NVIDIA: OpenAI, Future of Compute, and the American Dream | BG2 w/ Bill Gurley and Brad Gerstner

00:00:00I think that OpenAI is likely going tobe the next multi- trillion dollarhypers scale company. Okay,[Music][Applause]>> Jensen. Great to be back, of course,with my partner Clark Tang. You know, Ican't believe it's>> Welcome to Invidia. Uh, oh, and niceglasses.>> Um, those actually look really good onyou. The problem is now everybody'sgoing to want you to wear them all thetime. They're going to say, "Where arethe red glasses?">> I can vouch for that. So, it's been overa year since we did the last pod.>> Yeah.>> Over 40% of your revenue today isinference, but inference is about readybecause of chain of reasoning. Yeah.>> Right. It's about>> It's about to go up by a billion times,>> right? By by a million x by a billion.>> That's right. That's the part that mostpeople have, you know, haven'tcompletely internalized. This is thatindustry we were talking about. This isthe industrial revolution.>> Honestly, it's it's felt like you and Ihave had a continuation of the pod everyday since then. You know, in AI time,it's been about a hundred years. I wasre-watching the pod recently and themany things that we talked about thatstood out. The most the one that thatwas probably most profound for me

00:01:15wasyou pounding the table>> that you know remember at the time therewas kind of a slump in terms ofpre-training>> and people were like, "Oh my god,>> the end of pre-training,>> right? The end of pre-training. We'renot going. We're overbuilding. This isabout a year and a half ago.>> And you said inference isn't going to100x, a thousandx, it's going to 1billionx.>> Mhm.>> Which brings us to where we are today.You know, you announced this huge deal.We ought to start there.>> I underestimated.Let me just go on record. I estimated wewe now have three scaling laws, right?We have pre-training scaling law. Wehave post-training scaling law.Post-training is basically like uh AIpracticing. Yes,>> practicing a skill until it gets itright. And so it tries a whole bunch ofdifferent ways and and uh in order to dothat, yeah, you've got to do inference.So now training and inference are nowintegrated in reinforcement learning.>> Really complicated. And so that's calledpost training. And then the third>> is inference. The old way of doinginference was one shot,>> right?>> But the new way of doing inference,which we appreciate, is thinking. Sothink before you answer.>> Yeah. And so now you have three scalinglaws. The the longer you think, thebetter the quality answer you get. While

00:02:30you're thinking, you do research, you gocheck on some ground truth. And you youlearn some things, you think some more,you go learn some more, and then yougenerate an answer. Don't just generateright off the bat.>> And so thinking, post-training,pre-training, we now have three scalinglaws, not one. You knew that last year,but is your level of confidence thisyear in the inferences going to 1billionx and where that will take thelevels of intelligence is it higher? Areyou more confident this year than youwere a year ago?>> I'm more confident this year and thereason for that is because look at theagent systems now>> and AI is no longer a language model andAI is a system of language models andthey're all running concurrently maybeusing tools. Some of us using tools,some of us doing research and yeah,there's a whole bunch of stuff and it'sall multimodality and look at all thevideo that's being generated. I mean,it's just crazy stuff. Yeah.>> It really brings us to, you know, kindof the seminal moment this week thateverybody's talking about the massivedeal. You announced a couple days agowith OpenAI Stargate where that you'regoing to be a preferred partner, investhundred billion dollars in the companyover a period of time. they're going tobuild 10 gigs and if they used Nvidiafor those 10 gigs that could be upwardsof 400 billion in revenue to Nvidia.

00:03:45Sohelp us understand you just tell us alittle bit about that partnership whatit means to you right and why thatinvestment makes so much sense forNvidia.>> So first of all that I'll answer thatlast question first and then I'll comeback and present my way through. I thinkthat OpenAI is likely going to be thenext multi- trillion dollarhypers scale company.>> Okay.>> I think you and I>> Why do you call it a hypers scalecompany?>> Hypers scale like uh like Meta is ahypers scale. Uh Google's a hypersscale.>> They're going to have consumer andenterprise services and and uh they arevery likely going to be the world's nextmulti-t trillion dollar hypers scalecompany. Yes.>> And I think you would agree with that.>> I agree.>> If that's the case, the opportunity toinvestbefore they get there, this is some ofthe smartest investments we can possiblyimagine. And you got to invest inthings, you know,>> right?>> And it turns out we happen to know thisspace.>> And so the opportunity to invest inthat,>> the return on that money is going to befantastic. So we love the opportunity toinvest. We don't have to invest, right?>> And it's not required for us to invest,but they're giving us the opportunity toinvest. Fantastic thing. Now let mestart from the beginning. So we'repartnering with with OpenAI in several

00:05:01projects. First the first project is thebuildout of Microsoft Azure. We're goingto continue to do that>> and that that partnership is goingfantastically. We're going we haveseveral years of buildout to do hundredsof billions of dollars of work just todo there.>> Right.>> The second is the OCI buildout>> and uh I think there's some five, six,seven gigawatts that are about to bebuilt out. And so we're working with OCIand OpenAI and and SoftBank to buildthat out,>> right?>> Those projects are contracted. Um we'reworking on it. Lots of work to do. Andthen the third the third is Core Weave,right? And so uh all of Core Wee 4, I'mtalking about OpenAI still. Yes.>> Okay. Everything in the context ofOpenAI.>> And so so the question is what is thisnew partnership? This this newpartnership is about helping OpenAIworking partnering with OpenAI to buildtheir own selfbuildAI infrastructure for the first time,>> right?>> And so this is us working directly withOpenAI at the chip level, at thesoftware level, at the systems level, atthe AI factory level um to help thembecome a fully operated hypers scalecompany that I mean this is going to goon for some time. And it's going tosupplement it's going to supplement theamount of

00:06:16you know they're going throughtwo exponentials as you know>> right>> the first exponential is the number ofcustomers is growing exponentially andthe reason for that is the AI is gettingbetter the use case is getting betterjust about every every application isconnected to open now>> and so they're going through the theusage exponential the second exponentialis the computational exponential ofevery use.>> Yes.>> Right. Yes.>> Instead of just a oneshot inference isnow thinking before it answers. Yeah.>> And so these two exponential iscompounding their compute requirements.And so we we got to build out these allthese different projects. And so thislast one is an additive on top ofeverything that they've alreadyannounced, all the things that we'realready working on with them. It'sadditive on top of that>> and uh it's going to support the youknow this incredible exponential growth.>> One of the things you said there that'sreally interesting to me is kind of youknow they're going to be highprobability multi-t trillion dollarcompany in your mind. I think it's agreat investment. At the same time, youknow, they're selfbuilding. You'rehelping them self-build their datacenters. So, here to four, they've beenoutsourcing to Microsoft to build thedata center. Now, they want to buildfull stack factories themselves.>> They want to do they want to they wantto basically have a relationship with usthe way that that Elon and X hasrelation.>> Correct.>> I mean, Elon and Exel built.

00:07:31>> Exactly. But I think that's>> this is a very big deal>> when you think when you think about>> the advantage that Colossus had. Yeah,>> they're building full stack. That is ahyperscaler because if they don't usethe capacity, they could sell it tosomebody else.>> In the same way Stargate, they'rebuilding monstrous capacity. They thinkthey'll need to use most of it, but itputs them in a position to sell it tosomebody else as well. It sounds verymuch like AWS or GCP or Azure. That'swhat you're saying.>> Yeah, I I think they'll likely use itthemselves and um just think the case ofX, they'll likely use it themselves. Umbut they would like to have the the samedirect relationship with us, directworking relationship and directpurchasing relationship. Um uh Meta justas with Zuck and Meta has with us uhit's exactly a direct>> um our relationship with uh between usand Sunund and Google direct ourpartnership with Satia and Azure direct.Isn't that right? And so they've gottento a large enough scalethat they believe it's time for them tostart building these directrelationships. So, I'm delighted tosupport that and and all of and andSatia knows it and Larry knows it andeverybody everybody's aware of what'sgoing on and everybody's very supportiveof it.>> So, one of the things I find mysterious,right? You know, you just mentionedOracle 300 billion

00:08:47Colossus what they'rebuilding. We know what the sovereignsare building. We know what thehyperscalers are building. You know,Sam's talking in terms of trillions. Butof the 25 sellside analysts on WallStreet who cover your stock, if I lookat the consensus estimate, it basicallyhas your growth flatlining starting in2027.>> 8% growth 2027 through 2030. Okay, thatis the 25 people in their only job. Theyget paid to forecast the growth rate forNvidia. So clearly>> we're comfortable with that by the way.>> Right. Look, we're comfortable withthat. Okay, we have no trouble beatingthe numbers on a regular basis,>> right? No, I understand that. But, butthere is this interesting disconnect.>> No,>> right. I hear it every day on CNBC andBloomberg.>> And I think it goes to, you know, someof these questions around, you know, uh,shortages leading to a glut that theydon't believe. They say, "Okay, we'llgive you credit for 26, but 27, youknow, maybe we'll have too much andyou're not going to need that." But itis interesting to me and I think it'simportant to point out that the yourconsensus forecast is that this won'thappen right and we also put togetherforecast uh you know for the company

00:10:03taking into account all of these numbers>> and what it shows me is still eventhough we're two and a half years intothe age of AI a massive divergence ofbelief>> between what we hear Sam Alman sayingyou saying Sundar saying Satcha issaying and what Wall Street stillbelieves and you know again you'recomfortable with that.>> I also don't think it's inconsistent.>> Okay. So explain that a little bit.>> So first of all uh for the builderswe're supposed to be building foropportunity>> right?>> We're we're builders.>> Let me give you three points to thinkthrough and and and these three pointsuh it'll help you um hopefully uh bemore comfortable with Nvidia in thisfuture. So the first point and this isthe laws of physics point. This is themost important point that generalgeneral purpose computing is over andthe future is accelerated computing andAI computing.>> That's the first point.>> And so the way to think about that isthere's how much how many trillions ofdollars of computing infrastructures inthe world that has to be refreshed.>> Right. Right. And when it gets refreshedit's going to be accelerated comput.>> That's right. And so the first thing youhave to realize is that general purposecomputing and nobody disputes that.Everybody goes, "Yeah, we completelyagree with that. General purposecomputing

00:11:18is over. Moore's law is dead."People say these things. And so whatdoes that mean? So general purposecomputing is going to go to acceleratedcomputing. Our partnership with Intel isrecognizing that general purposecomputing needs to be fused withaccelerated computing to createopportunities for them.>> Is that right? And so one,>> general purpose computing is shifting toaccelerated computing and AI. Two, thefirst use case of AI is actually alreadyeverywhere,>> right?>> It's in search recommener engines,>> isn't that right? In shopping. The basichypers scale computing infrastructureused to be CPUs doing recommenders,right?>> Is now going to GPUs>> doing AI,>> right?>> So you just take classical computing,it's going to accelerated computing AI.You take hypers scale computing is goingfrom CPUs to accelerated computing andAI and then now that's the second pointjust feeding the metas the Google's thebite dances the Amazons>> and take their classical traditional wayof doing hyperscaling and moving into AI>> that's hundreds of billions of dollars>> and and because that may be four billionpeople on the planet today if you

00:12:33takeTik Tok meta into account>> that's Google into account who arealready demanding workloads that aredriven by accelerated comput.>> That's exactly right. And so there asimp without even thinking about AIcreating new opportunities. It's aboutAI shifting how you used to do somethingto the way new way of doing something.Okay. And then now let's talk about thefuture. I just so far I've only spokenkind of largely about>> just mundane stuff.>> Just mundane stuff. The old way is nowwrong. You're going to go, you're nolonger going to use uh uh fuel lightlanterns. You're going to go toelectricity. That's all.>> Right.>> Okay. And you no longer, you know, propplanes. You're going to go to jets.That's all. And so, you know, so far,you know, that's all I've talked about.And then>> now that the incredible thing is whenyou go to AI, when you go to acceleratedcomputing, then what happens?>> What are the new applications thatemerge as a result? And that's all theAI stuff that we're talking about.And that's the that opportunity. What isit? How do what does that look like?Well, the simple way of thinking aboutthat is where motors replace labor andphysical activity. We now have AI. TheseAI supercomputers, these AI factoriesthat I talk about,

00:13:48they're going togenerate tokens to augment humanintelligence, right? And humanintelligence represents what 55 65% ofthe world's GDP. Let's call it $50trillion.And that $50 trillion is going to getaugmented by something. And so let's youjust let's come back to a single person.Suppose I were to hire a $100,000employee. And I augmented that $100,000employee with a $10,000 AI. Yes.>> And that $10,000 AI as a result madethat $100,000 employee twice moreproductive, three times more productive.Would I do it? Heartbeat. I I'm doing itacross every single person in ourcompany right now. Right.>> Every single co-agents.>> That's right. Every That's right. Everysingle software engineer, every singlechip designer in our company already hasAIS working with them.>> 100% coverage.>> As a result,>> the number of chips we're building isbetter. The number is growing. The paceat which we're doing it is right. And sowe're we're growing faster as a company.As a result, we're hiring more people.Our productivity is greater. Our topline's greater. Our profitability isgreater. What's not to love about that?Now apply the Nvidia story to theworld's GDP.>> Yeah.>> And so what's likely to happen is thatthat $50 trillion is augmented by

00:15:04let'spick a number>> 10 trillion that $10 trillion>> needs to run on a machine. M>> now the reason that AI is different thanit in the past>> in a way software was written a priori>> and then it runs on a CPU and it doesn'tit runs it a a person would operate it>> in the future of course AI is generatingtokens>> but a machine has to generate the tokensand it's thinking>> so that software is running all the timewhereas in the past the software waswritten once now the software is in factwriting all the time it's thinking>> in order for the AI to think it needs afactory. And so let's say that that 10trillion of token generated>> 50% gross margins and 5 trillion of itneeds a factory needs an AIinfrastructure. So if you told me thaton an annual basis the capex of theworld was about $5 trillion>> I would say the math seems to makesense.>> Yeah.>> And that's kind of the future, right?Yeah. the going from Excel generalpurpose computing to acceleratedcomputing replacing all the hypersscales with AI and then now augmentinghuman intelligence for the world's GDP>> and today that market is about ourestimate is

00:16:19about 400 billion annually.>> Yeah.>> So the the TAM you know is is a four to5x increase over where it is today.>> Yeah. Eddie last night, Eddie Woo atAlibaba said between now and the end ofthe year and excuse me, now and the endof the decade, they're going to increasetheir data center power by 10x. Right.>> Right. You just said how much?>> 4x.>> There you go.>> There you go.>> Yeah.>> They're going to increase power by 10x.And we we correlate to power.>> Nvidia's revenue is almost correlated topower. Isn't that right?>> Yeah,>> that's right. Yeah. Because>> one other thing, what he what else didhe say?>> Yeah.>> He said token generation is doublingevery few months.>> Yeah.>> What's that saying? The the perf perwatt has to keep on going exponentially.That's why Nvidia's like cranking it outwith perf per watt.>> And revenue per watt is, you know, wattis basically revenues in this future.>> Embedded in this assumption, I find itvery fascinating historical context,right? For 2,000 years, basically GDPdid not grow. Okay? And then we get theindustrial revolution, GDP accelerates.We get the digital revolution, GDPaccelerates.

00:17:34And it basically whatyou're saying, and Scott Besson has saidit, he said, I think we're going to have4% GDP growth next year. Basically, whatyou're saying is the world's GDP growthis going to accelerate because now weare giving the world billions ofco-workers that will do work for us. Andif GDP is an amount of output for afixed amount of labor and capital,right, it has to accelerate.>> It has to,>> right?>> It has to look at what's going on withAI as a result of the technology of AI.And that technology of AI, let's justcall it the large language models andall the AI agents. It's now creating anew industry of AI agents. There's noquestion about that. Okay. So, so that'sOpenAI is the fastest growing revenuecompany in history,>> right? and they're growingexponentially, right? And so, so AIitself is a fast growing industrybecause of AI needs a factory behind it,right?>> An infrastructure behind it.>> There's this industry is growing. Myindustry is growing.>> And because my industry is growing, theindustry underneath it is growing.Energy is growing.>> Power shell.>> This is the>> This is like renaissance for the energyindustry, isn't that right? Nuclearenergy, you know, gas turbines. I mean,look at all of those companies in the inthe infrastructure

00:18:50ecosystem underneathus. They're doing incredibly well.>> Everybody's growing. These numbers haveeverybody talking about a glutterbubble, right? Zuckerberg said last weekon a podcast, you know, he said,"Listen, I think it's quite possible atsome point that we will have an airpocket and Meta may in fact overspend by$10 billion or whatever, but he said itdoesn't matter. It's so existential tothe future of his business that it's arisk that they have to take. But whenyou think about that, it sounds a littlebit like prisoners dilemma. Right. Andwalk us again through>> these are very happy prisoners.>> Walk us again through Right. Today ourestimate is that we're going to have a100 billion of AI revenue in 2026excluding meta and excluding you knowthe GPUs running recommener engines.Okay. So there's>> or search or>> correct. So there there's other stuffbut let's call it 100 billion.>> What is that industry anyways?>> What is the industry already in hypersscale? What is the hypers scales youknow between trillions?>> Yeah exactly by the way that industry isgoing to AI>> before anybody starts at zero. You gotto start there.>> But I think the skeptics would say weneed to go from a 100red billion of AI

00:20:05revenue in 26 to at least a trillion ofAI revenue in 2030. Okay. You just weretalking a minute ago about five trillionwhen you look at kind of global GDP. Ifyou do did a bottoms up, can you seeyour way to a trillion dollars ofAIdriven revenues from a hundred billionover the course of the next 5 years? Arewe growing that fast? Yes. And I wouldalso say we're already there.>> Okay. So explain that.>> Because the hyperscalers,they went from CPUs to AI. Okay. Theirentire revenue base is all now AIdriven.>> Correct.>> Uh you can't do Tik Tok without AI.>> Correct.>> You can't do YouTube short without AI.You can't you know you can't do any ofthis stuff without AI. Uh the theamazing things that that Meta is doingfor for uh um uh you know customizedcontent, personalized content. You can'tdo that without AI. It's all of thatstuff used to be humans, you know, doinguh content a priori creating fourchoices that are then>> selected by a recommener engine.Correct. And now it's infinite number ofchoices generated by an

00:21:20AI, right?>> But those things are already like we hadthe transition from CPUs to GPUs largelyfor those recommener engines and nowthey're going>> and that's fairly new I would>> in the last three or four years. Zuckwould tell you I was at Sigraph and Zuckwould tell you you know they were lategetting to GPU for sure for sure.>> GPUs for for meta is what couple yearsand a half>> it's pretty new search with GPUs>> for sure>> brand spanking new>> for sure for sure>> brand spanking new>> search for GPUs on GPUs>> so your argument would be theprobability that we're going to have atrillion dollars of AI revenues by 2030is near certain because we're almostalready there. Okay,>> let's just talk about incremental fromwhere we are.>> Now we can talk about incremental fromwhere we are today, right? As you doyour bottoms up or your tops down, Ijust heard your tops down aboutpercentage of global GDP.>> Yeah.>> What is the percentage probability thatyou think we'll have a glut will runinto a glut in the next three or four orfive years?>> Right. It's a distribution of we don'tknow the future. It's a distribution ofpower>> until until wefullyconvert all

00:22:35general purpose computing to acceleratedcomputing and AI. Until we do that,>> yes,>> I think the chances are extremely low.>> Okay. Okay. And that will take a fewyears.>> That'll take a few years.>> Yeah.>> Yeah. Let me ask one more and then>> until all recommener engines are AIbased. until all content generation isAI based because content generationconsumer oriented content generation isvery largely recommener systems and soon so forth um and all of that's goingto be AI generated until until all ofthe stuff what classically was hypersscale now transitions to AI you knoweverything from shopping to e-commerceto you know all that stuff untileverything goes over>> because but all this new build rightwhen we're talking about trillions.We're investing ahead of where we are.Um, you know, is that like at will? Areyou obliged to invest the money even ifyou see a slowdown or a kind of a glutcoming or is this one of these thingsthat you're just waving the flag to theecosystem to say get out and build andat some point in time if we see some ofthis slow down, we can always pull backon the level of investment.>> Actually, it's the other way becausewe're at the end of the supply chain,right? And so, we respond to demand.>> Okay?

00:23:50And right now all the VCs willtell you and you guys know>> the demand the short there's a shortageof compute in the world not becausethere's a shortage of GPUs in the world.>> Okay, if they give me an order I'llbuild it.>> Mhm.>> Right. We've over the last couple yearswe've really plumbed the supply chain.So all of the supply chain behind mefrom wafer starts to co-ass HBM memoriesyou know all of that technology. We'vereally geared up.>> Yeah.>> If we need to double we'll double.>> Yes.>> Okay. So the supply chain is ready. Nowwe're just waiting for demand signalsand when when uh the the CSPs and thehyperscalers and our customers uh dotheir annual plan and they give us youknow their forecast um we respond tothat and we build to that.>> Now what's what's going on of course isthat every one of their forecasts thatthey provide us turns out to have beenwrong>> right>> because they under forecasted>> and so now we're always in a scramblemode. Mhm.>> And so we've been in the scramble modenow for you know a couple of years>> and it's whatever forecast we've beengiven has been always>> significant increase from last year butnot not enough.>> Satcha last year seemed to be pullingback a little bit. You know seemed to beyou know some people called him theadult in the room tamping down kind ofsome of these these

00:25:06expectations. A fewweeks ago he said hey I've also builttwo gigs this year and we're going toaccelerate in the future. Do you seesome of the traditional hyperscalersthat may have been moving a littleslower than let's call it a coreweave oror or Elon X or maybe a little slowerthan Stargate? Do you see them all? Itit sounds like to me they're all leaningin more now and they're all also>> because of the second exponential.>> Okay.>> We've already had one exponential wewere experiencing which was the adoptionrate of AI, the engagement of AI wasgrowing exponentially. Yes. The secondexponential that kicked in wasreasoning.>> Yeah.>> That was the conversation we had oneyear ago.>> One year ago.>> Yeah. We said, "Hey, listen. The momentyou take AI from one shot,>> memorizing an answer and Right.Memorizing and generalizing, that'sbasically pre-training.">> Yeah.>> So memorizing an answer, you know what's8* 8? Just memorize it. Okay. And somemorizing an answer and generalizing,that was one shot AI. Now a year ago,reasoning came about>> for sure>> research came about, tool use cameabout, and now you're a thinking AI>> 1 billion X.>> It's going to use a lot more compute.>> Certain hypers scale customers to yourpoint

00:26:21had internal workloads that theyhad to migrate anyways from um fromgeneral purpose computing to acceleratedcomputing. So they built through thecycle. I think maybe some hyperscalershad different workloads so they weren'tquite sure how quickly they could digestit but everyone>> has now concluded that they dramaticallyunderbuilt.>> One of the applications that my favoriteis just good oldfashioned dataprocessing>> structured data and unstructured data.>> Just good oldfashioned data processing.>> And very soon we're going to announce avery big initiative of accelerated dataprocessing. M>> data processing represents the vastmajority of the world's CPUs today.>> It still completely runs on CPUs.>> You know, if you go to data bricks, it'smostly CPUs. You go to snowflakes,mostly CPUs. Uh SQL processing atOracle, mostly CPUs. Everybody's usingCPUs to do SQL structured data.>> In the future, that's all going to moveto AI data.>> That is one gigantic massive market>> that we're going to move to. But youneed you need a you need everything thatNvidia does requires acceleration layersand requires you know>> domain specific right data processinglies recipes we got to go build that butthat's coming

00:27:37>> so one of the push backs I you know Iturned on CNBC yesterday they were likeoh glut bubble when I turned onBloomberg it was about roundtpping andcircular revenues okay and so for thebenefit of people you know at home knowthese arrangements are when companiesenter into a misleading transactionction that artificially inflates revenuewithout any underlying economicsubstance. So in other words, growthpropped up by financial engineering, notby customer demand. And the canonicalcase everybody's referencing of courseis Cisco and Nortell from the lastbubble 25 years ago. So when you guys orMicrosoft or Amazon are investing incompanies that are also your bigcustomers, in this case you guysinvesting in Open AI while Open AI isbuying tens of billions of chips, justremind us and remind everybody else likewhat is it what are the analysts onBloomberg and otherwise getting wrongwhen they're hyperventilating aboutcircular revenues or about roundtpping?>> 10 gigawatts is like $400 billion,right? Something like that. and and that$400 billion dollars will have to belargely funded by their offtake,>> right? Their revenue>> which is growing exponentially.It has to be funded by their capital,

00:28:52the money they've raised through equityand whatever debt they can raise.>> Those are the three vehicles. And theequity that they could raise and thedebt that they could raise has somethingto do with the confidence of therevenues that they could sustain>> for sure. And so smart investors andsmart lenders will consider all of thesefactors.>> Fundamentally, that's what they're goingto do.>> That's their company. It's not mybusiness. And of course, we have to stayvery close to them to make sure that webuild in support of their continuedgrowth.>> Okay? And so, um, there's the revenueside of it and has nothing to do withthe investment side of it. Theinvestment side of it is not tied toanything. It's an opportunity to investin them. And as we were mentioningearlier, this is likely going to be thenext multi- trillion dollar hypers scalecompany. And who doesn't want to be aninvestor in that? You know, my onlyregret is that that they invited us toinvest early on.>> I remember those conversations>> and we were so poor, you know, that wewere so poor, we didn't invest enough,you know, and I should have given themall my money. And the reality is if youguys don't do your jobs and keep upwith, you know, if if Ver Rubin doesn'tturn into a good chip, they can go getother chips and put them in these datacenters,>> right? There's no obligation that theyhave to use your chips and and like yousaid, you're you're

00:30:07looking at this asan opportunistic equity investment. Theother thing I would say>> and we've made some great investments. Igot to put it out there, you know, weinvested in XAI, we invested inCorewave.>> Incredible. Yeah.>> Yeah. How smart was that?>> Yeah. As I go back to this, the otherfundamental thing it seems to me is, youknow, you're putting it out there.You're saying this is what we're doing.And the underlying economic substancehere, right? It's not that you're justsome somehow sending revenues back andforth between the two companies. We gotpeople sending money every month forChat GPT, a billion and a half monthlyusers using the product. You just saidevery enterprise in the world is eithergoing to do this or they will die. Everysovereign views this as existentialto their national security and economicsecurity as as nuclear power. Whatperson, company or nation saysintelligence is bas basically optionalfor yeah for us. I mean it's fundamentalto them.>> Well the automation of intelligence>> I beat the demand question to death. Solet's jump in a little bit to systemdesign and I'm I'm I'm going to turn toClark here in a sec second on that. Butin 2024, you switched to your annualrelease cycle,>> right, with Hopper.

00:31:23>> You then had a massive upgrade whichrequired, you know, significant datacenter overhaul with Grace Blackwell. In2025, and in the back half of 26, we'regoing to get Vera Rubin. 27 we'll getUltra and 28 Fineman. How is the annualrelease cycle going? Okay. What were themain goals of going to an annual releasecycle? And did AI inside Nvidia allowyou to execute the annual release cycle?>> Yeah, the answer is yes. On the back onthe last question, without it, um,Nvidia's velocity, our pace, our scalewould be limited.>> And so without without AI these days,um, it's just simply not possible tobuild what we built. Now, um, why do wedo it? There's something that rememberuh uh Eddie said it uh at his earningscall um or his conference. Uh Satia hassaid it, Sam has said it. The tokengeneration rate is going upexponentially.>> Yeah.>> And the customer use is going upexponentially.>> Um I think they're at 800 million weeklyactive users or something like that.>> Yes. I mean that's less than two yearsfrom chat GPT, right?>> And each of those users

00:32:38is generatingmassively more tokens because they'reusing inference time reasoning.>> That's right. Exactly. And so so thefirst thing is because the tokengeneration rate is going up soincredibly two exponentials. Yeah.>> On top of each other,>> we have to un unless we increase theperformanceat incredible rates,>> the cost of token generation will keepgrowing because Mo's law is dead, right?Because transistors basically cost thesame every single year now. And power islargely the same. And between those twofundamental laws, unless we come up withnew technologies to drive the cost down,even if there's a slight difference ingrowth, you give somebody a discount ofa few percent. How's that going to makeup for two exponentials?And so we have to increase our perperformance annually at a pace thatkeeps up with that exponential.>> Yeah. So in the case of in the case ofum uh going from uh I guess uh uh Keplerto yeah all the way to Kepler all theway to um uh Hopper was probably100,000x>> that was the beginning of the AI journeyfor Nvidia. 100,000x in 10 years. Okay.Between Hopper and Blackwell, weincreased because of

00:33:53MVLink 72,>> right?>> 30x>> in one year and then we'll get another Xfactor again with Reuben and then we'llget another X factor with Fineman. Andthe way we we do that is because thetransistors aren't really helping usvery much, right? Moors law is largelythe density is growing up but going upbut the performance is not. And so ifthat's the case, one of the challengesthat we have to do is we have to breakthe entire problem down at the systemlevel and change every chip at the sametime and all the software stack and allthe systems all at the same time. Theultimate extreme code design. Nobody'sever codees at this level before,>> right?>> We change the CPU, revolutionize theCPU, a GPU, the networking chip, theMVLink scale up, the Spectrum X scaleout. Somebody said I heard somebodysaid, "Oh yeah, it's just Ethernet."Yeah. Right. Okay. So, Spectrum XEthernet is not just Ethernet. Andpeople are starting to discover, oh mygod, the X factors is pretty incredible,>> right?>> You know, Nvidia's Ethernet business,the just Ethernet business is thefastest growing Ethernet business in theworld.>> Yeah.>> And so, so scale out and of course nowwe have to uh build even larger systems.So, we scale across>> um multiple AI factories connectedtogether. And then we

00:35:08do this at anannual pace. And so we now have anexponential of exponentials goingourselves from technology. And thatallows our customers to drive the costof tokens down. Keep making those tokenssmarter and smarter with pre-trainingand post-training and thinking. And as aresult, when the AI gets smarter, theyget more used. When they get more used,they're going to grow exponentially.>> For people who may not be as familiar.>> Yeah. What is extreme code design?>> Extreme code design means that you haveto optimize the model, algorithm,system, and chip at the same time.>> You have to innovate outside the box,>> right?>> Because Moore's law said you just haveto keep making the CPU faster andfaster. Everything got faster. You wereinnovating within a box.>> Just make that chip faster. Yeah. Well,if that chip doesn't go any faster, thenwhat are you going to do? Innovateoutside the box. And so Nvidia reallychanged things because we did twothings. We invented CUDA, invented GPUs,>> and we invented the idea of code designat a very large scale.>> That's why there's all these industrieswe're in. We're creating all theselibraries and code design. Number one,full stack extreme

00:36:24is even beyondsoftware and GPUs. It's now at the datacenter level switches and networking andyou know all of that all of thatsoftware in the switches and thenetworking and the nicks the scale upthe scale out>> optimizing across all of that as aresult of that blackwell to hopper is30x>> no moors law could possibly achieve that>> right>> and so that's extreme>> and that comes from the extreme codedesign>> that's because Nvidia has that's why wegot into networking and switching andscale up and scale out and scale acrossand building CPUs and building GPUs andbuilding nicks. You know that that's thereason why Nvidia is so rich in softwareand people we we check in moreopen-source software in the world thanjust about anybody except one othercompany. I think it's AI2 or somethinglike that.>> And so so we have such enormous richnessof software and that's just in AI. Don'tforget computer graphics and digitalbiology and autonomous vehicles and youknow the amount of software we produceas a company is incredible that allowsus to do deep and extreme code designs.>> I heard from one of your competitors youknow yes he's doing this because ithelps drive down the cost of

00:37:39tokengeneration but at the same time yourannual release cycle makes it almostimpossible for your competitors to keepup. um the supply chain gets locked upmore because you're giving three-yearvisibility to your supply chain. So nowthe supply chain has confidence as towhat they can build to. So do you thinkabout this? Wait wait wait before youbefore you ask the question. Think aboutthis.>> In order for us to do several hundredbillion dollars a year>> of AI infrastructure buildout.>> Yes.>> Think about how much capacity we had togo start a year ago.>> Yes. We're talking about buildinghundreds of billions of dollars of>> wafer starts and DRAM buys and are youyou guys talking>> Yeah.>> This is now at a scalethat hardly any company can keep upwith.>> So would you say your competitive moatis greater today than it was three yearsago?>> Yeah. You know, first of all, there'sjust more competition than ever before,but it's harder than ever before. Mhm.>> And the reason why I say that is becauseum wafer cost is getting wafer costs aregetting higher>> which means that unless you do codedesign at an extreme scale you're justnot going to be able to deliver the Xfactor growth

00:38:55>> number one number and so you you knowunless you unless you're working on sixseven eight chips a year>> right>> that's amazing thing it's not aboutbuilding an ASIC it's about building anAI factory>> system>> and this system that has a lot of chipsinand they're all co-designed>> and together they deliver that you knowthat 10x factor that we get>> uh almost regularly. Okay. So number onethe code design is extreme. The secondthing is that the scale is extreme.>> When your customers deploy a gigawattthat's a you know 400,000 500,000 GPUsright>> getting 500,000 GPUs to work together isa miracle.>> I mean it's just a miracle. And so theyyour customers are taking enormous riskon you to go buy all of this. You got toask yourself what customer would place a$50 billion PO>> on an architecture,>> right?>> On an unproven architecture, a new one,>> right? A new architecture.>> Yeah. You just take out a whole newchip. You're as excited as you are aboutit, you know, and everybody's excitedfor you>> and and you just show the first silicon,>> right? Who's going to give you $50billion PO,

00:40:10right?>> And why would you start $50 billionworth of wafers for a chip that justtaped out?>> But for Nvidia, we could do that becauseour architecture is so proven. So numberthe the scale of our customer is soincredible. Now the scale of our supplychain is incredible,>> right?>> Who's going to start all of that stuff,pre-build all of that stuff for acompany unless they know that Nvidia candeliver through? Isn't that right? andthey believe that we can we can deliverthrough to all of the customers aroundthe world. They're willing to startseveral hundred billion dollars at atime.>> This is just the scale is incredible.>> To that point, you know, one of thebiggest key debates and controversies inthe world is this question of GPUsversus A6,>> Google's TPUs, Amazon's Tranium, and itseems like everyone from ARM to OpenAIto Enthropic are, you know, rumored tobe building one. Mhm.>> Last year you said, you know, we'rebuilding systems, not chips, and you'redriving performance through every singlepart of that stack. You also said thatmany of these projects may never get toproduction scale.>> But given like the>> most of them,>> most of them,>> given the seeming success of Google'sTPUs,>> yeah,>> you know,

00:41:25how are you thinking aboutthis evolving landscape today?>> Yeah.>> And>> first of all,>> Mhm. the the advantage that that Googlehad is foresight.>> Remember, they started TPU1before everything started.>> You know, this is no different than astartup. You're supposed to build astartup. You're supposed to create astartup before the market grows.>> Mhm.>> You're not supposed to come up as astartup when the market's a trilliondollars large. Mhm.>> You know the this fallacy and and allVCs know this this fallacy that a largemarket if you could just take a fewpercent market share you could be agiant company.>> That's actually fundamentally wrong.>> Yeah.>> You're supposed to take 100% of a tinycompany, a tiny industry, which is whatNvidia did did, right? Which is whatTPUs did. There were only the two of us.>> But you better hope that that industrygets really big. You're creating anindustry.>> That's right. Right. and and I mean theNvidia story you know which>> and so that's the challenge for the forpeople who are building A6 now it lookslike a juicy market>> but remember this juicy market hasevolved from a chip called a GPU to Ijust described an AI factory and youguys just saw I just announced a chipcalled CPX for

00:42:40context processing anddiffusion video generation a veryspecialized workload but an importantworkload inside a data center I justprelude it to maybe AI data processingprocessors because guess what? You needlong-term memory. You need short-termmemory. The KV cache processing isreally intense.>> AI memory is a big deal.>> You know, you kind of like your AI tohave good memory and just dealing withall the KV caching around the system.Really complicated stuff. Maybe it wantsto have a specialized processor.>> Maybe there's other things, right? Soyou you see that Nvidia's our our ourviewpoint is now not GPU. Our viewpointis looking at the entire AIinfrastructure and what does it take forthese incredible companies to get all oftheir workload through it which isdiverse and changing.>> Look at the transformer. The transformerarchitecture is changing incredibly. Ifnot for the fact that CUDA is easy tooperate on and iterate on, how do theytry all of their vast number ofexperiments to decide which one of thetransformer versions, what kind ofattention algorithm to use? How do youdisagregate? CUDA helps you do all thatbecause it's so programmable. And so theway to think about our our our

00:43:55businessnow is you look at when when all ofthese ASIC companies or ASIC projectsstart 3, four, five years ago, I got totell you that industry was superadorable and simple.>> There was a GPU involved,>> right?>> But now it's giant and complex>> and in another two years it's going tobe completely massive. The scale isgoing to be so large. And so I thinkthat the battle the the battle ofgetting into>> a very large market as a nent player isjust hard you know as you guys know>> even for the customers who perhaps aresuccessful with AS6.>> Yeah.>> Isn't there an optimal balance in theircompute fleet like it's you know I thinkinvestors are very much binarycreatures. They just want a yes or noblack and white answer. But even even ifyou get the ASIC to work, isn't there anoptimal balance because you think I'mbuying the Nvidia platform, CPX is goingto come out for prefill for, you know,for video generation,>> maybe a decode, you know, you know, aplatform video.>> Exactly.>> Yeah.>> So there will be like many different>> um chips or parts to add to the Nvidiaecosystem,>> accelerated

00:45:11compute fleet,>> right? as new workloads are, you know,are are born.>> That's right.>> And, you know, people trying to tape outnew chips today are not reallyanticipating what's happening a yearfrom now. They're just trying to get achip to work.>> That's right.>> Set another way, Google's a big GPUcustomer.>> Google's a big GPU customer. Um, if youlook at and Google is a very specialcase. I mean, we just have to, you know,show respect where respect is reallydeserved. I mean,>> uh, TPU is on TPU7.>> Yes.>> Right. And so um and and uhand it's a challenge for them as well,right? And so so the work that they dois incredibly hard. So so I think thefirst thing to to let let me do a um youknow remember there there are threecategories of chips.There's the category chips that arearchitectural.>> X86 CPUs, ARM CPUs, Nvidia GPUs.ArchitecturalAnd it has an ecosystem above and and umuh the architecture uh allows uh hasrich IP and rich ecosystem verycomplicated technology. It's built bythe owners like us. Okay.

00:46:26>> There's AS6.I worked for the original company LSILogic who invented the idea of A6.>> As you know LSI Logic is not hereanymore,>> right? And the reason for that isbecause AS6 is really fantastic. Whenthe the market size is not very large,>> um it's easy to have somebody uh be acontractor to help you put the packagingof all that stuff together and do themanufacturing on your behalf and theycharge you 50 60 points of margin. Butwhen the market gets large for an ASIC,there's a new way of doing things calledcoot,>> customerowned tooling. And who would whowould do something like that? Um Apple'sas Apple Apple's smartphone chip thevolume is so large they would never gopay somebody else 50 60% gross margin tobe an ASIC. They do customer owntooling.>> And so um where will TPUs go when theywhen it becomes a large business?Customer own tooling.>> There's no question about it. And so abut there's a place for A6.>> Uh video transcoders will never be toolarge.Um, smart nicks will never be too large.And so when when there's 10, 12, 15 asprojects going on

00:47:41at an ASIC company,I'm not surprised by that,>> you know, because there there probablyfive smart nicks and four transcodersand, you know, are they all AI chips? Ofcourse not. You know, and and ifsomebody were to build an embeddedembedding processor for a specificrecommener system and that was an ASIC,of course you could do that. But wouldyou do that as the fundamental computeengine for AI that's changing all thetime? You've got low latency workload.You got high throughput lo workload. Youhave token generation for chat. You havethinking workload. You have AI videogeneration workload. Is there a you knownow you're talking about>> the workhorse backbone of youraccelerated?>> That's what Nvidia is all about. Again,dumb this down. It's like playing chessand checkers, right? The fact of thematter is the folks who are starting AS6today whether it's Tranium or whetherit's some of these other acceleratorsetc. they're building a chip that's acomponent of a much larger machine.you've built a very sophisticatedsystem, platform, factory, whatever youwant to call it, and now you're openingup a little bit, right? So, youmentioned CPXGPU, right? That is, itseems to me that it in some ways you'redisagregating the workloads to the bestslice

00:48:56of the hardware for thatparticular domain.>> Well, we did we announced this thingcalled Dynamo,>> right? disagregated orch AI workloadorchestration and we open sourced itbecause the future AI factory isdisagregated>> right>> and you launched MV Fusion>> that even said to your competitors>> including Intel which you just investedin>> that's right>> you know the way in which youparticipate in this factory that we'rebuilding because nobody else is crazyenough to try to build the entirefactory but you can plug into that ifyou have a product that's good enough,compelling enough that the end usersays, "Hey, we want to use this insteadof an ARM GPU or we want to use thisinstead of your inference accelerator,etc. Is that correct?">> We're delighted delighted to connect youin. Yeah.>> Tell us a little bit fusion such a greatidea and we're so happy to partner withIntel on that. It takes takes the Intelecosystem, you know, there most of theworld's enterprise still runs on Intel.It takes the Intel ecosystem, takes theNvidia AI ecosystem, acceleratedcomputing, and we fused it together,>> right?>> And we did that with ARM,>> right? And there are several otherswe're going to be doing it with. And anduh that that opens up opportunities forboth of us. It's a

00:50:11win for both of us.Great great win. I'll be a largecustomer of theirs and uh they're goingto expose us to a a much much largermarket opportunity.>> Yeah. that's deeply related to this ideais the argument you've made that kind ofum c shock some people where you say ourcompetitors building A6 they couldliterally all their chips are cheaperalready today but they could literallyprice them at zero our objective is theycould price them at zero and you wouldstill buy an Nvidia system because thetotal cost of operating that systempower data center land etc theintelligence out is still a better betthan buying a chip even if it's given toyou for free>> because the land power and shell isalready $15 billion,>> right?>> Yeah.>> So, we've taken a crack at kind of themath on that. But walk us through yourmath because I think for people whodon't spend as much time here that itjust doesn't compute. How could itpossibly be that you were pricing yourcompetitor's chips at zero given theexpense of your chips and it still is abetter bet? There's two ways to thinkabout it. Um, one way is um, uh, let'sjust think about it from a perspectiveof revenues.>> Yes.>> Okay. So everybody's power limited andlet's say uh you were able to secure twomore gigawatts of power.>> Well,

00:51:26that two gawatts of power youwould like to have translate torevenues.>> Yes.>> So yourperformance or tokens per wattwas twice as high as somebody else'stoken per watt because you did I diddeep and extreme code design,>> right?and my performance was much higher perunit energy,>> then my customer can produce twice asmuch revenues>> from their data center.>> And who doesn't want twice as muchrevenues? and and and if somebody gavethem a 15% discount,>> you know, the difference between ourgross margins, which is called the 75points, and somebody else's grossmargins, call it the 50 to 65 points, isnot so much as to make up for the 30times difference between Black Wall andHopper. Let's pretend Hopper Hopper isan amazing chip, an amazing system.Let's pretend somebody else's ASIC isHopper.>> Yeah.>> Black Wall's 30 times.So you've got to give up 30x revenues inthat one gigawatt.>> Mhm.>> It's too much to give up. So even ifthey gave it to you for free, you youyou only have 2 gigawatts to work with.Your

00:52:41opportunity cost is so insanelyhigh.>> You would always choose the best perfper watt. So I heard this from one ofthe CFOs at one of the hyperscalers thatgiven the performance improvement rightthat's coming out of your chips againprecisely to that point tokens per pergig um and power being the limitingfactor right that they had to upgrade uhto the new cycle so when you look aheadat Ruben at Ruben Ultra at Fineman>> does that trajectory continue>> we're building what six seven chips ayear now>> yeah and and each one>> that's part of that system.>> That's right. And those that systemsoftware is everywhere and it takes>> it takes the integration and theoptimization across all of those sixseven chips to deliver on the 30xblackwell. Now imagine I'm doing thisevery single year. Bam bam bam bam bambam. And so if you build one ASIC inthat soup of AS6 in that soup of chipsand we're optimizing across that, youknow, it's a hard problem to solve.>> This does bring me back to where westarted about the competitive moat.We've been covering this and investorsfor a while. We're investors throughoutthe ecosystem and in competitors ofyours,

00:53:57you know, from Google to toBroadcom. But when I really just firstprinciples around this and say>> are you increasing or decreasing yourcompetitive mode,you move to an annual cadence.You're co-developing with a with asupply chain.The scale is massively bigger thananybody anticipated which requires scaleboth of balance sheet and ofdevelopment.Right? the moves you made both throughacquisition and organically with thingslike Envy Fusion, CPX, which we justtalked about. All of those thingstogether cause me to believe that yourcompetitive mode is increasing visav atleast in so far as building out thefactory or the system.>> It's at least surprising.>> But but but I think it's interestingthat your multiple is much lower thanmost of those other people. And I thinkpart of that has to do with this law oflarge numbers. A $4.5 trillion companycouldn't possibly get any bigger. But Iasked you this a year and a half ago asyou sit here today. If the market'sgoing to AI workloads are going to 10xor 5x, you know, we know what capex isdoing, etc. Is there any conceivable

00:55:12world in your mind where your top linein 5 years isn't two or 3x bigger thanit is in 2025? Like what's theprobability that it's actually not notmuch higher than it is today given thoseadvantages?>> I'll answer it this way. Our opportunityas I described it is much larger thanthe consensus.>> I'll say it here. I think Nvidia willlikely be the first 10 trillion dollarcompany. And I would I've been here longenough, it wasn't that long ago, just adecade ago, as you well remember, thatpeople said there could never be atrillion dollar company. Now we have 10,right? And today the world's bigger,>> right? And today, this is this is theback to the exponentials around GDP andthe growth.>> The world is bigger and and and peoplemisunderstand what we do. They they uhthey remember we're a chip company,right? And we are we build chips. Boy,do we build chips and build the mostamazing chips in the world. But NVIDIAis really an AI infrastructure company.You know, we are your AI infrastructurepartner and our partnership with OpenAIis a perfect demonstration of that.>> Yeah.>> That we are their AI infrastructurepartner and we work with people in a lotof different ways. You know, you you youwould uh we we don't require anybody

00:56:27tobuy everything from us.>> Um we don't we don't require that theybuy uh the full rack. They could buy achip. They could buy a component. Theycould buy our networking. They could buyour We have customers buying only ourCPU. You know, just buy our GPUs and buysomebody else's CPUs and somebody else'snetworking. You know, we're kind of okayselling any way you like to buy. Youknow, my only request is just buy alittle something from us. You know, yousaid, you know, this isn't just aboutbetter models. We also have to build. Wehave to we have to have worldclassbuilders. And you said, you know, themost world-class builder maybe that wehave in the country is Elon Musk. And wetalked about Colossus one and what he,you know, what he was doing there,standing up a couple hundred thousand,you know, at the time H100s, H200s in acoherent cluster. Now he's working onColossus 2, you know, which may be500,000 GBs, um, millions of H100equivalents in a coherent cluster. Iwould not be surprised if he gets to agigawatt before anybody else does inone.>> Yeah. So say a little bit about that.The advantage of being, you know, thebuilder who, you know, isn't justbuilding the software and the models,but understands what it takes to tobuild those clusters.

00:57:42>> Well, you know, these AI supercomputersare complicated things. The technologyis complicated. Procuring it iscomplicated because of financing issues.Securing the land power and shell,powering it is complicated. building itall, bringing it all up. I mean, theseare this is unfortunately the mostcomplex systems problem humanity hasever endeavored.And and so Elon has has a greatadvantage that in his head um all ofthese systems are interoperating and umand the and the interdependencies um areare um you know resides in one headincluding the financing.>> Yes.>> And so>> he's a big GPT. He's a big supercomputerhimself.>> He's the Yeah. the ultimate GPU. Yeah.Yeah. And so so he has a great advantagethere. Yeah.>> And and he has a great sense of urgency.He Yeah. He has a has a real desire toto build it and and so when when willcomes together with>> with skill.>> Yeah.>> You know, unbelievable things canhappen.>> Yes.>> Yeah. Quite unique.>> Something you've been so involved in isI want to talk about sovereign AI. Iwant to talk about China and the globalAI race that's going on. You know, whenI look back at you 30 years

00:58:58ago, youcouldn't have imagined you were going tobe hanging out in palaces with airs andthe king this week and you're at theWhite House all the time. The presidenthas said that you and Nvidia arecritical to uh US, you know, nationalsecurity. So when when you look at thatfirst just contextualize for me likeit's hard to believe that you would bein those places if sovereigns didn'tview this at least as existential asimportant as maybe we did nuclear in the1940sright we don't have a Manhattan projecttoday at least funded by the governmentbut it's funded by Nvidia it's funded byopen AI it's funded by Meta it's fundedby Google we have companies today thesize of nation states and thank god Godfor America, right? Who are fundingsomething that it appears to mepresidents and kings think are think isexistential to their future economic andnational security. Would you agree withthat?>> Nobody needs atomic bombs. Everybodyneeds AI.>> Well said.>> Okay.>> Here. Here.>> Yeah.>> Here.>> And and so that's a very very largedifference. Um AI AI as you know ismodern software. I just that's where Istarted from general purpose computingto accelerated computing from humanwritten code line at

01:00:13a time to AIwritten code that foundation can't beforgotten we've reinvented computingthere's not a new species on earth wejust reinvented computing and everybodyneeds computing it needs to bedemocratized>> which is the reason why everybody all ofthese all of the countries realize theyhave to get into the AI world becauseeverybody needs to stay in computing.There's nobody in the world that says,"Guess what? You know, I used to usecomputers yesterday. I'm pretty goodwith, you know, clubs and fire tomorrow,you know, and so everybody needs to moveinto computing. It's just it's justbeing modernized. That's all." Okay.Number one, um it it is the case thatthat in order to participate in AI,you have to encode within AI, your yourhistory, your culture, your values.>> And and of course, AI is getting smarterand smarter so that even the core AI isable to learn these things fairlyquickly. You don't have to start fromthe ground, you know, from ground zero.And so I I think that that every countryum needs to have some sovereigncapability. I recommend that they alluse OpenAI, they all use Gemini, theyall use, you know, these open models useGrock and I think they I recommend theyall

01:01:28do that. I I recommend they all useanthropic.>> Um but they should also dedicateresources to learn how to build AI. Andthe reason for that is because they needto learn how to build it not just forlanguage models, but they need to buildit for industrial models, manufacturingmodels, national security,>> national security models. There's awhole bunch of intelligence they had togo cultivate themselves. So they theyought to have sovereign capability.Every country should develop it.>> And is that what you see? Is that whatyou're hearing around the world?>> They all realize it.>> They all realize>> they all realize it. And they they allare going to be customers of OpenAI andThrobic and Grock and Gemini, but theyall really need to also build their owninfrastructure. And this is this is thebig idea that that what Nvidia does iswe're building infrastructure. Just asevery country needs energyinfrastructure, the communications andinternet infrastructure, now everysingle country needs AI infrastructure.So you let's start with the rest of theworld. You know, our our good friendDavid Saxs. Um the AIS are doing a heckof a job.>> We are in so lucky.>> Yeah.>> To have David and Shriram in WashingtonDC. Um doing you know and David doingYeah. doing AI in the AISR. Uh this atwhat a what a smart move

01:02:43by PresidentTrump to put them in the White House.>> Um because during this pivotal time>> Yes.>> the technology is complicated.>> Shriram is the only person in WashingtonDC that I think knows CUDA.>> Yeah. Um and andwhich is strange anyways. But but I Ijust love the fact that during thispivotal time when technology iscomplicated, policy is complicated, theimpact to the future of our nation is sogreat>> that we have somebody who isclear-minded, dedicating the time tounderstand the technology>> and thoughtful to uh help us throughthat.>> And it would seem to me, yeah, I'm goingback to the Manhattan Project analogy.>> Yeah.>> Right. that you have a president whounderstands>> how existential this is. You havegovernors like Greg Abbott in Texas whowant to remove regulations to acceleratebecause they understand how important itis. You have secretaries right at energyand Doug Bergram at interior and Lutnikat commerce who also understand howimportant this is, how pro- energy theyare. Could you imagine?Could you imagine the alternativeif we had an administration right nowwho is not proen energy and want energyto grow in our nation so that we couldhave

01:03:58AI?>> I find it>> I just can't even think about it.>> I find it ironic that that that just acouple years ago we were saying China'sbuilding a 100 nuclear reactors. They'reso far ahead of us. Like that's the thethe primitive to AI. But now you havepeople when we go to build it, everybodysays, "Oh, it's a glut, right?" Like itseems to me that this is something thatthe government, it is in their interest.And we have industry and governmentworking together in a way that I haven'tseen in a long time. You've been arounda long time. You you're very close withPresident Trump at this stage. Help usunderstand like what is the nature ofindustry government relationships? Wesaw that dinner last week with all theCEOs. You know, you spent a lot of time.Is it unique? Have you seen anythinglike this in your career over the last30 years?>> It was it was hard to go to DC in thepast as you know. Uh getting anappointment is almost impossible,>> right?>> Uh President Trump has a open door toleaders who wants to come in and uh helpthem understand the future. Um this isan administration that believes ingrowth.>> Fundamentally, President Trump wantsAmerica to grow.>> Yeah. If we can grow economically, wewill be strong militarily. If we couldbe if we could

01:05:13grow economically, wewill be secure. I've never met somebodywho is secure who's poor.>> Being being rich as a nation is anessential part of national security. Andhe knows that.>> He also wants America to win the AI thethe AI race. This is going to be a verylong-term race. And um and heunderstands that this is a pivotal time.He wants the technology industry to run.He wants everybody in the world to bebuilt on American technology.>> These are sensible, logical things.>> You know, the opposite is strange to me.If I take everything and I just reversedit, we want our country not to grow>> and because we don't want our country togrow, we don't need any energy becausewe know we need energy to grow and solet's not have any energy and in fact wedon't want our technology industry tolead. uh he understands that ourtechnology industry is our nationaltreasure.>> Correct.>> And that technology like corn and steeland things in the past are now suchfundamental trade opportunities. It's anessential part of trade. And why wouldyou not want American technology to becoveted by everyone so that it could beused for trade?>> Right. So let's talk about

01:06:28you know>> the internet.>> Yeah. Google spread around the world.Yeah, we had democratic values spreadaround the world by way of search andGoogle didn't have to go to Washingtonto get permission to do it. It justhappened. We diffused our technologyaround the world. David Sachs has beencrystal clear of the need to accelerateexport licenses so that the American AIstack wins around the world. Right?We're talking chips, we're talkingmodels, we're talking data centers, etc.We know a year and a half ago thatwasn't happening. was a concept that wascalled small yard tall fence orsomething like that. A small yard tallfence and andthe irony of it was it was described insuch a way and it was recommended inpolicy in such a way it was a small yardtall fence around America.That was the strange part. I thinkPresident Trump's got it right that wewant to maximize exports.We want to maximize American influencearound the world. We're supposed tomaximize those things.>> And do you see those licenses coming?Are you seeing the acceleration inWashington? I know it's being said atthe top, but are you seeing it flow downthrough government that's acceleratingus around the world?>> Secretary Lutnick was all over it.>> Great. Yeah.>> So, now let's talk about

01:07:44China. You knowwhat most people may not realize is Ithink you understand China as well asany leader in the United States.>> We've been there for 30 years.>> Been there for 30 years. What mostpeople don't realize is is up until acouple years ago, you had dominantmarket share within China in terms of>> 95% market share.>> 95% market share in the most importantthing arguably.>> And you have said that our biggest goalthat we as a country could have underthe guise of somehow trying to slow themdown is we've unilaterally disarmed. Weforced Nvidia out of China, which hasallowed Huawei to accelerate on the backof monopoly profits within China. And Ijust saw this morning, you're seeingannouncements out of Huawei and Baba andothers that they're going to build datacenters around the world. Now Huawei hasa three-year plan to pass Nvidia fundedby the monopoly profits in the biggestAI market in the world. So it's lookinglike your admonitionthat this is a huge mistake to handChina, you know, monopoly markets iscoming true. The president said, youknow, a a after

01:08:59kind of the ban on H20s,now we have a situation where you cansell, you know, chips to China, butthere's a 15% export tax. But now itappears that the Chinese perhapsoffended by statements out of the UnitedStates are saying no, Nvidia is notallowed to sell here. Now where do westand today between Nvidia and China?And can you reiterate kind of what youthink we as a country should be doing toput ourselves in a best position to winthe AI race around the world?>> We have a competitive relationship withChina. We should acknowledge that Chinarightfullyshould want their companies to do well.We I don't I don't for a second begrudgethem for them. They should do well. Theyshould they should give them as muchsupport as they like. It's all theirprerogative.And don't forget that China has some ofthe best entrepreneurs in the worldbecause they came from some of the bestSTEM schools in the world.They're they're the most hungry in theworld.>> Yes.>> 996 as you know.>> This is a very>> producing the most AI engineers in theworld.>> 96. So the audience knows 9 in themorning to 9

01:10:14at night 6 days a week.>> That is their culture.>> Yeah.>> Okay. We're up against a formidable,innovative, hungry, fastm moving,underregulated.>> Yeah.>> Okay. People don't realize this. Theyare very lightly regulated,>> right? Less regulated, ironically, thanwe are in a capitalist system.>> That's right. People think that they'recentrally governed. But remember, thegenius of China was distributed economicsystems.>> Yeah.>> And so all of these 33>> provinces and all the mayor economy hasdriven enormous amount of internalcompetition, internal economic vibrancy,which of course has some of its sideeffects. But this is a vibrant,entrepreneurial, high techch, modernindustry. And two, one, uh, some of thethings I I heard, uh, they could neverbuild AI chips.That just sounded insane. Two, uh, thatChina can't manufacture. China can'tmanufacture.If there's one thing they could do ismanufacture. And three, they're yearsbehind us. Is it two years, three years?Come on. They're nanconds behind us.>> Nanoc.

01:11:29>> Yeah, they're nanconds behind us. And sowe've got to go compete.>> Yeah,>> we've got to go compete.>> And so so the question then becomes umwhat's in the best interest uh what's inthe best interest of China of course umis that they have a vibrant industry. Uhthey also publicly say and rightfully Ibelieve they believe this is that theywant China to be an open market. Theywant to attract uh foreign investment.They want companies to come to China andcompete in the marketplace,>> right?>> And I believe that they I hope I believeand I hope that would return to that>> in our context. Answering your question,what do I see in the future? I do hopebecause they they say it um theirleaders say it and I take it at facevalue and I believe it because I thinkit makes sense for China that what's inthe best interest of China is forforeign companies to invest in China,compete in China>> and for them to also have vibrantcompetition themselves and they wouldalso like to come out of China andparticipate around the world. That is Ithink is a fairly sensible outcome andwe what we need to do as a country is toenable our technology industry whichtoday is the I'm privileged to beworking in an industry that is ournational treasure. We have toacknowledge

01:12:44it is our national treasure.It is our best industry.>> It is our single best industry.>> Yeah.Why would we not allow this industry togo compete for its survival for for thisindustry to go and proliferate thetechnology around the world so that wecould have the world be built on top ofAmerican technology so that we canmaximize our economic success magn uhmaximize our geopolitical influence umaximize this this technology industryduring such a vibrant time such apivotal time to allow it to thrive.>> The skeptic says Jensen just wants tosell more chips and if he can sell themto China, great. He'll sell them toChina. He doesn't care about, you know,what that means for America. That's askeptic. Now,>> can I just can I just address theskeptics? Just because I want America inecosystem and economy to grow doesn'tmake me wrong.>> Right. Right.>> Okay. So, first of all, everything thathas been said so far that's been made upso far about U. China has proven to bewrong. The facts are just wrong. Theground truth is wrong. And sojust because we want America to win,just because we want this industry togrow, doesn't make me wrong.>> Correct. And I think anybody who knows

01:13:59you and now the president, certainlymyself, you deeply care about thecountry. You deeply want the UnitedStates of America to win the global airace. You just happen to believe and Ithink you have as much experience ormore experience than anyone that itenures to our advantage the probabilityof us winning the global AI raceactually goes up if you are competing inChina because it allows us to tap intohalf of the world's AI engineers keepingthem you know in this ecos let's beclear with the companies we're talkingabout here bite dance Alibaba etc theseare companies that are largely owned byAmerican investors Yeah. Right.>> Right. Like these are global companiesthat are building recommener enginesthat>> by the way extraordinary technologies,>> incredible companies. And so I think andI'm hopeful>> that the argument that you're makingvisav China, which is a harder argumentthan diffusion to the rest of the world.I understand that. And that's why Ithought when the president said, youknow, I don't know, it's a flip of acoin. Maybe Jensen's right. Maybe theother guys are right. If Jensen'swilling to put a little bit of 15% intothe US Treasury as a hedge on that, thenI'll go for it. Um, but I was reallydisappointed on the heels of that.>> Mhm.>> Um, I

01:15:15think if the Chinese feel likethey're being taken advantage of thatwe're going to send them chips that are,you know, 10 years old or something,then I then I get why why they had thatresponse.>> H20 is really quite spectacular still.And I of course it it's it's not as goodas Blackwell and and I get that. Yeah.>> Um I look, you know, we're I'm patientand and I believe that they're they'rewise. They're thinking through theirsituation. Um they they have they havelarger agendas to to deal with um relyou know visibly uh the United States.There are a lot of discussions going on.But I'll come back to the ground truthfundamental truth. I believe that is inthe best interest of China that Nvidiais able to serve that market and competein that market. I fundamentally believeis in the best interest of China. It isof course um fantastic in the fantasticinterest of the United States.>> Yeah,>> it is fun. But those two truths cancoexist. It is possible for both to betrue and I believe it is both true.>> And so I I um uh I I'm rather,>> you know, even though I tell all of ourinvestors>> Yeah.>> that our guidance includes no China.>> Yeah. And I appreciate all of ourinvestors to include no China in

01:16:30any ofour guidance.>> We've got plenty of growth opportunitiesoutside and we you know we've got all ofthat is true.>> It doesn't make China not important tous. It's very important to us. Anybodywho thinks that the Chinese market isnot important is has their head deep inthe sand.>> Yeah.>> And so this is s one of the mostimportant markets in the world. Smartmarkets as you know, smart people doingsmart things and we want to be there.>> Yeah. And I think it's in the bestinterest of both countries that we arethere. And so I think when I take a stepback, I am confident that ultimately thewisdom will prevail.>> Yes.>> Yes.>> I've I've always been confident thatwisdom prevails. I've always beenconfident that that truth prevails anduh it's taken me this farand uh I believe I believe that to befundamentally true now. And so thesethings will get sorted out and we willhave the opportunity to go compete inthat China market.>> I'm not very political, but very topicalis the administration's decision tocharge 100,000 per H1B visa.>> Mhm.>> You've spent a lot of time with thepresident. Um you've called him oursecret weapon in AI.>> I also know you want to recruit the bestand brightest to our country.>> Yeah. So, how do you

01:17:45think about thedecision to charge 100,000 per H1B visa?Does this make it easier or harder torecruit talent? And, you know, doesperhaps it's a little different forlarge companies or small companies?Like, how do you think about it?>> I'm going to start with it's a greatstart.>> Hold on. You said it's a great start.>> It's a great start.>> I'm just going to start there. And thereason for that is this.>> That implies I don't I hope it's not theend,>> but I think it's a great start. I justhope it's not the end. Here's what Ifundamentally believe.America has one a singular brandreputation that no country in the worldhas. And no country in the worldis in a position or in the horizon to beable to say come to America and realizethe American dream. M>> what country has the word dream behindit?>> Yes, it's part of its brand.>> We are utterly singular and you'retalking to somebody who represents theAmerican dream. My parents didn't haveany money. Sent us over here. We startedfrom nothing. You guys know I, you know,bust tables, wash dishes, clean toilets,and here I am.

01:19:00>> Yeah.>> This is the American dream. PresidentTrump knows that>> we want legal immigrants.>> Yeah. there's a difference between legalimmigrants and illegal immigrants.>> But the idea that it's a country that'sfree for all doesn't make sense.>> And so now the question is how do we gofromthe the idea that we want to protectfundamentally the American dream todealing with illegal immigrants at sucha large scale? Um how do we find alogical pragmatic solution?>> Right? So, the idea that he that thatthat we put a $100,000 price tag on H-1Bum probably sets the bar a little toohigh,>> but as a first bar, it at leasteliminates um illegal immigration, andthat's a good start.>> How does it how does eliminate illegalimmigration?>> Well, it at least it eliminates abuse of>> abuse of H-1B. Yeah. Yeah. At least. Andand that's a good start. and at least wecan have a conversation.>> So, one of the things that we know aboutPresident Trump, he's he he's a goodlistener.>> He actually listens. I mean, he listensto you, he listens to me, and>> he doesn't have to. And he listens to alot of people, and he's integrating

01:20:15alot of information, and this isobviously a very complicated issue.>> And so, so I think that that this is afine start. It's a fine start. But I II'm not confused that that anyone in theadministration, anyone in the WhiteHouse is confused thatlegal immigration, immigration is thefoundation of the American dream and isthe ultimate brand that we want toprotect and that's the future we want toprotect.>> And I would also say it seems to me thatcertainly Saxs and other people in theadministration know that we have torecruit the world's best and brightest.we should not sacrifice the greatness ofthe brand. Um, charging $100,000 orlet's say, you know, it got lowered to50 or whatever the case is,>> it does seem like it it it it tilts theplaying field in favor of big companieswho can effectively sponsor all thesepeople,>> right?>> And it's more challenging for thestartup ecosystem where people arealready super expensive and now I got topay this fee on top of it.>> It also has an un an unintendedconsequence. It might um accelerateinvestment outside United States,>> right? And so there there are unintendedconsequences, but like I said,>> start somewhere, move towards the rightanswer, right? You know, often timespeople want to go directly from a wronganswer, wrong condition.

01:21:30We don't wantthis condition where we're at,>> right?>> And directly jump to the perfect answeris hard to find,>> right?>> Just start somewhere. It's theentrepreneurial way.>> It's important to me, you know, thepresident talked about before when hewas running for office, he wanted tostaple a green card to the, you know, tothe diplomas of these STEM students. sosmart>> people coming to the United States fromfrom China AI researchers studying atStanford like we want to keep them herewe want to get you know and by the way>> if their families can't get here>> they're going to leave after a few yearsso you might even want to make it easierfor their families to come here andothers are you confident that we have astrategic plan in this administrationyou know this is a start but yourconvers conversations they give youconfidence that we have a broaderstrategic plan to make sure we'rerecruiting the best and the brightest.>> I don't know that I have an answer forthat.>> Okay.>> Um but I understand that where we're atis not where we want to be.>> Yeah.>> And I don't think anybody's lost losttheir focus on, you know, the Americandream, the importance of immigration,the importance of attracting all of theworld's best talent to United States,create the conditions for them to stayhere. I there are things that are doneum from time

01:22:45to time that works againstwhat I just described,>> right?>> Um making making foreign studentsuncomfortable,>> right,>> and being here in the brand threatensthe brand.>> Um uh let's not let's not forget that>> that it's okay to be competitive withChina,>> but be careful not to be tough onChinese.>> And so we need to make sure that thatslippery slope isn't crossed. Yeah.>> Um, you know, and so there there all ofthese things that goes along with withfinesse and nuance.>> But the fact of the matter is we knowwhere we want to be. We know we're in adifficult situation. We don't want to behere and President Trump doesn't havemuch time to move us in that direction.>> Right.>> And so to the extent that we move inthat direction,>> I believe it's a good start.>> Agreed.>> Yeah. I heard from a Chinese researcherleading one of our leading labs in theUS>> that three years ago 90% of the top AIresearchers graduating from universitiesin China>> wanted to come to the United States anddid come to the United States to work inour leading labs>> and he guessed that today that's closerto 10 or 15%.>> Right? So seen a precipitous drop.>> That's precisely

01:24:01a concern that we have,>> right? So have you seen this? Have youyou you know you're you're payingattention to both markets. Do you seethis? And what are the things we need todo in order to reverse that?>> Definitely see a greater concern of ofChinese students um who who come hereand um uh remain here.>> Yeah. and or many of them who come herefor for school and are thinking aboutgoing elsewhere, right?>> Many of them thinking about Europe,>> right?>> And so so I think I think we need to besuper super concerned about this. Thisis this is>> this is a source>> of existential crisis.>> This is definitely the early indicatorsof a future problem.>> Right. Right. you know, smart people'sdesire to come to to America and smartpeop smart students desire to stay,>> those are what I would call KPIs.>> Yes.>> Early indicators of future success.>> Yes.>> I think of it a bit like the Warriors.You know, if they have an advantage ofrecruiting all the best players in theNBA, right, then they can continue towin championships. Y,>> but the second that recruiting pipeline,>> right, because of the brand of theWarriors gets diminished or somethingelse happens,

01:25:16>> um, then they're not going to be able torecruit the best future players andyou're not going to win championships.And when I you talk about the Americandream so eloquently, that being brandUSA,>> right? The right to come here and to todo what you've done. And you know, so Ihope that the feedback to thisadministration, it's not just theadministration, it's also just how we asa country talk about immigration.>> That's right.>> Right. This needs to be the place thatwelcomes the best and the brightest,that attracts, has a strategic plan forrecruiting the best and the brightestand making sure that this is the placethat they want to work.>> As you know, there's in in there's athere's a phrase, and I didn't hearabout this phrase until just a few yearsago, China hawks. Yes.>> And and apparently that if you're aChina hawk, you get to wear that labelwith pride. It's almost like a badge ofhonor,>> right?>> It's a badge of shame.>> There's no question it's a badge ofshame.>> There's no question that although theywant what's in the best interest of ourcountry, and we all want what's in thebest interest of our country,>> um destroying that pipeline,>> right,>> of the American dream,>> Yeah.>> is not patriotic,>> right?They they think they think they're doingthe right thing for our country, butit's not patriotic. Not Not even alittle bit.>> And so we we need

01:26:32to um continue to beuh the great country we are, to have theconfidence,>> right,>> of a great country.>> Yes. Well said.>> And to have the confidence of a greatcountry and and have somebody who wantsto compete with us and to have theattitude, bring it on.>> Right. Right.>> Bring it on.>> Right. because I believe in I believe inour people. I believe in our people. Ibelieve in the people that are here. Ibelieve in our culture. I believe in ourcountry. I believe in our system. Bringit on. And is it your take that that'swhere the president is? Like he's a he'sa pragmatist. He's a he's a believer inthe growth and the ability of the UnitedStates to compete. Um it seems to methat's where he is.>> There's no question President Trump isthe bring it on president.>> Right. Right. And he doesn't seem to melike the reason I'm confident and I'vesaid on this pod that I think he'll geta big deal done with China.>> I I really really do hope so.>> Yeah.>> And and I I think he he speaks he hespeaks positively um uh uh with greatrespect and um great eloquence aboutabout his relationship and theimportance of China. Uh, not one timehave I ever heard him say

01:27:47the worddecouple, which we heard a lot in thelast administration,>> right?>> Um, you can't decouple against um thesingle most the two most importantrelationships for the next century. Thatdoesn't make any sense at all.Decoupling is exactly the wrong concept.>> Right?>> I mean, it seems to me he and ScottBesson are saying, "Listen, we need tomake America great. We need tore-industrialize America. We need tobalance and make sure that we have fairtrade. Um that we protect industriesthat we need to help build that Chinahelps us do that recognizing that wehave helped them do it over the courseof the last 25 years. But thatultimately he said the best way tounderstand me is I'm a great dealmaker.I make deals right whereas I think inother camps there are people who areiconoclastic or dogmatic. Uh you knowit's the mere shimer view of China thatthere's a great power struggle. one wasmust win and one must lose versus thisidea>> the idea that every country has to lookexactly like ours,>> right?>> You know, and we we want diversity.>> You want America to win, but thatdoesn't have to come at the expense ofpoking an eye and telling somebody elsethey have to lose>> because we're that confident.>> Yeah, we're that confident.>> Because we're that mighty. Because we'rethat incredible. I've

01:29:02got no trouble, asyou know, I've got no trouble workingwith all my colleagues in the ecosystem,right? And notice we just did theultimate deal,>> right? Partnering with Intel,>> a company that spent most of its lifetrying to put us out of business, right?>> And I had no trouble partnering withthem,>> right?>> You know, and so, and the reason forthat is because number one, bring it on.>> Yes.>> And number two, the future is so muchgreater.>> It doesn't have to be all us or them. Itcould be us and them.>> Yeah.>> Yeah. But nonetheless, bring it on.>> Yeah. Agreed. You know, you you youmentioned something that's profoundlyimportant to both of us. You and I havetalked a lot about this, the Americandream,>> you know, and it was, I think, AbrahamLincoln who said, "Fundamental to theAmerican dream is the right to rise.">> Yeah. That's right.>> The belief that your kids can do betterthan you did.>> That's right.>> Right. And you you've experienced theright to rise. We've all experienced theright to rise in America.>> So, yeah, you go to Wikipedia, you lookup American dreams, my picture,>> right?>> Yeah. And the ultimate American dream.And yet we live at this time wherebecause of the nature of thesetechnological systems, we have companiesthat are going to be worth 10 trillion.We'll probably have individuals that areworth a trillion. Those are theincentives that give people the theencouragement to rise. But at the

01:30:17sametime, when we head into this age ofabundance, something that I was deeplyworried about was that too many peopleget left behind.>> Yeah.>> Right. And they feel left out and leftbehind. So it makes sense for them toattack this system of capitalism.Something that you and I worked ontogether and I'm deeply grateful for wasthe idea of invest America that we haveto start every kid at birth on thecapitalist right to rise journey givethem a thousand bucks in great companieslike Nvidia>> social security>> and and and open a high etc. Um and theybenefit>> right as the as the comp country winsthey win and they own it individuallythey can see it on their>> every kid is a shareholder in the futureof America>> of America. So on the 200 because ofyour support and I wanted to take thechance on this podcast and the supportof>> Well, I want to thank you for startingit, for driving it.>> Yeah.>> Yeah. What a great idea.>> And you know, so this>> you're a genius.>> The Please this passed>> in the big beautiful bill.>> Most people don't even realize that yet.>> Starting in 2026, every child bornforever more in the history of thiscountry>> will start off with an investmentaccount at birth. Yeah.>> Seen a thousand bucks in the bestAmerican companies

01:31:32and your company hasagreed to add to the accounts of notonly the kids who work for youremployees but maybe even kids of others.I'm going to adopt schools, you know,and lots of philanthropists andcompanies. We think every company acrossAmerica>> wonderful way for companies to giveback,>> right?>> Yeah. as part of the 401k.This seems to me to be part of thechange in the social contract that needsto occur because if if we're seeing thisexponential progress, we know that thethe evolution of government in thesocial contract needs to keep up withit. Um obviously President Trump and andbipartisan group in the House and Senatepassed this into law. So maybe just talkto us a little bit when you think aboutthe the pace and magnitude of changesthat are coming, right? Um I know youbelieve it will be a net good, but therealso going to be a bunch of peopledisplaced along the way. We probablyneed things like this and other things,>> right? In order to, you know, bringeverybody along for the journey. There'sseveral things that that President Trumphas done and let me just start startthere has done that is incredibly goodfor bringing everybody along. The firstthing is reindustrializing America.>> Yeah.President Trump, Secretary Lutnik,

01:32:48theyou know they're all in behind that allthe work that they're doing encouragingcompanies to come build here in theUnited States, investing in factoriesand uh reskilling and upskilling thatskilled labor workforce, right?>> Incredibly valuable to our country. umthe idea that that we no longer uh uh uhmake it only that you get a PhD or yougo to you know one of the great schoolsand only in that way can you build agreat life right>> and and deserve to have a great livinguh we've got to change all that doesn'tmake any sense we love craft right>> I love people who make things with theirhands and and we're now we're now goingto go back and build things>> build magnificent incredible things Ilove that>> yes>> that's going to transform form America.There's no question about that. There'sa whole there's a whole a whole band ofan economy, a whole band of society thatthat um uh has been largely left behindbecause we outsourced everything,>> right?>> Now, I'm not suggesting we insourceeverything,>> right?>> You know, all the people arguing about,you know, manufacturing tennis shoes andtoothpicks, I mean, you know, that'sthat's denigrating a perfectly gooddiscussion into some insane level. Youknow, we we've got to

01:34:03um recognize thatre-industrializ reindustrializingAmerica is is just fundamentally goingto be transform transformative numberone. Number two,>> and aspirational.>> Oh, it's fantastic.>> Elon taking us to Mars, watchingspaceships caught with, you know, uhchopsticks out of the sky. This is notonly great for the industrializing baseof America, it's aspirational for>> Fantastic. That's right.>> And then and then, uh, of course AI.Yeah,>> it is the greatest equalizer. Just thinkeverybody can have an AI now. The theultimate equalizer. We've closed thetechnology divide. Remember the lasttime that somebody had to learn haswants to use a computer uh for theireconomic um or career benefit. They haveto learn C++ or C or at least Python.Now they just have to learn human,>> you know. And so, and if you don't knowhow to program an AI, you tell the AI,"Hi, I don't know how to program an AI.How do I program an AI?" And the AIexplains it to you>> or does it for you.>> It does it for you. And so, it'sincredible, isn't that right? It's Andwe've now closed the technology dividewith technology.>> Yeah.>> This is something that every everybody'sgot to engage. You know, OpenAI has 800million active users.

01:35:18Um gosh, it reallyreally needs to be 6 billion.>> Yeah.>> Right. It really needs to be 8 billionsoon. And so so I think that that that'snumber one. Then number two, uh and thennumber three, you know, I think the theum AI will change tasks.>> Yeah.>> The thing that people confuse is thereare many tasks that will beeliminated. There are many tasks thatwill actually be created.>> But it is very likely that for manypeople their jobs are gainfullyprotected. Right.>> And so, for example, I'm using AI allthe time. You're using AI all the time.My analysts are using AI all the time.My engineers, every one of them use AIcontinuously.>> And we're hiring more engineers. We'rehiring more people. We're hiring acrossthe board. The reason for that isbecause we have more ideas.>> Yes,>> we can now go pursue more ideas. Thereason for that is because our companybecame more productive. And because webecame more productive, we became morerich. We became more rich, we can hiremore people to go after those ideas,>> right?The I the concept that that AI comesalong and therefore there's going to bea mass destruction of jobs starts withthe I starts with the premise that wehave no more ideas.

01:36:35>> Right.>> It starts with the premise we havenothing left to do. Everything we'redoing in our lives today.>> Yeah.>> This is the end.>> Yeah.>> And if somebody else were to do that onetask for me, I have one task left. Now Ihave to sit there and wait forsomething.>> Yes.>> You know, wait for retirement, sit on myrocking chair. That idea doesn't makesense to me.>> And so I think I think that intelligenceis not a zero sum game. The moreintelligent people I'm surrounded by,the more geniuses I'm surrounded by,surprisingly, the more ideas I have, themore problems I imagine that we can gosolve,>> the more work we create, the more jobswe create. And so I think for for um Idon't know what the world looks like ina million years that's going to be leftfor my my children. Um but for the nextseveral decades, my sense is thateconomy is going to grow. Lots of newjobs are going to be created. Every jobwill be changed. Some job some jobs willbe lost. And um uh we're not going tobe, you know, riding horses on streetsand those things it'll be fine. Youknow, humans are are famously skepticaland terrible at understandingcompounding systems and they're evenworse at

01:37:50understanding exponentialsystems that accelerate with size.>> We've talked about exponentials a lottoday. You know, the great futurist RayKerszswhile said in the 21st century,we're not going to have a hundred yearsof progress. We're likely to have 20,000years of progress.>> Right.>> Right.You said earlier, we're so fortunate tobe living at this moment andcontributing to this moment. I'm notgoing to ask you to look out 10 or 20 or30 years because I think it's sochallenging. But when we think about2030

01:38:25things like robots,>> 30 years is easier than 2030.>> Oh, really?>> Yeah. Yeah.>> Okay. So, I'll get I I'll grant youlicense to go out 30. as you think outover the course of I like these shortertime frames because they have to marrybits and atoms bits and Adams the hardpart of building this stuff rightbecause everybody's saying it's going tohappen>> is interesting but not helpful>> exactly but if we have 20,000 years ofprogress reflect on that statement byRay reflect on exponentials and how allof our listeners whether you work ingovernment whether you're in a startupwhether you're you know running a bigcompany need to be thinking about theaccelerating rate of change, theaccelerating rate of growth and how youwill, you know, be co-intelligent inthis new world.Well, there there are a lot of thingsthat that that many people have alreadysaid and and um and they're all they'reall very sensible. I think the the um uhin the next 5 years, one of the thingsthat that is really cool that's going toget solved is the fusion of artificialintelligence and megatronics, robotics.And so we're going to have we're goingto have, you know, AIS that are that aregoing to be wandering around us.>> And uh we all and that everybody

01:39:40knowsuh we all all know that we're going toall grow up with our own R2-D2.>> Yeah.>> And that R2-D2 remember everything aboutus and coach us along the way and be ourcompanion. We already know that.>> Okay. and and so the idea and and andthe idea that every human will havetheir own GPUs associated with them inthe cloud um and that they're 8 billionpeople 8 billion GPUs that's a you knowviable outcome.>> Yeah. you know, and so>> and each having their own model that'sfine-tuned for them,>> fine-tuned for them. And they that thatAI is in the cloud is also embodied in awhole bunch of it's embodied in yourcar. It's embodied in your own robot.It's everywhere with you.>> And so that I think that that future isa very sensible thing. the idea that uhwe're we're going to understand the theinfinite complexity of biology andunderstanding the system of biology uhand how it how how to predict it andhave digital twins of every everybody umour own digital twin for health carelike we have a digital twin for shoppingat Amazon why wouldn't we have ourdigital twin at healthcare of course wewould and so uh you know a dig a systemthat that predicts how our how we'regoing to age what disease will likelygoing to have and uh anything

01:40:55that'sthat's about to happen maybe even nextweek or you know tomorrow afternoon andpredict it early. Of course, we're goingto have all that. And so I think all ofthat is a given.Um I think the the the the part thatthat I'm asked a lot by uh CEOs that Iwork with about now given all of that,what happens? What do you do? And thisis this is the this is a common sense ofof things that move fast,>> right?>> If you if you have a if you have a trainthat's about to get faster and fasterand go exponential, the only thing thatyou really need to do is get on it.>> Yeah.>> And once you get on it, you'll figureeverything else out along the way.>> Right.>> And so to predict where that train'sgoing to be,>> right,>> and try to shoot a bullet at it. Yeah.or predict where that train's going tobe and it's going exponentially fasterevery second and go figure out whatintersection to wait for it,>> right?>> That's impossible.Just get on it while it's going kind ofslowly,>> right?>> And go exponential along the way. A lotof people think this just happenedovernight. You know, you've been you'vebeen at this for 35 years. I rememberhearing Larry Pageige say probablyaround 2005 or 2006

01:42:11that the end stateof Google will be when the machine canpredict uh the question before you evenanswer it before you even ask it andgive you the answer without having tolook right. I heard Bill Gates say in2016>> because contextually you must be askingabout you must be wondering about that.>> I heard Bill Gates say in 2016 whensomebody said hasn't all the things beendone we've had the internet we've hadcloud we've had mobile social etc. Hesaid, "We haven't even started." Hesaid, "What do you think? Why would yousay that?" He said, "We won't even beginuntil machines go from being dumbcalculators to beginning to think forthemselves, to think with us,>> right?" Um, kind of that is the moment,>> right, that we're in. I think to haveleaders like you, leaders like Sam andElon, Satcha, etc., it's such anextraordinary advantage for thiscountry, right? and to have thecooperation that we see between a systemof risk capital that I take, you know,that I'm part of which can provide therisk capital for people to do. We don'twe're not relying on government having aManhattan project. We can actually dothis ourselves and together for thebenefit of the country. It's anextraordinary time>> and at a scale that's unimaginable.>> Right. Right. It's an extraordinary

01:43:26time. But I also think, you know, one ofthe things that I'm just grateful isthat we have leaders who also understandtheir responsibility to the fact that weare creating change at an acceleratingrate. And we know while it will mostlikely be great for the vast majority,there'll be challenges along the way.And we'll deal with those as they come>> and raise the floor for everybody andmake sure that this is an a win, notjust for some elite bureaucrats at thetop hanging out in Silicon Valley.>> And don't scare them. Bring them along.Don't scare them. Bring them along.>> And we will.>> Yeah.>> So, thank you for that.>> Exactly.

01:44:03[Music][Applause]As a reminder to everybody, just ouropinions, not investment advice.