes, opens to win. And rowin apple is, do clothes never gotto be modulate you can be integrated. And like, you know, clay was amazing and one of the greatest strategic, but I think that's just representative to me of lake. Everybody thought that like the apple model .

sucked yeah I mean, IT sucks in less year at scale. And at the time, there was very little to believe that in video was going to have the scale required to justify this investment or that there was a market to let them achieve the scale to just try this.

That's the thing. Even if you were to say, okay, Jenny, I believe you and I agree with you that this is a good model, if you can pulled off at the time, you could be done. valentine. Er whoever looked around and maybe done was still looking around because they probably still held the stack like where's the market that's going to enable the scale you need to run this playbook?

All right. So you, onna, take us to two thousand eleven twelve, where we happen back in here.

If only the world were works like fixing. And I were actually like a truly straight line. It's never a straight line.

We will get there. And that is what saves. And video makes this whole thing work. But they have some big adventures in between. So a stocking heard it's two thousand eight and i'm just completely te speculating on my own.

But there in the penalty box, they're committed to continuing to investing quota and making general purpose computing on GPU a thing I do wonder if they felt like, well, we got to do something to a peace. Shareholders here, we got to show that were trying to be commercial here. So it's two thousand eight.

What's going on in two thousand eight? Need of in the tech world, it's mobile. So in two thousand, they launch the tega chip and platform within video.

This may not be saved.

The this is now what save the company. This is more climb car style. I think that maybe that's too well on in video. What was tegor p people might recognize that name.

IT was a full on system on a ship for smart phones competing directly with calm with samsung like IT was a processor like a an ARM based c CPU, plus all the other stuff you would need for a system on a chip to power android heads, head. This is like a wild departure for IT leverages none of the videos, core scale sets exit maybe graphics being part of smart phones. But like, come on, if there's ever a use case for integrated graphics, it's smart phones.

right? right? Low power, smaller footprints.

You total do you know this is one of my favorite ts about the whole research. Do you know what the first product was that shipped using a tiger, a chip?

Uh, no. IT was .

the microsoft zoo H. D. Media player. That just tells you pretty much everything you need to know. Uh, IT did though. The tiger system IT is so around sort of to this day, empowered the original tesla model s touch screen.

So before any of the autopilot autonomous driving stuff, they were the process are powering just the infotainment, the touch Green invocation in the model s and I think that actually starts to help. In video, get into the automotive market. The tegor platform still to this day is the main processor of the ninja switch.

Oh, they repurposed IT for that.

you for that. And they I think they still have there in video shield profiteering gaming device stuff that I don't know that anybody buys those.

Ah this makes so much sense because they basically have walked away from every console since the playstation three. yeah. And so it's interesting that they have the thriving gaming division that doesn't power any of the councils accept the intent do of switch. And I always sort of onder red like, why did they take on the switch business? Because I cannot really had IT done.

It's not for the graphics cards. IT was as somewhere to put .

the tiger's stuff fascinating, quick IDE. It's funny how these G, P. U companies have not been good at transitioning to mobile.

There is a funny naming thing, but do you know what happened to? So there's the A T I radio, which became the A M D radio desktop series. They tried to make mobile GPU. IT didn't go great, and they ended up spinning that out and selling all that IP to another company. Do you know the company?

Oh, I do not. Was a apple .

IT is quite calm. And IT today is quite comes mobile. G, P, U, division and quality comes good at mobile. And so is that the natural home for IT? Do you know what that line of mobile G P U processors .

is called?

No, IT is the R D O A R D N O processors. And do you know why it's called the R D O or r dino?

That sounds super familiar.

but no, the letters are rearranged from radio.

That's great. great.

See your say and videos mobile graphic s efforts didn't quite panel.

No, we didn't talk about this as much in the SONY episode. But my impression is the whole android value chain ecosystem is that there is no profits to be made anywhere. And google keeps that that way on purpose.

Ironically, they make a lot of money now .

on the play store, uh yeah, the play store and ads.

right? And do you think the primary way that they monotoned IT is not having to pay other people to acquire the search traffic, right?

But I mean, for like partners, like if you are making oh everything from chips all the way up through hardware in the android process and I don't think you're make IT like maybe if you were the scale player, like these things are designed to sell for dirt cheap as in products like there's no margin to be here here. Yeah yeah. Also, before we continue, you just sit the side bar on the M, D model graphics shape.

I see your side bar. I'm going to raze you one more sidebar that we have to include that, you know, because the N Z S. Guys told us about this. So when in video is going after mobile, they buy a mobile baseball company called I sera. British company called I sera in two thousand eleven.

Go with this. Oh yes, I so good place to back .

later to a be a thin bb ba. And then a few years later, when they end up, we pretty much shut down the whole thing. They shut down what they move from my sa. They lay everyone off the ice. A founders who meet a lot of money when invidia bought them, they go off and they found a company called the graph core that, uh, we're going to talk about a little bit of the the episode is the the primary .

sort of invidia bare cases.

video bare cases, invidia killers out there. They've never ried about seven hundred million and veta capital. So a mobile.

in some ways, it's kind of like BIOS and jet dot com. If jet had been successful, I think that sort of the graph core to NVIDIA analogy.

yes. Well, I injuries still out if anybody's going to be really successful in competing with the video, although I think the market now is probably ironically big enough in video and can be oil and there can be plenty of big other companies take up anyway. Okay, back to this time.

So in video is bumping along through all of us in the early late two thousands, early twenty tens. Some years growth is like ten percent. Maybe it's flat. And others like this company is completely inside with in two thousand and eleven, they wife on earnings, again stuck, goes through another fifty draw down.

It's because I was gonna say that I don't even if you can say about jenson, like here we are, the company is screwed again like everybody else want to give them up, but obviously not them. So what happens? Basically a miracle happens. I don't know that there's any other way that you can describe this except like a miracle. So maybe this is actually not a great strategy case study of jenson because IT required a miracle.

Well, Jenny would say was intentional, that they did know the market timing and that this strategy was right and the investment was paying off, and that they were doing this the whole time.

In fact, even in the bentham's interview.

I think he said ben basically lays out like, how did all these implausible things happened exactly the right time? And in his responses, oh yes, we planned at all. IT was so intentional.

jenson did not plan alex net or see a coming because nobody saw alex net coming. So in two thousand and nine, a printer computer science professor and also undergrad lum of princeton, just like is truly who and ondertook named to special as artificial intelligence. Computer science starts working on an image classifying project that he calls image net.

Now the inspiration for this was actually uh, way old project from like the eighties at princeton called word net that was like classifying words, which is classifying image image name. And her idea is to create a database of millions of labeled images, like images that they have a correct label applied them, like, this is a dog, or this is a strawberry, or something like that. And that with that database, then artificial intelligence, image recognition, Albertha could run against that database and see how they do so.

Like, oh, look at this image of, know you and I were looking, looking like that, a strawberry. But you don't give the answer to the algorithm, and the algorithm figures out of fit things is a strawberry doctor. Whatever in our collaborators start working on this is super cool.

They built the database to use a mechanical turk, amazon mechanical turk, to build IT. And then one of them, not exactly sure who if IT was failure or somebody else has the idea of, like, we know we've got this database, we want people to use IT. Well, let's make a competition.

This is like a very stand thing in computer science academy of like, let's have a competition and i'll grow them competition. So will do this annually. Anyone, any team, can submit their algorithms against the image and database, and we'll compete like who can get the lowest error rate, like the most number of images, percent the images correct.

And great. That brings her great. Reno becomes popular in the A.

I. Research community. SHE gets poached away by stanford the next year.

archive. O, cakes went there too. So that's fine.

And she's still there. I know that I can read. I couldn't resist that does.

She's like a kindred spirit to me. Do you know? I know you do know a bit. Most listeners do not know what her in doubt. Tennie chary is at stanford today.

I do. SHE is the sopa chair, yes.

the scotia capital professor of computer science at stanford. So cool. Why does he become this equality val chair? And what does all this have to do with in video? Well, in the twenty twelve of competition, a team from the university of toronto.

So bits and algorithm that wins the competition. And IT doesn't just win IT by lake a little bit. He wins IT by a what? So the way they measure this is one hundred percent of the images in the database.

What percent of them did you get wrong? So IT wins IT by over ten percent. I think IT had a fifteen percent alright, or something in the next.

like all the best previous ones have been like twenty five point something percent. Yes, this is like someone breaking the forbid at mile. Actually, in some ways it's more impressive than the forbid. IT, at my think, didn't brute force their way all the way there. They like try to completely different approach, and then boom, showed that we could get way more accurate than anyone else ever thought.

So what was that approach? Well, they called the team, which was composed of alex crisis. I was the primary leader, the team.

He was A P. H. D. Student and collaboration with ellia sutzkever and jeff hinton.

Jeff hn was the H. D. Advisor of alex. They called IT alex net. What is IT? IT is a convolution neural network, which is a branch of artificial intelligence called deeper learning.

Now, deeper learning is new for the use case, but ben is, you weren't exactly right. IT had been around for a long time, a very long time. And deep learning neural networks, this was not a new idea 啊。

The algorithms had existed for many decades, I think, but they were really, really, really computationally intensive. They required to train the models to do A A deep neural network. You need a lot of computer, like on the order of unity grains of sand that exists on earth. IT was completely impossible with a traditional computer architecture that you could make this work in any practical applications.

And people were forecasting too, like when with more law, when will we be able to do this? And IT still seemed like the far future, because not only did more slaw need to happen, but you also needed the invidia approach of massively paralyzed architecture. Or suddenly you could get all these incredible performance gains, not just because you're putting, you know more transistors in given space, but because you're able to run programs in parallel. Now yes.

so alex net took these old ideas and implemented them on GPU. And to be very specific, you implemented them in CUDA on invidia. G, P, S.

We can not overstate the importance of this moment, not just for video, but for, like, computer science, for technology, for business, for the world, for a staring at the screens of our phones all day, every day. This was the big bang ging moment for artificial intelligence. And in video. And kuta were right there.

Yep, this funny. There's another example within the next couple years, twenty twelve, twenty thirteen, where NVIDIA had been thinking about this notion of general purpose computing for their architecture for a long time. In fact, they even thought about, should we we launch r GPU as G P G P U. General purpose graphics processing units. And of course, they decided not to do that, but just built CUDA.

which is code word for a lake. We've been searching for years for a market. For this thing.

we can find a market. So you do is anything and deep a lot. competition. And so in twenty thirteen, brian cuttin ZARA, who's a research scientist at in video, published a paper with some other researchers at stanford, which included Andrew, where they were able to take this unsupervised learning approach that had been done inside the google brain team, where they had sort of the oogly st.

Brain team had sort of published their work on this, and I had a thousand nodes. And this is a big part of the sort of early neural network hype cycle of people trying cl stuff. And this team was able to do IT with just three notes.

So totally different models, super paralyzed, lots of compute for a super short period of time in a really high performance computing way. Or hpc is IT sort become known. And this ends up being the very core of what becomes cool. Nn, which is the library for deep neural networks that actually baked into CUDA, that makes IT easy for data scientists and research scientists everywhere who aren't hardware engineers or software engineers to just pretty easily, right, high performance, deep neural networks on and video hardware. So this alex, nothing plus then brian and Andrew wpp er IT just collapses all these sort of previously thought to be impossible lines to cross and just makes IT way easier and way more performance and way less energy intensive for other teams to do .

in the future and specifically to do deep learning. So I think at this point, like everybody knows that this is pretty important, but it's not that much of a leap to say if you can train a computer to recognize images on its own, that you can then train a computer to see on its own to drive a car on its own. One to play, chess to play, go to make your photos look really awesome when you take them on the latest iphone, even if you don't have everything right .

to eventually let you describe the scene, and then have a transformer model paint that scene for you in a way that is unbelievable that a .

human didn't make IT. yep. And that most importantly, for the market that Johnson and video looking for, you can use the same branch of A I to predict what type of content you might like to see next to show up in your feed of content, and what type of ad might work really, really, really well on you.

So basically, all of these people who are just talking about a bit, a lot of you recognize their names. They get stepped up by google. Fai goes to google.

brian went to buy you and he's back at the video now doing .

apply jeff and goes to facebook. So you know, all the other markets like given throw out say you don't believe in self driving cars as you don't think it's going to happen. Any of the other stuff like this IT doesn't matter like the market of advertising of digital advertising that this enables is a freaking multi drilling .

delight market. And it's funny because like that feels like who that's the killer use case, but that's just the easiest use case. That's the most like obvious well label data said that these models don't have to be amazingly good because they're not generating unique output.

They're just assisting in making something more efficient. But then like flash for ten more years an hour in these crazy transformer models with, I know a hundred and millions or billions of parameters, things that we thought only humans could do are now being done by machines. And it's like it's happening faster than ever.

yes. So I think to your point, David, it's like, oh, there was this big cash cow enabled by no neural networks and deeper learning in advertising. sure. But that was just the easy stuff.

right? But that was necessary though this was finally the market that enabled the building of scale and the building of technology to do. In genson interview, ban actually says this. When he sort of realizing this, talking to denser, he says this is been talking the way value across on the internet in a world of zero marginal cost, where there's just explosion and abundance of content that value accused to those who help you navigate content.

Talk about aggregation theory uh and then he says, what i'm hearing from you, Jenny, is that, yes, the value accused to people that help you navigate the content, but someone has to make the chips in the software so that they can do that effectively. And it's like a sort used to be with windows was the consumer facing layer and intel was the other piece of the winter l monopoly. Ly, this is google and facebook and hold us to other companies on the consumer side, and they're all dependent on in video. And that sounds like a pretty good place to be. And indeed, IT was a pretty .

good place to be.

amazing place to be. Oh my god, the thing is, like the market did not realize this for years. And I mean, I didn't realize this, and I you really didn't realize this. We were the class of people working in tech as venture catholic.

that of do you know the market .

in recent oh, no.

Oh, this is awesome. Okay, it's a couple years later, so it's like getting more obvious, but it's twenty sixteen. And mark rees and gave an interview.

He said we've been investing in a lot of companies, applying deep learning to many areas, and every single one effectively comes in building on invidious platforms. It's like when people were all building on windows in the nineties or all building on the iphone in the late two thousands. And then he says, for fun, our firm has an internal game of what public companies we'd invest in. If we were a hedged fund, we'd put in all of our money to. In video.

this is like he was paradigm, right, that called all of their capital and one of their funds and put IT into big coin. When I was, like, three thousand dollars, like we also have been doing this. So literally a video stack in twenty like recent, like this is now known twelve, thirteen, fourteen, fifteen.

IT doesn't treat above like five bucks a share. And in video today, as we record this, as I think about two twenty a share, the high in the past year has been well over three hundred. Like if you realized what was going on and and get in a lot of those years, he was not that hard to realized what was going on.

Wow, like IT was huge. It's funny. So there is even and .

will get what happened in twenty seventeen and twenty eighteen. We crippled down in a little bit, but there was a massive stock run up to like sixty five dollars a share in twenty eighteen. And even as late as I think, the very beginning of twenty nineteen, you could have gotten that I tweet this, and we will put the graph on the screen. And the youtube version here, you could have gotten IT in that crash for thirty four bucks S A share.

twenty sixteen.

If you zoom out on that graph, which is the next tweet here, that you can see that like in retrospect, that little crash, this looks like nothing. You don't even pay attention to IT and the crazy run up that they had three, fifty year, whatever the all time I was.

Yeah, it's wild. A few more wild things about this. It's not until twenty sixteen again, alex, that happens in twenty twelve. It's not until twenty sixteen that NVIDIA gets back to the twenty billion dollar market cap peak that they were in two thousand and seven when they were just a gaming company. That's almost ten years.

I really hadn't thought about IT the way that you're describing IT, but the breakthrough happened in twenty ten, twenty 10， twenty twelve. Lots of people had the opportunity, especially because frickin jenson is talking about IT on stage. You talking about our earnings calls at this point.

He's not keeping in this a secret.

No, he's like trying to tell us all that this is the future. And people are still skeptical. Everyone's not rushing divide the stock or watching this freaking magic happen using their hardware, using their software on top of IT.

And like even semon ductor analysts who are like students of listening to jenson talk and following the space very closely. So I think he sounds like a crazy person when he's up there. Are spouse that the future is neural networks.

And we're going to go all in. And we're not pivoting the business, but from the amount of attention that is giving in earnings calls to this versus gaming, I mean, everyone is just like a are you off your rocker? I think .

people are just lost trust in interest now after like there were so many years of like they were so early with kuta and early too. Didn't you know that? It's like they would know, alex, that was gna happen, right?

Jenson felt like the G, P, U, platform could enable things that the C. P. U.

Paradise could not. And he would, like, had this faith that something would happen. But I can to know this was good happen. And so for years he was just saying that like we're building that .

they were to be more specific, IT was that will look, the GPU has accelerated the graphics workload. So we have taken the graphics workload off of the CPU. The CPU is great at to your primary workforce for all sorts of flexible stuff, but we know graphics needs to happen in its own separate environment and heavily fancy fans on in and get super cold.

And IT needs. These matrix transforms the math that needs to be done as matrix multiplication. And there was starting to be this belief that, like, oh, well, because the professor, the apocalypse fessor, told me that he was able to use this program, the matrix transforms to work for him.

You know, maybe this matrix math is really useful for other stuff. And sure, IT was for a scientific computing. And honestly, like IT fell so hard into invidious lap that the thing that made deep learning work was massively paralyzed.

Matrix math, and they're like, envy is just like staring down their G. P. U. Like, I think we have exactly what you are looking for. Yes.

this that same interview with brand katayama, he says about when all this happened. He says the deep learning happen to be the most important of all applications that need higher competition under statement of the century. And so once in video saw that IT was basically instant, the whole company just latched onto IT.

There are so many things to log jenson for. You know, he was painting a vision for the future, but he was paying very close attention. And the company is paying very close attention to anything that was happening. And then when they saw that this was happening, they were not asleep at the switch.

Yeah hundred percent. It's interesting thinking about the fact that in some ways that feels like an accident of history, in some ways IT feels so intentional that graphics is an embarrassingly parallel problem because every pixel on a screen is unique. They don't have a core to drive every pixel on the screen.

There's only ten thousand cars on the most recent in video. Graphic cards, but there is not, which is crazy, right? But there is more pixels on a screen. So you know, they're all doing every single pixel at the same time, every clock generation. But IT worked out so well that neural networks also can be done entirely in parallel like that, where every single computation that is done is independent of all the other computations that need to be done.

So they also can be done on this super parallel set, of course, is just, you got a wonder, like when you canna reduce all this stuff to just math IT is interesting that these are two very large applications of the same type of math in the search space of the world. Of what other problems can we solve with parallel matrix multiplication? There may be more, there may even be bigger markets out there.

told IT. Well, I think they probably will be a big part of jenson's vision that he paints for any video now, which will get a this is just the beginning. There's robotics, there's autonomous vehicles, there's the army verse.

It's all coming. It's funny. We just talked about holic. Nobody saw this before they run up in twenty six and twenty seventeen. And there were all these years where like the market and new, you know whether he made money in his personal account or not, you know what does him? But then in twenty eighteen, another class of problems that are embarrassingly paralyzed is, of course, gypo currency mining. And so a lot of people were going out in buying consumer in video, you know, graphics cards and using them to set up cyp to mining riggs in twenty six and twenty seven. And then when the crypto winter hit twenty eighteen in the end of the ico craze and all that, the mining big demand, if I even this has become so big for in video that their revenue actually declined.

right? Yeah so couple of interesting things here. Let's talk about technically why so the way crypt opining works is effectively get in check.

You're effectively brute forcing an encryption scheme. And when you're mining, you know you're trying to discover the answer to something that is hard to discover. So you're guessing if that's not the right thing, you're incremental. You're guessing again and that's a vast simple vacation and not technically exactly right, but that's the right way to think about IT.

And if you we're going to guessed and check at a math problem and you had to do that on the order of a few million times in order to discover the right answer, you could very unlikely discover the right answer on the first time. But you know that probably ticket is only can happen to you once if r and so, well, the cool thing about these chips is that a, they have a crap ton, of course. So the problem like this is massively paralyzed, because instead of guessing in checking with one thing, you can guess in check with ten thousand at the same time, and then ten thousand more, and then ten thousand more.

And the other thing is IT is matrix math. So yet again, there is a third application beyond gaming, beyond neural networks. There is now the third application in the same decade for the two things that these chips are uniquely good at. And so it's interesting that like you could build hardware that's Better for crypto mining or Better for A I and both of those things have been built by NVIDIA and their competitors now. But the sort of like general purpose GPU happen to be pretty darn good at both of those things.

Well, at least way, way, way Better than a CPU.

也 as some of videos started up， competitors put IT today. And the rebirth is the one that i'm thinking of. They sort of say, well, the G, P.

U. Is a thousand .

times Better, you know, much, much Better than a CPU for doing this kind of stuff, but it's like a thousand times worse than IT should be. There are exist much more optimal solutions for, you know, doing some of this this A I stuff interesting.

really makes the question of like how good is good enough in these use cases.

right? And now I mean IT to flash way forward the game that in vidia and everyone else these upstarts are playing is really it's still the accelerated computing game, but now is how do you accelerate workloads of the GPU instead of off the cp?

interesting. Well, vector group that would to stock the timer again because there are another fifty percent draw down. This is just like every five years, this is gotto happen.

which is fascinating because at the end of the day was, I think, completely outside their control. If people were buying these chips for a use case that they didn't build the chips for, they had really no idea what people were buying them for. So it's not like they could even get really good market channel intelligence on, are we selling to crypto miners or are we selling to, you know, people that are going to .

use these for gaming the best buy e and then people go .

by and and best buy e right? And some people are buying a wholesale like if you're actually starting a data center to mine, but a lot of people are just doing this in the erased ment with consumer hardware, so they don't have perfect information on this. And then of course, the Price crashing makes IT either unprofitable or less profitable to be a minor. And so then your demand dries up for this thing that you U A didn't ask for and b had poor visibility into knowing if people are buying in the first place. So the management team just looks terrible to the street at this point because they had just no ability to understand what was going on in their business.

And I think a lot of street was still, still IT is handover of skepticism about because there's deeper learning thing like what jensen OK and so kind any excuse to sell off IT took anyway, that was shortly to the fifty percent depth because with the use case, and specifically the enterprise use case for G P S for deep learning like IT just takes off.

And so this is really interesting if you look at NVIDIA, um one of the report finances a couple different ways, but one of the ways they break IT out is few different segments is the gaming consumer segment. And then there are data in our segment and like data really good in the data that we are talking on. Google isn't going and buying you do in in video a GPU and hooking a muck to the laptops of their software engineer earthly.

Is stadia still a thing like I think that's used for cloud gaming .

and some like but it's happening in the da center is point right.

right? I guess what i'm saying and my argument is every time I see data in revenue, I in my mind, I sort of make IT synonymous with this is their ml segment.

Yes, yes, that's what i'm said. I I agree. yeah. Now, the data set, this is really interesting, again, because they used to sell these cards that would get packaged, put on a shelf, a consumer would buy them, yeah, they made some specialty cards for the scientific computing, marketing stuff like that. But this data center opportunity, you like men, do you know the Prices that you can sell gear to data centres for a like IT, makes the R T X thirteen nine, you look like pints.

And the R T X thirty ninety, which is the most expensive high, and graphics card at that you can buy as a consumer, was three thousand dollars. Now it's like two thousand dollars. But if you're buying, I don't know what's the latest. It's not the a one hundred. Is the eight one hundred.

the one hundred? They just announced each one hundred.

And that's what like twenty or thirty grand in order to just .

get one card yeah and people are buying a lot of things.

Yeah, crazy. It's crazy. It's unna. I tweet about this.

I was sort wrong with like everything there's nuance know tesla has announced making their own hardware. They're certainly doing IT for the on the car. The inference stuff link the false of dreaming computer on tesla s they now make those chips themselves, the tesla dojo, which is the training center that they announced.

They announce they are also going to make their own silicon for that. They actually done that yet. So they're still using and videos, tips for their training, the current compute cluster that they have that they are still using. I want to say I did the math I D like assume some pricing. I think they spent between fifty and one hundred million dollars that they paid in video for all of the .

computer in that all to one customer.

There's one customer for one use case at that one customer.

crazy. I mean, you see this show up and they're earning. So where is the part of the episode we were close enough to today that is best illustrated by the today number? So i'll just flash forward to what other segment looks like now.

So two years ago, they had about three billion of revenue and IT was only about half of their gaming revenue segment. So gaming, you know, through all of this through two thousand and six to alexa t all the way, you know, other decade forward to twenty twenty, gaming is still king. IT generates almost six billion in revenue.

The data center revenue segment was three billion, but had been pretty flat for a couple years. So then insanely over the last two years, IT three axed the data center, the segment three x and is now doing over ten and a half billion a year in revenue. And it's basically the same sizes as the gaming segment. It's nuts. It's amazing how IT was like sort of obvious in the mid twenty tens. But when the enterprises really showed up and said we're buying all this hardware and and putting in our data centers and the whether that's the hyper scales that like cloud folks, google, microsoft, amis on putting in, in their data centers or whether it's companies doing IT in their own private clouds or whenever they want to call IT these days on prime data centers, everyone is now using machine learning hardware in the .

data up and in video is selling IT for very, very, very healthy Grace margins. Apple level Grace margin? Yes, exactly. So speaking of the data center, couple things. One in this is so in video in twenty eighteen, they actually do change the terms of the user agreements of the consumer cards of g force cards that you cannot put them in data centers anymore.

They're like we really do need to start segmenting a little bit here, and we know that the enterprises have much more willingness to pay IT is worth that. Mean, you buy these crazy data center cards and they have like twice as many transistors and actually they don't even have video outputs like you can use the data center GPU like the a one hundred does not have video out so they actually can be used as graphic cards.

Oh yeah there is there's a cool um linus tech tips video about this where they get a hold of an a one hundred somehow. And then they run some benched Marks on IT, but they can actually .

like drive a game on IT.

Fascinating yeah so fun data .

center stuff is like super high horse power, but of course like useless to run a game on because you can't pipit to A T V or a monitor. But then it's interesting that the sort of artificially doing IT the other way around and saying, for those of you who don't want to spend thirty thousand dollars on this center, trying to like make your own little rigg at home, your own little data center of rig at home.

no, you cannot really don't think about go in the fries and ying. Ironic as that tell, the whole thing started. But anyway, in twenty twenty, they acquire an israeli data center computer company company called melon ox that I believe focuses on like, uh, networking computer within the data center here for about seven billion, integrate that into, you know, their ambitions in building out the data center. And the way to .

think about what men ox enables them to do is now they're able to have super high band with super low latency connectivity in the data center between their hardware. So at this point, they've got envy link, which is there. It's like the, what is apple color of her prieta in or connect or I think A M.

D. Calls at the infinity fabric. It's the like super high band with chip to chip connection. So I think about what melon ox lets them do is and lets them have these extremely high band with switches in the data center to then let all of these different boxes with the video hardware and then communicate superfast to each other.

And because, of course, these data is that the other thing about no customers, like a tesla example I gave there, are buying cards the end Price because they're buying solutions from invidia. They're buying big boxes with lots of stuff.

And you say solutions, I hear gross margin.

yeah. That's such a great quote we should like frame that and put in on the world quiet museum IT is true.

The acquiring melac not only like enables this, now we have the super high concept vy thing. But this is what leads to this introduction of this third leg of the stool of computing for any videos that they talk about now, which is, you had your C P U. great.

Is your work course. You know, it's your general purpose computer. Then there is the G P U, which is really A G P G P U, that they've really beefs ed up. And they're really like for the enterprise, for these data centres.

They've put tensor cores in IT to do the machine learning specific four by four by four metres multiplications super fast and do that really well. And they've put all this other non gaming data center specific A I modules onto these chips in the hardware. And now what are saying is you've got to C, P U.

You ve got your G, P U. Now there's A D, P U. And this data processing unit that's like kind of born out of the melanotan stuff, is the way that you really efficiently communicate and transform data within data centers. So the unit of how you think about like the black box just went from a back on iraq. Now you can kind of a think about your data center as the black box, and you can write at a really high abstraction layer, and then NVIDIA will help handle how things move around the data center.

All right? Listeners are next sponsor is a new friend of the show. Huntress interest is one of the fastest growing and most loved cybersecurity companies today. Its purpose built for small emd size businesses and provides enterprise grade security with the technology, services and expertise needed to protect you.

They offer a revolutionary approach to manage cyber security that isn't only about tech, it's about real people providing real defense around the clock.

So how does IT work? Well, you probably already know this, but IT has become pretty trivial for an entry level hacker to buy access and data about compromised businesses. This means cybercriminal activity towards smaller medium businesses is at an all time high.

So hunches created a full managed security platform for their customers to guard from these threats. This includes end point detection and response, identity threat detection, response security awareness training and a revolutionary security information and event management product that actually just got launched.

Essentially, IT is the full sweet of great software that you need to secure your business, plus twenty four, seven monitoring by elite team of human threat hunters in a security Operation center to stop attacks that really software only solutions could sometimes miss. Countries is democratizing security, particularly cyber security, by taking security techniques that were historically only available to large enterprises and bringing them to businesses with as few as ten, a hundred or a thousand employees at Price points. That makes sense for them.

In fact, it's pretty wild. There are over a hundred and twenty five thousand businesses now using countries, and they rave about IT from the hill tops. They were voted by customers in the g two rankings as the industry leader in end point detection and response for the eight consecutive season and the industry leader in manage detection and response again this summer.

Yeah, so if you want cutting edge cyber security solutions, backed by a twenty four, seven team of experts who monitor, investigate and respond to threats with precision, head on over to hunters dot com slash acquired or click the link in the show. Notes are huge things to hunters.

Okay, so I said one more thing on the data center.

Yes.

that one more thing is a easy to forget. Now I know because we've just been deep on this in video was gna buy ARM.

Do you remember this? Yes, they were. And in fact, this is going to be looking corporate communication, this nightmare.

Everyone out there, jenson, their I R person, different tech people who are being interviewed on various podcast, were talking about the whole strategy and how excited they are, own ARM and how and videos is gonna. You know, it's good on its own, but IT could be so much Better if we add ARM. And here's all the cool stuff are going to do with that, and that doesn't happen.

S, and now you've got dozens of hours of people talking about the strategy. zero. Almost like it's funny that now after listening all that, i'm sort like disappointed with nvidia's ambition on its own without having the strategic .

assets of ARM. Yeah, we should revisit our at some point. We did do the soft bank acquiring our episode years and years ago now.

But you know you think I am like they are a CPU architecture company whose primary use cases mobile and smart phones. It's like everything that intel screw up on back in the misguided mobile era. Now they're going in buying like the most important company in that space.

You know, it's just like again, in the panta interview, jen talks all about this. Maybe this is just justifying in retrospect, man, I don't think so. Like blink, he was about the data center yeah like everything ARM does is like great, fine.

But like we want to the data center and we say we want to own the data center. We want to own everything in the data center. And we think ARM chips, ARM cp use can be really a really important part of that. ARM is not focusing right now enough on that.

Why would they? Their core market is mobile. We want them to do that. We think there's a huge opportunity. We want to doll them in and do that.

And indeed, this year in video announced they are making a data center C P U and ARM based data center C P U called Grace to go with the new hopper architecture for their latest G P U. So there's Grace and hopper, of course, the rear admiral Grace hopper. I think I think that's right.

The am choice in the navy has great computer scientist pioneer. So yeah like data, it's it's big. Interesting of the objectors to .

that acquisition, and it's a good objection. And this is ultimately, I think, why the abandoned S I. Get the regulating pressure on this arms business is simple.

They make the IP, so you can license one of two things from them. You can license the instruction set. So even apple, who designs their own ships, is licensing the ARM instruction set.

And so in order to use that, I don't know what that actually is, twenty keywords or so that that get compiled to assembly language torn n on whatever the chip is. You if you want to use these instructions, you have to license that from ARM. great.

And if you don't want to be apple and you don't want to go build your own trips so you don't want to be in video, whatever, but you want to use our that instruction that you can also license these off the shelf chip designs from us, and we will never manufacture any of them. But you take one of these two things who you license from us. You have someone like T, S, M, C.

Make them great. Now you're a fabulous I conductor company, and they sell to everyone. And so of course, the regulatory body is gna step in being like, wait, wait.

So in video year, a files chip company, you a vertically integrated business model or you going to stop allowing ARM licenses to other people and n vd goes on. No, no, no. Of course, we would never do that over time. They might do some stuff like that.

But the thing that they were sort of like, where do we just believable beating the drama that the strategy was going to be is right now, our whole business strategy is that kuda at everything felt on top of our whole software services ecosystem is just for our hardware. And how cool would I be if you could use that stuff on ARM designed IP, either just using the I. S.

A, or also using actual designs that people license from them? How cool would I be if, because we are one company, we are able to make all of that stuff available for ARM chips as well? Yeah plausible, interesting. But no surprise at all that they face to a regulating pressure .

to go through with this no but clearly you an idea rattled around and gensys head a bunch in in videos because um well, let's cat us up to today. So they juster G, T C. At the end of march.

The big uh developer, the big G P U developer conference that they do every year that they started in two thousand nine as part of building the whole guity system out is so freaking impressive. Now like there are now three million registered included developers, four hundred and fifty separate S D case in models for CUDA. They announced sixty six zero new ones.

At this gtc we talk about the next generation G P U architecture with hopper and then the Grace CPU to go along with IT. I think hoper, I could be wrong on this. I think hopper is gonna be the world's first foreign omeo process chip using T S M C.

New foreign omeo process, which which is, that's right. amazing. Talk a lot about the universe, where I talk about on universe in a second.

But you think this licensing thing they usually do, their investor day, their own day at the same time is gtc. And in the analyst day, just IT gets up. There is just so funny. Go through the whole history of this now of like looking for a market, trying and to find some market of any size is like we are targeting a trillion dollar market. He's like a start up raising a seat round walk with it's stick.

We'll put this graphic up on the screen for those watching the video is a articulation of what the segments are of this trillion dollar addressable opportunity that in video has in front of IT. My view of this is if their stock Price wasn't what IT was, there is no way that they would try to be making this claim that they're going after a trillion dollar market. I think it's shy.

There's a lot of question there.

But the fact that they are valued today, I mean, what's their market cap right now, something like half a trillion dollars, they need to sort of justify that unless they are willing to have IT go down. And so they need to come up with a story about how they are going after this enormous opportunity, which maybe they are. But IT leads to things like an invest day presentation of let us tell you about our trillion dollar opportunity ahead. And the way that they actually articulated is we are going to serve customers that represent a hundred trillion dollar opportunity, and we will be able to capture about one percent of that.

It's just like if we can see the company pitted.

if we just get one percent of the market.

well, that's they really talk about this in narratives in a minute. But this is a generational company. This is unbelievable, is amazing there so much to admire here. This company did what like twenty something billion in revenue last year in his worth, half a trillion dollars.

They did twenty seven billion dollars last year in revenue.

Google ad words revenue in the fourth quarter of twenty twenty one was forty three billion. Google as a whole did two hundred and fifty seven billion in revenue. So you got to believe if you're an invidious shareholder.

right? There are the eight largest company in the world by market cap, but these revenue numbers do, you know, are in a different .

order of magnitude. De.

you gotta believe it's on the come. Yeah, you do. I mean, and video has literally three times the Price to sales ratio of apple or Prices to revenue as apple and nearly two x microsoft.

And that on revenue, I mean, fortunately, this NVIDIA story is not speculative in the way that early stage start up is speculative like even if you think it's overvalued, IT is still a very cash general business. Yes, they generate eight billion a free cash flows every year. So I think they're sitting on twenty one billion in cash because the last few years have been very cash energy very suddenly for them. So the takeaway there is, by any metric, Price to sales, Price earnings, all that there are much more Richard valued then uh, an apple or microsoft, these fan companies. But IT is you know extremely profitable business even on .

an Operating profits perspective. Well, so that any prize data goodness and makes money.

it's crazy. Now have a sixty six percent gross margin. So that illustrates to me how seriously different cities they are and how much of a mote they have verses can predators in order to Price with that kind of margin? Because think back, i'll put IT up on the screen here.

But back in ninety nine, they had a growth margin of thirty percent on their graphics chips and that in twenty fourteen, they broke the fifty percent mark. And then today in the slide really illustrates that it's architecture systems data center kuta CUDA x like it's like the whole stack of stuff that they sell l as a solution. And this sort of all bundle together and bundle is the right word. I think they get great economics because they are bungling so much stuff together. The sixty six percent growth .

margin business now yeah well and thinking about increasing that growth margin further and what we are talking about a minute ago with ARM in the licensing. So at the analyst day around gtc this year, they say that they're going to a start licensing a lot of the software that they make separately licensing and separate from the hardware like good. And there's a quote from James here.

The important thing about our software is that it's built on top of our platform. IT means that IT activates all of the videos, hardware chips and system platforms. And second, darling, the software that we do, our industry defining software. So we've now finally produce a product that an enterprise can license. They've been asking for IT.

And the reason for that is because they can't just go to open source and download all the stuff and make IT work for their enterprise no more than they could go linux, download open source software and run a multi billion dollar company with IT were joking a few minutes to go about, you say, solution and I C margin, you know yeah, like open sore software companies have become big for this reason on data, bricks, confluent and elastic. These are big companies with big revenue based on open source because enterprises they like, oh, I want that software. If they are not just going go to give you A D P.

Morgan, you're not going to go to get hub and be like, great, I got IT. Now you know you need solutions. So to jenson in in video, they see this as an opportunity to i'm sure this isn't gonna canalized hardware customers. Further, I think this is going to be incremental selling on top of what already doing.

That's an important point. And I think this is a playback team that I had. But often times when someone has hardware that is differentiated by software and services and then they decide to start selling those software gn services on a car, is a strategy conflict to your classic vertical versus horizontal problem unless you are good at segmentation and that sort of what n via is doing here, which is what they're saying, we're only going to a license to people that there is no way that they wanted just bought the hardware and gotten this stuff for free anyway. So if we don't think it's going to account ze and there are completely different segment that we can do things in pricing and distribution channel and terms of service that clearly walls off that segment, then we can behave in a completely different .

way to that segment. Further returns on our assets that we've generated.

Yeah, IT is a little tim cook though in you know tim cook beaten the service is narrative drum. I mean, that is going to you hear public company CEO who has a high market cap, everyone asking where the next phase of growth is going to come from and saying we're going to sell services and look at this growing business line of licensing that we have good this.

But who else is going to do at wearing .

a leather jacket at is a great point.

will. Talk about the .

okay. So few other things. Just talk about the business today that I think you're important to know, just as you sort of like think about sort of have a mental model for what in video is, is about twenty thousand employees.

We mentioned they did twenty seven billion in revenue last year. We talked about this very high revenue multiple or earnings multiple or everyone to frame at relative defame companies. They're growing much faster than apple, microsoft, google. They're growing at sixty percent a year. This is a thirty year old company that grew sixty percent in revenue last year.

Yeah, if you're not used to like wrapping your mind around that, like start up double and triple, but like in the first five years that they exist, google has had this amazing run where there are still growing at forty percent. Microsoft went from ten to twenty percent over the last decade. Again, amazing.

They're accelerating. But like infinity is growing us. Sixty percent, right? I don't care what your discount rate is.

Having sixty percent growth in your dcf model versus twenty or forty will get you a lot more multiple. Inflation be damned. Inflation be damned.

okay. Couple other things about specific segments of the business that I think you're pretty interesting. So they have not slept on gaming like we keep eating this and video data center enterprise machine learning argument yeah .

we even talked about retracing.

And right? Yeah, this R T X set of cards that they came out with, the fact that they can do retracing and real time. Holy crap, for anyone is looking for sort of a fun dive on how graphics works. Go to the wikipedia page for ray tracing. It's very cool. You model where all the lights sources are coming from, where all the POS would go in three d the fact that in video can render that in real time at sixty frames a second or whatever, while you're playing a video game is nuts. And one of the ways that they do that, they invented this new technology that's extremely cool, is called D L S S deep learning super sampling.

And this, I think, is like where NVIDIA really shines, bringing machine learning stuff and gaming stuff together, where they basically have faced this problem of, well, we either could render stuff at low resolution with less frames, because we can only render so much per amount of time, or we could render really high resolution stuff with less frames. And nobody likes less frames, but everyone likes high resolution. So what if we could cheat death? And what if we could get high resolution and high frame rate? And they're sitting around thinking, how on could we do that? Like, you know what, maybe this fifteen, your bet that we've been making on deep learning can help us out.

And what they discovered here in an invented in D, L, S, S. And AMD does have a competitor to this is a simple sort of idea. But this deal, S, S concept is totally amazing.

So what they basically do is they say, well, it's very likely that you can infer what a pixel is going to be based on the pixel around IT. Also pretty likely you can infer what a pixel gonna based on what I was in the previous frames. And so let's actually render IT at a slightly lower resolution so we can bump up the frame rate. And then when we're out putting IT to screen, we will use deep learning to artificially at .

the final stage of the graphics pipeline. Yes, yes, that's awesome.

It's really cool. And when you watch the side by side on all these youtube videos, IT looks amazing. I mean, IT dos involve really tight embedded development with the game developers. They have to reduced stuff to make A D L S, S enabled. But IT just looks phenomenal. And it's so cool that when you're looking at this four or even eight k output of a game yet for frame rate, you like wo in the middle of graphics pipeline, this was not this resolution and then they magically upscaled. It's basically making the like enhances joke, like a real thing that's so awesome.

I'm remembering back to the river a one twenty eight in the beginning of when they went to game developers, they were like, yeah, yeah. And all the blend modes in in direct dex, you know, you don't need just use this.

Yes, exactly, exactly. And they have the power to do IT. And I mean, they have the stick and the care IT with .

game developers to do IT. I mean this point, no game developer is not gonna make their games optimize for the latest and video hardware.

The other thing that is funny that's within the gaming segment because they didn't want to create a new segment for IT is script. So because they have poor visibility into IT before, they weren't liking the fact that I was actually reducing the amount of cards that were available to the retail chin nel l for their gamers to go and buy. What they did was they artificially crippled the card to make IT worse at crypt, ominous.

And then they came out with a dedicated crypt of.

And so like the charitable pr thing from NVIDIA is, hey, you know, we really we love gamers and we didn't want to make IT so that the gamers couldn't get access to know all the cards they want. But really, they're like, hum, people are just like strait up performing and orbital by crypt on mining on these cards. Let's make that more expensive on the cheap cards and let's make dedicated crypto harbor for them to bye to do those.

Let's make that arbitrage. Yes, your arbitrage is my opportunity. So magically.

their revenue is more predictable now and they get to make more money because much like their sort of terms of service data center thing, the terms of service, their way to be a lot to create some segmentation and those more profitability. Evil, evil genius life. The last thing that you should know about and videos s gaming segment is this really weird concept of adding board partners. So we've been oversimplifying on this whole episode saying, oh, you know, you go and you buy your R T X thirty nine T I. At the store and you run your favor game on IT.

But actually, you're not buying that from a video of the vast maturity of the time you are going to some third party partner asis msi h zo tag as one if there's also like a bunch of really low and ones as well who in video sells the cards to and those people install the cooling in the branding and all the stuff on top of IT, and you buy IT from them. And it's really weird to me that in video does that. I love how consumer .

gaming graphic cards have become the modern day equivalent of a hot rod.

Oh, dude, as you can imagine for the episode, i've been hanging a lot on the n video sub read IT. And like it's not actually about in video or in video the company or in video of the strategy. It's like show off your sick photos of your glowing rig, which is pretty fun.

Ty, but like IT feels like a remnant of old and video that they still do this, like they do make something called the founders edition card. And it's basically a reference design where you can buy IT from and video directly. But I don't think the vast majority of their sales actually come from that.

So it's like, um what are the android funds that google makes?

So I I suspect that shifts more over time. I can't imagine a company that wants as much control as and video does loves the ad and board partner thing, but they're built a business on IT and so they're not really willing to canabal ze an alienate. But I bet if they are their way and they're becoming a company that can more often have their way, they'll find a way to to can just go more direct.

make sense.

Two other things I want to talk about, one is automotive. So this segment has been like very small for a revenue perspective for a long time and seems to not have a lot of growth.

But jensen says in his pitch decks is going to be a three hundred billion dollar part of the dam. And I think right now.

it's something like, is that a billion dollars in revenue I think is like a billion dollars, but IT doesn't really grow.

I don't even know if that much.

Don't call me on that. So here's what's going on with automotive, which is pretty interesting. What n via used to do for automotive is what everyone used to do for automotive, which is make fairly commodity components that automated kers buy and then put in there. Every technology company has had their fanciful attempt to try to create a meaningful ly differentiated experience in the all have failed. You think about microsoft and the ford sink.

ford think o you think .

about car play, kind of maybe a little bit works. And the only company that really been successful has been tesla at starting like a completely new car company. That's the only way they're able to provide a meaningful differentiated experience. And video is my perception of what they are doing is they're pivoting this business line, this like flat, boring, undifferentiated business line to say maybe evs, electric vehicles and autonomists driving is a way to break in and create differentiated experience even if we're not onna make our own cars. And so I think what's really happening here is when you hear them talk about automotive now and they've got this very fancy name for IT, it's the something drive platform.

Oh, I drive is that is something like that.

something like that. But dealing with envy as product naming is madding. But this drive platform IT kind of feels like they're making the full E V A V hardware software stack, except for the middle glass and wheels, and then going to car companies and saying, look, you know how to do any of this.

This thing that you need to make is basically a battery and a bunch of GPU and cameras on wheels. And like you're issuing these press releases saying you're going in that direction. But is none of this is the core competency of your company except the sales and distribution. So like what can we do here? And if NVIDIA is successful in this market, it'll basically look like, you know an NVIDIA computer, full software hardware with a car chasa around that, that is branded by whatever the car company .

is like the android market.

yeah. And I think we will see if the shift to autonomous vehicles is a real b near term and see enough of a dislocation in that market to make IT so that someone like NVIDIA, a component supplier, actually can get to own a bunch of that value chain versus the automated manufacturer cannot forever stubbly getting to keep all of vin control. The experience you which .

to do a mini bowlen bear on this here before we get to the broader on the company. You know the bull case for that is we were in friend of the show Jimmy messing with in slack lotus is one of their partner. The lotus couldn't go build autonomous driving software like I don't think so for ari. And no.

you know not at all there going to be an video card effectively.

yeah.

okay. Last segment thing I want to talk about is how we open the show talking about the NVIDIA omniverous. And this is not on the verse like metaverse IT is similar in that is kind of a three d simulation type thing. But it's not an open world that you wander around in the same way that meta is talking about or that you think about unfortunate something like that. What they mean by the universe is pretty interesting.

So a good example of IT is this the earth to this digital twin of earth that they're creating, that has these really sophisticated climate models that they're running, that basically is a proof concept to show enterprises who wants to license this platform. We can do super realistic simulations of anything that's important to you and what their pitches to the enterprise is. Hey, you've got something.

Let's say IT is a bunch of robots that need to wander around your warehouse to pick and pack. If it's a amazon who actually amazon is the customer, they showcase amazon and all the fancy videos and they say you're gonna using our hardware often are to train models to figure out the routes for these things that are driving around your data centres. You're going to be licensing certainly some of our hardware to actually do the influence to put on the robots that are driving around.

When you want to make a tweet to a model. You're not just going to like deploy those to all the robots. You cannot want to run that in the omniverous first. And then when it's working, then you wanted to play y in the real world and their own universe. Pitch is basically it's an enterprise solution that you can license from us where any time you're going to change anything in any of your real world assets, first model IT in the on universe. And I think that's a really powerful, like I believe in the future of that in a big way because I think now that we have the compute, the ability to gather the data and the ability to actually, you know, run the simulations in in a way that has efficient way of running IT in a good user interface to understand the data, people are gonna stop testing in production with real world assets, and everything's gonna be model in the on universe first, before rolling out.

This is what an enterprise metaverse is going to be. This is not designed for humans. Humans may interact with this.

There will be U. I. You'll be able to be part of if the purpose of this is for simulating applications. And most of IT, I think, is gonna run with no humans there.

Pretty crazy.

Yeah, good to 想 找 一个 good。

You want to talk bar ball is on the company.

Let's do IT analysis.

So I mean, they paint the book case for us when they say there's one hundred trillion our future, we're going to capture one percent of IT. There are three hundred billion from automotive. Here's the four, five segments that add up to trillion dollars of opportunity.

sure. That's like a very neat way with a bow on IT in a very wish. You washing your hand, wave your way of articulating IT. So the question sort of becomes where's AMD fall in all this? There a legitimate second place competitor for high and gaming graphics. And I think we will continue to be that feels like a place where these two are going to keep going head to head the bare cases that there's a tiktok rather than a thor's competitive advantage for in video, but most high and games you can play on both M, D and n video hardware at this point.

The question for the data center is, is the future these general purpose G, P, U, that NVIDIA continues to modify the definition of GPU to include specialized you know functions as well? All the other stuff they are putting on there in their hardware? Or is there someone else who is coming along with a completely different approach to accelerated computing, whether accelerating workloads off the GPU onto something new like a surprise or like a graph core that is going to eat their lunch in the enterprise A I data center market?

That's an open question. You know, it's interesting like people have been talking about that for a while. The other big bear case that people have been talking about again for a while now is, you know, the big, big customers of in video that are paying them a lot of money.

The tesla is the google, facebooks, the amazon's, the apples, and not just paying them a lot of money and getting you do assets. The value of that, do you paying high girls margin dollars to in video for what they're getting that those companies are gonna want to say, you know, it's not that hurt to designer on silicon, bring all this stuff in house. We continue IT to exactly are use cases sort of similar to the servers graf core bare case on the video. I think in both of these cases, you IT hasn't happened yet.

Well, there have been a lot of people who have made a lot of noise. There have been few that have executed on IT like apple has their own gp s on the m ones. Tesla switching hasn't happened yet. A switching the to their own for the full self driving, their doing their own stuff on the car and .

they're switching, that is switch on the inference side on device. Yes, that has happened. Video is strong in that. But I think the real thing of what .

is the data center and google is probably the biggest fair case there is interesting to talk about these companies in particularly, sorry, risk. What they're doing is such a gigantic s swing in a totally different take then what everyone else has done for fox, who hasn't sort of follow the company there making a chip that's the size of a dinner plate. Everyone else's chip is like a thumbnail, but they're making a dinner plate size chip. And you know the yields on these things kind of suck. So like they need all the redundancy on those chips to make IT.

So that, oh my god.

the amount of expense to do that, right? And you can put one on a way for these wafers are crazy expensive to make.

wow. So you get poor yields in the wrong places on a wafer. And like that whole wafer is toast.

right? So a big part of the design of three verses, this sort of redundancy in the ability to turn off different pieces that aren't working. They draw sixty times as much power their way more expensive.

Like if video is gonna you a twenty or thirty thousand dollars, sir is going to sell you a two million dollars to do A I training. And so IT is this bet in a big way on hyper specialized hardware for enterprises that want to do these very specific AI workloads. And it's deployed in these beta sites in research labs right now.

And you know, not there yet, but it'll be very interesting to watch if they're able to meaningfully compete for what everyone thinks will be a very large market. These enterprise AI workloads that I mentioned, google that made a bunch of noise about making their own silicon in the data center and then stayed the course and stayed really serious about IT with their TPU. Their business model is different, so nobody knows what the bill of materials is to create a TPU.

Nobody knows really what they cost to run. They don't retell them. They're only available in google cloud. And so google is sort of counter positioned against NVIDIA here where they're saying we want to differentiate google cloud with this offering that depending on your workload, that might be much cheaper for you to use TPU with us then for you to use in video hardware with us or anyone else. And they're probably willing to eat margin on that in order to grow google cloud share in the cloud market.

just so kind .

of the android strategy, but run in the data center. One thing we haven't mentioned.

but we should, is cloud is also part of the NVIDIA story too. Like you can get in video gp u in A W S has her an and google cloud. And that is part of the gore story for in video too.

In n videos starting their own cloud, you can get direct from n video cloud base G P A.

center G P S. Just yeah.

I will be very interesting to see how this all shakes out with the then video that starts .

tup s and with google. I mean, all that said, like I think we look in videos very, very, very, Richard, valued on evaluation bases right now, very with another very in there IT depends if .

you think their growth will continue. Are there a sixty percent growing company year over year over year for a while, then they're not originally valued. But if you think it's a covet hick up or crypto hike up.

but to the the bull bear case and cannot both the startups and the big tech companies doing this stuff in house, it's not so easy here you know like yeah facebook and tesla, google and nazon and apple are capable of doing a lot. But who would just told the whole story? This is fifteen years of CUDA, and the hardware underneath IT and the library is on top of IT. The in video has built to go recreate that and surpass IT on your own is such an enormous, enormous to bite.

yes. And if you're not a horse zonal player and your vertical player, you Better believe that the pod goal that the end is worth IT for you for this massive amount of cost to create what NVIDIA has created. You like NVIDIA has the benefits getting deserve every customer. If your google and their strategy is what I think IT is of not retAiling T P S at any point then your customers, only yourself. So you're constrained by the amount of people you can get to use google cloud.

What analysts with google, they have google cloud that they can sell IT through. Yep, power or so .

at the way. I want to do this section because in our n video episode recovered the first thirteen years of the company. We talk a lot about what does their power look like up to two thousand and six. And now I want to talk about, what does their power look like today? What is the thing that they have that enables them to have a sustainable competitive advantage and continue to maintain pricing power over their nearest competitor, be google, to request in the enterprise or A M D N gaming?

yes. And just do a numerous the powers, again, as we always do, counter positioning, scale economies, switching costs, network a economies, process power, branding and .

cornet resource. So there are definitely scale economies. The whole CUDA investment, yes, not at first, but definitely now, is predicated on being able to admit tize that a thousand plus employees spend over the base of the three million developers and all the people who are buying the hardware to use what those developers create.

This is the whole reason we spent twenty minutes talking about if you are going to run this playback, you needed an enormous market to justify the capeci you we're going to put IT in.

right? So very few other players have access to the capital and the market that and video does to make this type of investment. So they are basically just competing against A M D.

For this. Totally agree, scale economies to me is like the biggest one that pops out to the extent that you have log in to developing on cuter. What I think a lot of people really have locked in on CUDA and that's major switching cost. Yeah like if you're onna boot out in video, that means .

you're booting out cuddle is CUDA accorded resource?

Oh interesting maybe I mean, that only works with a video hardware.

You could probably make an argument, there's process power or at least there was somewhere along the way with them having the six month ship cycle advantage that probably has gone away since people trade around the industry laden. That wasn't sort of a hard thing for other companies to figure out.

Yeah, I think process power definitely was part of the first instantiation of in videos power to the extended had power.

right?

Yeah, I don't know as much today, especially because T, S, M, C, work with anybody.

In fact, smc is working with these new start up billion dollar funded silicon companies.

Yes, they are. yes. Yeah, it's fine.

I actually heard a rumor, and we can link to IT in the shown notes that the empire series of chips, which is the one immediately before the the harper, the sort of a series PS, are actually fact by samsung, who gave a sweet.

hard deal n vid, likes to keep the .

law alive around T, S, M, C, because they ve been this a great long time partner and stuff. But yeah, they do play manufacturers of each other. I even think that jenson said something recently, like intl l has approached us about fb, bing, some of our chips, and we are open to the conversation.

Yes, yes, that did happen.

So there was a big cyber security hack a couple of months ago by this group lapses. And they stole access to and videos, source code. And actually, Johnson went on yahoo finance and talked about the fact that this happened, I mean, is a very public incident.

And it's clear from the demands of lapse is where some of nvidia's power lies because that they demand two things. They said, one, get rid the cypher, the governors like, make IT so that we can mind, which may have been a red hearing that might have just been them trying to look like a bunch of, like, kyp do, minor people. And the other thing they demanded is that in video, open source, all of its drivers and make available and source code.

I don't think IT was for CUDA. I think IT was just the drivers. But IT was very clear that, like, we want you to make open your trade secrets so that other people can build similar things. And that, to me, is illustrative of the incredible value and pricing power that NVIDIA gets by owning not only the driver stack, but you know, all of cute how tightly couple their hardware software is in video.

Is we just added this restoration an episode, hamilton and Cheney. And video is a platform in my mind, no doubt, about IT, kuta and NVIDIA and general purpose computing on GPU as a platform. So whatever, you know, all of the stew of powers that go into making, that they go into making apple, microsoft, you know, and lock go to in media.

Yep, I think the stew of powers is the right way to phrase that. Yes, anything else here you want to to play book .

how to be the playbook. So mean, I have I just wrote down in advance one that is such a big one for me. And i'm bias because I I, I try to think about this in investing, particularly in public markets investing. But like, man, you really, really want to invest in whoever is selling the pigs and the shows in the gold rush, the AI. You know, M L, deep learning gold rush h those years, gosh oh my gosh, like we should all all be kicking ourselves of twenty twelve thirteen, maybe not twenty twelve but certain ly twenty fourteen, twenty fifteen in the twenty sixteen like uh you know american reason saying every start up that comes in here he wants to do A I and deep learning and they're all using in video like maybe we should imported in video like I don't know if any one of those started up any given one is gna succeed. But i'm pretty share in video it's gna succeed back .

then such a good point kicking myself. One I have is a being willing to expand your mission. So as funny how Jason really days would talk about to enable graphics to be a story telling medium. And of course, this LED to the invention of the pixel shader in the idea that everybody can sort of tell their own visual story, their own way in a social network, real time way, very cool.

And now it's much more that wherever there is a CPU, there is an opportunity to accelerate that CPU and NVIDIA will bring accelerated computing to everyone, and we will make all the best hardware, software and services solutions to make IT so that any computing workload runs in the most efficient way possible through accelerated computing. That's pretty different. That enabled graphics as a story telling medium. But also they need to sell a pretty big story around the town they're going after.

I think there's also something to a the whole and videos story and across the whole work at the company of you. Sera, try this thing at this point and start up. And but so few companies and founders can actually do IT just not dying.

Yeah, they should have died at least four, several times and they didn't. And part of that was brilliant strategy. Part of that was things going their way. But I think a lot part of a too was just the company and janson, particularly in the most recent chapters where there are already a public company to spin like, yeah, i'm willing to just sit here and endure this pain and I have confidence that like we will figure IT out the market will come, not onna declare game over.

One that I have is we mentioned at the top of the show, but the scale of everything involved in machine learning at this point, and anything semiconductors, is kind of unfathered. Uni mentioned falling down the youtube grab hole annalee channel, and I was watching a bunch of stuff on how they make the silicon wafers. And my god, floor planning. Is this just unbelievable exercise at this point in history, uh, especially with the way that they sort of overlay different designs on top of each other on different .

layers of the the trip? Funny how they keep creating .

these sort of real world large scale analogies to chip so flor planning the way that an architect with layout the fifteen rooms in a house or five rooms in a house or two rooms in a house on a chip is laying out all of the circuitry and wires on the actual chip itself except of course, there's like ten million rooms. And so it's incredibly complex. And the state that I was going to bring up, which was just mind bending to think about, is that there are dozens of miles of wiring. N, A, G, P, U, wow, that is mind .

bending because these things are like they are less than the size. Your pom, right, right? And that .

obviously wiring in the way you think about, like a wire to reach done and pick up my, but it's wiring in the E U. V eed substrate on ship exposure as pride, the term that i'm looking for, her photo photography exposure. But IT is just so tiny. I mean, you can say foreign ameers all you want, David, but that won't register with me how freaking tiny that is until you're sort of face with the reality of dozens of miles of critical wires on this trip.

Yeah it's not like to me that registers is like, yeah it's like a my I got the best version but yeah like that's what that means. Okay.

here's one that I had that we actually talked about what I think could be fun. So I generated a capex graph of .

fun all showed on screen .

here for those watching on video. Obviously, there is a very high looking line for amazon because building data centers and for film centers is very expensive, especially the last couple years when I do in this massive build out. But imagine without that line for minute, NVIDIA only has a billion dollars of capex per year.

And this is relative for people listening on audio, relative to a bunch of other me know.

thank companies. yes. So apple has ten billion dollars of spent on capital expenditure per year. Microsoft, google have twenty five billion. T S, M, C, who makes the trip has thirty billion. What a great capital efficient business that invidia has on their hands, only spending a billion dollars year in cappel. It's like it's a software business and then basically is what IT is.

right? Like T, S, M, C does the fabling video makes software. And I P.

yeah. So here, this is the best crafts for you did very clearly see the magic of the fabulous business model that more chain was so gracious to invent when he grew T. S. M. C.

Thank you. morus.

Another one that I went to point out, it's a freaking hardware company. I know within they're not a hardware company, but there are hardware company with thirty seven percent Operating margins. So this is even Better than apple and for non finance folks, Operating margins.

So we talked about their sixty six percent gross march, and that's like unit economics, but that doesn't account for all the headcount. And no, this is and just all of the fixed costs in running the business, even after you subtract all that out, thirty seven percent of every dollar that comes in gets to be kept by via shareholders. It's are really, really, really cash generating business. And so if they can continue to scale and can keep these Operating margins or even improve them because they think they can improve them as really impressive.

wow, I don't realize that's Better than apples.

yeah. I think it's not as good as like facebook and google because .

they just run these like basically .

zero costs digital manaus lies in some of the largest marketing history but is still very good. We want to think our long time friend of the show, vantine, the leading trust management platform, venta, of course, automate your security reviews and compliance efforts. So frameworks like soc two I saw twenty seven o one gdpr and hip compliance and modeling ing venit takes care of these otherwise incredibly time and resource training efforts for your organization and makes them fast and simple.

Yeah, fanta is the perfect example of the quote that we talk about all the time here and acquired jeff bases his idea that the company should only focus on what actually makes your beer taste Better. I E spend your time and resources only on what's actually gonna, move the needle figure product and your customers, and outsource everything else that does IT. Every company needs compliance and trust with their vendors and customers. IT plays a major role enabling revenue because customers and partners demand IT, but yet IT add zero flavor to your actual product that IT takes .

care of volvo for you, no where spread sheet, no fragment to tools, no media reviews to couple together your security and complaints requirements in is one single software pain of glass that connects to all of your services via s and eliminates countless hours of work for your organization. There are now A I capabilities to make this even more powerful. And they even integrate with over three hundred external al tools, plus they let customers build private integrations with their internal systems.

And perhaps most importantly, your security reviews are now real time instead of static, so you can monitor and share with your customers and partners to give them added confidence.

So whether you are start up or a large enterprise and your company is ready to automate compliance and streamline security, like vantage seven thousand customers around the globe and go back to making your beer taste Better, head on over to vantage outcomes losh required and just tell them that ban and David sent you.

And thanks to friend of the show, Christina anta CEO, all acquired listeners get a thousand dollars of free credit vented outcome slash acquired ah okay grading. So I think the way to do this one, David, is what's the a plus case? What's the c case? What's the f case?

I think so.

And the sort of an interesting way to do this one because you could do IT from a shareholder perspective, where you have to evaluate IT based on where it's trading today and sort of like what needs to be true in order to have A, A plus investment starting today.

that sort of thing. You mean like a Michael mobile expectations investing style P S.

exactly. Or you could sort close your eyes to the Price and say, let's just look at the company. If you're jenson, what do you feel would be N A plus scenario for the company regardless of the investment case? I kind of think you have to do the first one though, like I kind of think it's a cop out to not think about IT. Like what's the bull and bear investment case from here?

As we point IT out many times on the episode, there's a lot you got to believe to be a on video at this share Price.

So what are they? Well, one big one is that they continue their incredible dominance. And they are, what are they growing like?

Seventy five percent or something you over year in the the data center, and they just sort of continue to owe that market. I think there is a plausible story there around all the crazy growth margin expansion. They've from sort of selling solutions rather than, you know, fitting into someone else of stuff. I also think with the milano x acquisition, there's a very plausible story around this idea of a data processing unit and around being your one stop shop for A I data center hardware. And I think rather than saying, like all the upstart competition will fail, I think you canna have to say that n video will find a way to learn from them and then integrated into their strategy too.

which seems .

plausible. Yeah, but theyve been very good at changing the definition of GPU over time to mean more and more robust stuff and accelerate more and more computer or clothes. And I think you just have to kind of bet that because they have the developer attention, because they now have the relationships to sell under the enterprise, they're gonna continue to be able to do their own innovation but also fast follow when IT makes sense to, uh, redefine GPU as something a little bit half year and incorporate other pieces of hardware to do other workloads into IT.

Yeah, I think the question for me on an a plus outcome for NVIDIA from this shareholder perspective is, do you need to believe that all the real world ai use cases are gonna a happen? Do you need to believe that some basket, maybe not all of them, but that some basket of anton's vehicles, the on universe robotics, one or multiple of those three, are gonna happen? They're going to be enormous markets, and then invidia is going to be a key player in them.

I mean, I think you do because I think that's where all the data set of revenue was coming from. These companies that are going after those opportunities .

and restless with whether that is something you have to believe or whether that's optionality the reason that would be only optionality, only upside is if the digital AI. We know that, that's a big market. There is no question about that at this point.

Is that going going to continue to just get so big? Are we still always scratching the surface there? How much more A I is gonna be baked into all the stuff we do in the digital world.

And will in video, I continue to be at the center of that? I don't know. I don't have a great way to assess how much growth is left there.

And I just kind of the right question though, yeah, there are an interesting point right now.

Know there is all the early company stuff that we talked about in the first episode, but at the beginning of this episode, no, jenson was really asking you to believe so hey, we're building this cute thing. Just ignore that there is no real use case for IT our market now. There is a real, real use case of market for IT, which is seen learning, deep learning in the digital world. Under able, he also pitching now that that will exist in the physical world too.

Yeah, the a plus is definitely that IT does exist in the physical world, that they are the dominant provider of everything you need to be able to accomplish that. yep. And if the real world stuff, you know, these little robots that run around the factory floors and the autonomous vehicles and if that stuff doesn't materialized, then he others no way that I can support the growth that it's been on.

I think that's probably right. That would be my hunt. Oh, saying that though does feel like a little bit of a betting against the internet. You know like I don't know mean the digital world pretty big and and keeps getting bigger.

Yeah but I think we're saying the same thing. I think you're saying that these physical experiences will become more and more inter twined with your digital experiences.

yeah. I mean.

autonomous driving in electric vehicles is an internet bet. In part if you want to bet on the growth y internet on media drive less. But IT also means that you're just going to be on the internet when you're driving .

yeah yeah when you're in .

motion in the physical world.

That's actually that's a bull case for a facebook right is like is autonomous vehicles because if people are being driven instead of driving, that's more time they're on instagram. It's so true.

okay. What's the failure case? It's actually quite hard to imagine a failure case of the business in any short order. It's very easy to imagine a failure your case for the stock in short order if there's a cast cate set of events of people losing faith.

I think maybe the failure case is this amazing growth for the past couple of years was pandemic pull forward. It's so hard for you imagine that that's like to the degree of a paleo or as you or something like that. But the way I think a great company that is get everything pulled forward, I don't think in video get everything pulled forward. They probably got a decent amount pulled forward.

Hard to quantify, hard to know, but it's the right .

thing to be thinking about.

Yeah right. carrots.

H car outs. I get a fun one, small one. Well, he collection of small things, long time listening is probably no one of my favorite.

I think my favorite series of books that have been written in the past ten years says the expense series, amazing. Five, five, nine books, so great, the ninety book came far. He was just even with, like, a newborn.

I made time to read this book, newborn plus acquired dies, like I gotta. That's that, you know, recently, last month. So the authors have been reading short stories, like companion short stories, along side the main narrative over the last decade that y've been doing this. And they released a companion of all the short stories, plus a few new ones called memories legion. And this is really cool. Like, I mean, their great writer er is a great short stories to read, even if you don't know anything about the expanse story, but if you know the whole nine books and then these like just paint little to be little glimpses in the corners and like characters that just existing, you don't quest IT otherwise but like, oh, what's the back story of that? I'm been really enjoying that.

So it's like the solo of the fantastic piece in where to find them exactly.

It's like nine or ten of .

those cool man is a physical product actually for the episode we did with brad garner. Er an ultimate we needed a third camera and so I was out bought uh SONY R X one hundred or point shoot camera and uh recently took IT to disney land and I must say IT is so nice to have a point shoot camera again. It's like funny how it's gone full circle.

I was A D our person forever and then I ve got a miracle camera. Then I became mr less plus big lang zoom lens person. But it's kind of annoying to look that around.

And then once I started downgrading my phone from the massive awesome iphone with the three extreme, and I now have the iphone thirteen mini, I think that's what IT is with the two cameras and no zoom lens really disapointment. It's pretty awesome. IT fills a sort of spot in my camera line up to have a pointing shoot with a really long zoom lands on IT.

And of course, like it's not as nice as having a full frame mirrors with like an actual zom ones, but IT really gets the job done. And it's nice to have that sort of like real feeling miracle style image that is very clearly from a real camera and not from a phone. A is a it's slightly more inconvenient to Carry because you can need another pocket.

Yeah going to ask, can you put IT in your pocket?

Yeah, I put in a pocket. I don't have to have a so like a rapid strap around my neck, which is nice, nice. So the SONY x one hundred great little device is like the seventh generation of IT, and they've really refined the industrial design at this point.

Awesome, awesome. I actually thought my first camera cube like a travel camera cube thing for our uh alf seventies now that we have a four required for, went after the ultimate epsom. I was like, oh.

wow, to do more in person.

Yeah, yeah. Then brought down is like, for sure. I'm gonna need to bring this somewhere. These cameras are just, they're so good. They're so good.

Our listeners, thank you so much for listening. You should come chat about the episode with us in the slack. There is eleven thousand other smart members of the acquired community just like you.

And if you want more acquired content after this and you were all cut up, go check out our L, P, show by searching acquired L P show in any podcast player. Here are interview nick. Learn from trovo trip most recently, and we have a job board.

Acquire that of m flash jobs. Find your dream job created just by us, the fine folks at the acquired podcast. And we will see you next time.

next time. Easy you, easy you, busy you who got the true.