Latest / Bloomberg Intelligence / Bloomberg Technology Special: Nvidia CEO Jensen Huang

Transcript

Bloomberg Technology Special: Nvidia CEO Jensen Huang · Bloomberg Intelligence

• 0:02
  Speaker 1: Bloomberg Audio Studios, podcasts, radio news.
• 0:08
  Speaker 2: From Mahart where innovation, money and power Collie in Silicon
• 0:13
  Speaker 2: Vallet NBN.
• 0:14
  Speaker 3: This is Bloomberg Technology with Caroline Hyde and Ed.
• 0:17
  Speaker 4: Ludlow live from San Francisco to our TV and radio
• 0:35
  Speaker 4: audiences around the world. Welcome to a special edition of
• 0:39
  Speaker 4: Bloomberg Technology. I'm Ed Ludlow. In just a few moments
• 0:42
  Speaker 4: in video, CEO Jensen Wang will join us for a
• 0:45
  Speaker 4: live interview following their latest earnings report, the company posting
• 0:49
  Speaker 4: a revenue forecast that beat consensus that fell short to
• 0:53
  Speaker 4: some of the most optimistic estimates, stoking concern that the
• 0:57
  Speaker 4: explosive growth is waning. Let's get right to Bloomberg Semiconductor
• 1:01
  Speaker 4: correspondent Ian King, who joins me on set. Let's start
• 1:04
  Speaker 4: with the basics, the fiscal third quarter forecasts and what
• 1:07
  Speaker 4: we learned through it.
• 1:07
  Speaker 5: Yeah, that forecast was fine. If you compare it to consensus,
• 1:11
  Speaker 5: it was there or thereabouts, and for most of the
• 1:14
  Speaker 5: companies in the world, that would be great and everybody
• 1:16
  Speaker 5: would be happy.
• 1:17
  Speaker 3: But this is in video.
• 1:18
  Speaker 5: This is a company which beats pretty much every quarter
• 1:23
  Speaker 5: and by an order of magnitude, and it didn't do that,
• 1:26
  Speaker 5: and it didn't indicate it would do that, and that
• 1:28
  Speaker 5: raised a lot of questions. As we heard on the.
• 1:30
  Speaker 4: Call, there are many storylines. I think the demand from
• 1:32
  Speaker 4: the hyperscalas is clearly intact, but Blackwell was everything. I
• 1:36
  Speaker 4: want to play a SoundBite of what Jensen Wang said
• 1:39
  Speaker 4: about Blackwell. Listened to this.
• 1:41
  Speaker 6: The change to the mask is complete. There were no
• 1:45
  Speaker 6: functional changes necessary, and so we're sampling functional samples of Blackwell,
• 1:55
  Speaker 6: Grace Blackwell in a variety of system configurations as we speak.
• 2:00
  Speaker 4: The main point here is that there was not a
• 2:03
  Speaker 4: design issue with Blackwell itself as had been reported, but
• 2:07
  Speaker 4: based on what Nvidia said, this was about the production mechanism.
• 2:11
  Speaker 4: Blue Beazine King, you did a very good job in
• 2:13
  Speaker 4: the top Life blog of explaining a GPU mask. Could
• 2:17
  Speaker 4: you just try to give a short version of that
• 2:19
  Speaker 4: to our audience.
• 2:20
  Speaker 5: Yeah, the mask is basically the blueprint, which basically that
• 2:24
  Speaker 5: is used to burn in the circuit onto the surface
• 2:27
  Speaker 5: of the chip to give it its function. Right, is
• 2:29
  Speaker 5: a very important step. That's the fundamental blueprint, and they
• 2:34
  Speaker 5: were saying we didn't get it wrong, but when it
• 2:36
  Speaker 5: came to manufacturing, it didn't produce as many good chips
• 2:39
  Speaker 5: as we wanted, so we made some tweaks to that.
• 2:41
  Speaker 5: Didn't have to redesign it, but made some tweaks, and
• 2:44
  Speaker 5: that is helping us to get a better yield.
• 2:46
  Speaker 4: It's worth noting at this stage the stocks down almost
• 2:49
  Speaker 4: seven percent in after hours, had been down more at
• 2:51
  Speaker 4: the conclusion of the call. I think basically because we
• 2:54
  Speaker 4: didn't learn enough about Blackwell. What they said was in
• 2:57
  Speaker 4: the fiscal fourth quarter of this year twenty five, there
• 3:01
  Speaker 4: will be several billion dollars of sales through Blackwell. Why
• 3:05
  Speaker 4: does the market want more and what does it want?
• 3:08
  Speaker 5: They wanted a reassurance from in videos management, and they
• 3:11
  Speaker 5: wanted reassurance in the form of details. They wanted, you know,
• 3:15
  Speaker 5: gens to Puddy's arm around everybody and say, don't worry,
• 3:18
  Speaker 5: it'll be fine. This is how much I'm going to get.
• 3:20
  Speaker 5: They asked consistently, We're asked questions of how many billions
• 3:24
  Speaker 5: and when will those billions exactly come in and collect. Kretz,
• 3:28
  Speaker 5: the CFO and Jensen Wang essentially avoided that question and
• 3:32
  Speaker 5: refuse to give that precise reassurance.
• 3:35
  Speaker 4: We're showing some of the after ours reaction, not just
• 3:37
  Speaker 4: in in video itself, but some of its peers, both
• 3:39
  Speaker 4: on the chip making side the server equipment providers, and
• 3:44
  Speaker 4: I think AMD and ARM in particular are very noteworthy.
• 3:47
  Speaker 4: Bloombogt and King stay with us. I want to go
• 3:49
  Speaker 4: out to Chicago and Bloomberg's Ryan for Lostelka on I
• 3:51
  Speaker 4: Equities team and Ryan. That's the broad summary from me
• 3:55
  Speaker 4: about the names moving and after Hours there's probably a
• 3:57
  Speaker 4: bigger picture after ours movement in the markets as well.
• 4:00
  Speaker 4: I'll start with Nvidia and work outward from that.
• 4:03
  Speaker 1: Sure, well, one thing I would say about in videos
• 4:06
  Speaker 1: after hours decline is that it does come after a
• 4:08
  Speaker 1: very strong year to day performance. I think it closed
• 4:11
  Speaker 1: up more than one hundred and fifty percent this year,
• 4:13
  Speaker 1: So even though the re forecast was maybe a little
• 4:16
  Speaker 1: bit shy of some of the most optimistic expectations, it
• 4:18
  Speaker 1: did beat expectations. And it is, you know, coming off
• 4:21
  Speaker 1: such a huge gain. So it's not necessarily surprising to
• 4:24
  Speaker 1: see a little bit of a consolidation now. And I
• 4:26
  Speaker 1: think even the decline is a little bit less and
• 4:28
  Speaker 1: the options market was anticipating. So just some context there
• 4:31
  Speaker 1: for looking at the declient. But you're right that pretty
• 4:34
  Speaker 1: much we are seeing widespread weakness following this report. All
• 4:37
  Speaker 1: the megacaps are, you know, modestly lower. We are seeing
• 4:41
  Speaker 1: much more pronounced weakness in other chip makers and chip
• 4:44
  Speaker 1: design companies and so forth. So yeah, certainly it does
• 4:47
  Speaker 1: seem like the initial read through here is negative, but
• 4:49
  Speaker 1: again it does come after a very strong start to
• 4:51
  Speaker 1: the year.
• 4:53
  Speaker 4: We made a big deal about this earnings print for
• 4:55
  Speaker 4: quite a long time. I look at something like an
• 4:58
  Speaker 4: investco QQQ, the ETF that tracks the nast that one hundred,
• 5:03
  Speaker 4: and I think it's down around a percentage point, right.
• 5:05
  Speaker 4: But in truth, Ryan, in the market's reaction, was this
• 5:09
  Speaker 4: the macro level event that we thought it was going
• 5:11
  Speaker 4: to be.
• 5:14
  Speaker 1: That's a great question. I would say that it's kind
• 5:16
  Speaker 1: of close enough in line with expectations, even if it
• 5:19
  Speaker 1: is a little bit shy the most optimistic ones that
• 5:21
  Speaker 1: I don't think this is going to really cause people
• 5:23
  Speaker 1: to really change how they're allocating, change their opinions on
• 5:26
  Speaker 1: AIS as kind of fundamental secular driver. But maybe in
• 5:29
  Speaker 1: the near term we do see a little bit of weakness.
• 5:31
  Speaker 1: I mean, again, some of these stocks have been moving
• 5:34
  Speaker 1: up so much there have been sort of kind of
• 5:36
  Speaker 1: growing calls about their valuation some concerns about that. I
• 5:39
  Speaker 1: don't know if this is the kind of absolute blowout
• 5:42
  Speaker 1: that we'll just you know, cause people to continue piling
• 5:44
  Speaker 1: into the way that they were doing earlier this year.
• 5:48
  Speaker 4: There is a lot in the news cycle inclusive of
• 5:50
  Speaker 4: and outside of in video. I think one of the
• 5:51
  Speaker 4: names that you note in after ours is super Micro.
• 5:54
  Speaker 4: It's down significantly. There's an video relationship to that, and
• 5:59
  Speaker 4: then there's the news around that name in and of itself.
• 6:03
  Speaker 1: Yeah. Absolutely so. We saw yesterday a short report came
• 6:06
  Speaker 1: out today it's delaying the filing of the ten K.
• 6:09
  Speaker 1: You know, both of those caused us a weakness in
• 6:11
  Speaker 1: the stock. I think today it was down more than
• 6:13
  Speaker 1: double digits, so you know, a lot of reasons to
• 6:15
  Speaker 1: be concerned there in general. And it is one of
• 6:17
  Speaker 1: Nvidia's biggest customers. I think it's the third largest, so
• 6:21
  Speaker 1: you know, this is just you know, another reason for
• 6:23
  Speaker 1: people who might have been kind of souring on super
• 6:25
  Speaker 1: Micro to you know, maybe be pulling the cell button.
• 6:27
  Speaker 3: A little bit.
• 6:29
  Speaker 4: Bloomdo's Ryan for Selica with the after hours action out
• 6:32
  Speaker 4: of Chicago. Thank you. Let's get the reaction from the
• 6:34
  Speaker 4: cell side with one of the stocks relative bears. D A.
• 6:38
  Speaker 4: Davidson's GIL Luria had a neutral rating and a street
• 6:41
  Speaker 4: low ninety dollars price target on shares of Nvidia. He
• 6:45
  Speaker 4: joins US now and GIL. I guess does that change
• 6:49
  Speaker 4: your barished perspective on the stock and your main takeaway
• 6:52
  Speaker 4: from the analyst school, the quarter.
• 6:56
  Speaker 2: Was a very good quarter. There ability to continue to
• 7:00
  Speaker 2: grow delivering chips at this rate at this scale is
• 7:05
  Speaker 2: fantastic and unprecedented. I don't think there's much of a
• 7:08
  Speaker 2: revenue issue here, of a growth issue here. I think
• 7:11
  Speaker 2: a little bit of a pushback is probably more around
• 7:15
  Speaker 2: the margin situation. You had a good conversation with Ian
• 7:17
  Speaker 2: about Blackwell and some of the moving pieces part of
• 7:20
  Speaker 2: what's happening with Blackwell. It's a new product, it's got
• 7:24
  Speaker 2: a few hitches along the way that put pressure on
• 7:27
  Speaker 2: gross margins, and then operating margins grew more than anticipated,
• 7:31
  Speaker 2: so less of the top line upside flowed to the
• 7:35
  Speaker 2: bottom line. And that's probably where investors are getting a
• 7:38
  Speaker 2: little bit more cautious than they were before, because in
• 7:41
  Speaker 2: the last several quarters, the huge upside to revenue flowed
• 7:45
  Speaker 2: all the way to the bottom line, creating huge upside
• 7:48
  Speaker 2: there to believe in and nvideo right now is to
• 7:52
  Speaker 2: believe twenty twenty six calendar is going to be about
• 7:54
  Speaker 2: two hundred billion of revenue and about four and a
• 7:58
  Speaker 2: half to five dollars of earnings. That's only possible if
• 8:01
  Speaker 2: they continue to get this type of growth on the
• 8:04
  Speaker 2: top line and more leverage. And the second part of
• 8:08
  Speaker 2: that equation looks a little less certain right now.
• 8:11
  Speaker 3: Gil.
• 8:11
  Speaker 4: When you said that two hundred billion dollar figure, Bloombergsy
• 8:14
  Speaker 4: and King, who's covered semi conductors at this company since
• 8:17
  Speaker 4: nineteen ninety eight, sat next to me still for what
• 8:19
  Speaker 4: it's worth, gave a little smile. I mean, it's an
• 8:22
  Speaker 4: astonishing figure. I think there's a lot of emphasis here
• 8:25
  Speaker 4: on Blackwell. How did you react to the news of
• 8:28
  Speaker 4: several billion dollars in fiscal fourth with Blackwell and the
• 8:32
  Speaker 4: explanation that this was a production issue impacting the ramp,
• 8:36
  Speaker 4: not a sort of core design issue around the product itself.
• 8:41
  Speaker 2: So, first of all, our estimate is far less than
• 8:43
  Speaker 2: two hundred billion for twenty twenty six.
• 8:45
  Speaker 4: But that's you're a relative bear, Gill, and I got
• 8:47
  Speaker 4: that bit right.
• 8:48
  Speaker 2: I think in terms of Blackwell, what happened to in
• 8:52
  Speaker 2: n videos. They used to be on a two year
• 8:54
  Speaker 2: cycle for the data center product, and they decided to
• 8:56
  Speaker 2: aggressively move to a one year cycle and then put
• 9:00
  Speaker 2: a lot of new feature functionality and bundle into the
• 9:03
  Speaker 2: Blackwell platform. So that was a very aggressive agenda, and
• 9:07
  Speaker 2: it's not that surprising that there's some challenges that they
• 9:10
  Speaker 2: have to encounter. If they can still ship in their
• 9:14
  Speaker 2: fiscal fourth quarter, that's a very good sign. And by
• 9:18
  Speaker 2: the way, their customers don't really care that much. Their
• 9:21
  Speaker 2: customers are just going to continue to buy the latest
• 9:24
  Speaker 2: and greatest, and if that's the Age two hundred, they'll
• 9:26
  Speaker 2: just continue to buy h two hundred. They have made
• 9:29
  Speaker 2: that abundantly clear on their conference calls, Microsoft, Meta, Amazon, Google, Tesla.
• 9:34
  Speaker 2: Elon Musk made it abundantly clear they'll buy as much
• 9:37
  Speaker 2: GPUs as in Video can produce, at least this year,
• 9:41
  Speaker 2: because they're saying the demand signals are there. As long
• 9:43
  Speaker 2: as the demand signals are there, they'll buy whatever and
• 9:46
  Speaker 2: video is selling, and towards the end of the year,
• 9:48
  Speaker 2: a piece of that will be Blackwell.
• 9:51
  Speaker 4: Gil Jensen Wong was it pains and gave a detailed
• 9:54
  Speaker 4: answer that basically summarizing video as a systems vendor. The
• 9:58
  Speaker 4: discussion around Envy link, CPUs, everything beyond GPU. How much
• 10:03
  Speaker 4: credit do you assign the company for that, the sort
• 10:05
  Speaker 4: of complete systems that they sell.
• 10:08
  Speaker 2: A tremendous amount of credit. Let's start with the beginning.
• 10:12
  Speaker 2: When history is written, we'll talk about how Jensen Wangen
• 10:17
  Speaker 2: Nvidia created the capabilities that make generative AI possible today,
• 10:21
  Speaker 2: in a similar way that Elon Musk is responsible for
• 10:24
  Speaker 2: the electric vehicle. And part of the genius wasn't just
• 10:27
  Speaker 2: seeing that this is coming, that accelerate computers coming, AI
• 10:30
  Speaker 2: capabilities is coming well before anybody else did. It's also
• 10:34
  Speaker 2: realizing that in order to create a moat around the chip,
• 10:37
  Speaker 2: they need to do a couple of things. One is
• 10:41
  Speaker 2: to own the proprietary software around that KUDA, but the
• 10:44
  Speaker 2: other thing is to enhance the bundle as much as possible.
• 10:48
  Speaker 2: These GPUs work in a raise when they work together,
• 10:52
  Speaker 2: ten thousands or sometimes tens of thousands of GPUs together.
• 10:56
  Speaker 2: When you box those up in a box with multiple
• 10:59
  Speaker 2: g CPUs, wiring, cabling, and firmware, you make it harder
• 11:05
  Speaker 2: for your competitors to catch up. And they've done a
• 11:07
  Speaker 2: remarkable job of that.
• 11:10
  Speaker 4: I also wonder about all the rest. The story thus
• 11:12
  Speaker 4: far has been about the hyperscale is a meta five names,
• 11:16
  Speaker 4: essentially being the core customer of Nvidia. We got some
• 11:19
  Speaker 4: detail on low double digit billions sales this year in
• 11:23
  Speaker 4: Sovereign AI. Do you see evidence that the market is
• 11:26
  Speaker 4: broadening out beyond just the cloud providers?
• 11:30
  Speaker 3: Not really.
• 11:31
  Speaker 2: It seems like those five are still a substantial amount
• 11:34
  Speaker 2: of the revenue. And those five have in their most
• 11:38
  Speaker 2: recent calls said some interesting things. One of them is
• 11:41
  Speaker 2: that they consistently said they are over investing. So they
• 11:44
  Speaker 2: said they definitely want to invest more. We're seeing that
• 11:46
  Speaker 2: in results today. But they also use the term over investing.
• 11:50
  Speaker 2: Over investing is something you do before you do the
• 11:53
  Speaker 2: other thing, and they made it clear that they're only
• 11:56
  Speaker 2: going to do so when they have the demand signals.
• 11:59
  Speaker 2: If or whatever reason, demand for AI capacity goes down,
• 12:04
  Speaker 2: for instance, because compute doesn't continue to drive performance of
• 12:10
  Speaker 2: foundational models, you're gonna see those same five actors pull
• 12:15
  Speaker 2: back and say, you know, maybe we have enough data
• 12:18
  Speaker 2: center capacity, or maybe what we have in the pipeline
• 12:20
  Speaker 2: is enough. And that's the concern for twenty twenty five
• 12:24
  Speaker 2: and twenty six is that those very companies could do that.
• 12:28
  Speaker 2: Beyond that, these are the same companies that are developing
• 12:32
  Speaker 2: their own chips. They're at various stages of this, but
• 12:35
  Speaker 2: Google's TPUs are probably as good as in videos. So
• 12:38
  Speaker 2: in terms of their internal use and selling to customers
• 12:41
  Speaker 2: such as Apple, Google's already caught up, Amazon's catching up,
• 12:46
  Speaker 2: Meta and Microsoft are earlier stages. But these same companies
• 12:51
  Speaker 2: that are buying every GPU from video they can will
• 12:54
  Speaker 2: be the ones that will buy possibly less next year
• 12:56
  Speaker 2: in the year after that.
• 12:58
  Speaker 4: I just want to reflect for our TV audience there
• 13:00
  Speaker 4: are people tuned in around the world. My understanding is
• 13:03
  Speaker 4: there are watch parties in cities from New York to London.
• 13:06
  Speaker 4: For this earnings, We've hyped it up for weeks. Probably
• 13:09
  Speaker 4: this is the most unique earnings print that I've covered
• 13:12
  Speaker 4: in my career. What about for you in the semiconductor space,
• 13:16
  Speaker 4: Why is this so different?
• 13:18
  Speaker 2: Well, so, first of all, I tried to participate by
• 13:21
  Speaker 2: wearing in video green today. But importantly and video is
• 13:26
  Speaker 2: important because of how much they've added to the cumulative
• 13:29
  Speaker 2: earnings of this in p. Five hundred and the cumulative performance.
• 13:33
  Speaker 2: But I wouldn't necessarily extrapolate their success or sometimes setbacks
• 13:38
  Speaker 2: to the rest of the market because what they do
• 13:41
  Speaker 2: is very specific. The growth and data center construction and
• 13:45
  Speaker 2: data center a GPU fulfillment into those data centers is
• 13:52
  Speaker 2: a very small set of companies. I'd argue and Video
• 13:55
  Speaker 2: is sucking the oxygen for the room from a lot
• 13:58
  Speaker 2: of other technology companies. Well, we celebrate with Nvideo because
• 14:01
  Speaker 2: they've added so much value. I wouldn't necessarily read through
• 14:04
  Speaker 2: to the rest of technology.
• 14:07
  Speaker 4: Gilerya of DA Davidson, the relatively most bearish or perhaps
• 14:11
  Speaker 4: the least bullish name on the street, Thank you so
• 14:14
  Speaker 4: welcome to this special edition of Bloomberg Technology for our
• 14:18
  Speaker 4: TV and radio audience around the world. Joining me now
• 14:21
  Speaker 4: is Jensen One, CEO of Nvidia, straight from the analyst's call,
• 14:25
  Speaker 4: and Jensen, good evening, good afternoon to you. I think
• 14:29
  Speaker 4: the market wanted more on Blackwell, they wanted more specifics,
• 14:33
  Speaker 4: and I'm trying to go through all of the call
• 14:34
  Speaker 4: and the transcript. It seems like a very clearly this
• 14:38
  Speaker 4: was a production issue and not a fundamental design issue
• 14:41
  Speaker 4: with Blackwell, but the deployment in the real world, what
• 14:45
  Speaker 4: does that look like? Tangibly and is there a sort
• 14:47
  Speaker 4: of delay in the timeline of that deployment and thus
• 14:50
  Speaker 4: revenue from that product.
• 14:54
  Speaker 7: I let's see, that's just the fact that I was
• 15:00
  Speaker 7: so clear and it wasn't clear enough kind of tripped
• 15:03
  Speaker 7: me up there right away.
• 15:05
  Speaker 8: And so let's see, we made a mass change to
• 15:08
  Speaker 8: improve the yield. Functionality of Blackwell is wonderful. We're sampling
• 15:12
  Speaker 8: Blackwell all over the world today. We show people giving
• 15:18
  Speaker 8: tours to people of the Blackwall systems that we have
• 15:21
  Speaker 8: up and running. You could find pictures of Blackwall systems
• 15:25
  Speaker 8: all over the web. We have started volume production. Volume
• 15:30
  Speaker 8: production will ship in Q four. Q four, we will
• 15:34
  Speaker 8: have billions of dollars of Blackwell revenues and.
• 15:40
  Speaker 3: We will ramp from there. We will ramp from there.
• 15:43
  Speaker 8: The demand for Blackwell far exceeds its supply, of course
• 15:49
  Speaker 8: in the beginning, because the demand is so great, But
• 15:53
  Speaker 8: we're going to have lots and lots of supply and
• 15:56
  Speaker 8: we will be able to ramp. Starting in Q four,
• 16:00
  Speaker 8: we have billions of dollars of revenues and we'll ramp
• 16:03
  Speaker 8: from there into Q one, into Q two and two
• 16:05
  Speaker 8: next year. We're going to have a great next year
• 16:07
  Speaker 8: as well.
• 16:08
  Speaker 4: Jensen, what is the demand for accelerated computing beyond the
• 16:11
  Speaker 4: hyperscalers and meta.
• 16:15
  Speaker 8: Hyperscalers represent about forty five percent of our total data
• 16:20
  Speaker 8: center business. We're relatively diversified today. We have hyperscalers, we
• 16:25
  Speaker 8: have Internet service providers, we have sovereign AIS, we have
• 16:33
  Speaker 8: industries enterprises, so it's fairly fairly diversified. A site outside
• 16:39
  Speaker 8: of hyperscalers is the other fifty five percent. Now, the
• 16:45
  Speaker 8: application use across all of that, all of that data
• 16:49
  Speaker 8: center starts with accelerated computing. Accelerated computing does everything, of course,
• 16:56
  Speaker 8: from well the models the things that we know about,
• 17:00
  Speaker 8: which is generative AI, and that gets most of the attention.
• 17:05
  Speaker 8: But at the core we also do database processing, pre
• 17:10
  Speaker 8: and post processing of data before you use it for
• 17:15
  Speaker 8: generative AI, trans coding, scientific simulations, computer graphics of course,
• 17:21
  Speaker 8: image processing of course. And so there's tons of applications
• 17:25
  Speaker 8: that people use are accelerated computing for, and one of
• 17:30
  Speaker 8: them is generative AI. And so let's see what else
• 17:33
  Speaker 8: can I say?
• 17:35
  Speaker 3: I think that's the.
• 17:36
  Speaker 4: Lever jump in Jensen. Please on sovereign AI. You and
• 17:39
  Speaker 4: I've talked about that before and it was so interesting
• 17:42
  Speaker 4: to hear something behind it that in this fiscal year
• 17:46
  Speaker 4: there will be low double digit I think you said
• 17:48
  Speaker 4: billions of dollars in sovereign AI sales. But to the
• 17:51
  Speaker 4: lay person, what does that mean? It means deals with
• 17:54
  Speaker 4: specific governments, if so, where.
• 17:57
  Speaker 8: It's not necessarily Sometimes it's deals with particular regional service
• 18:04
  Speaker 8: provider that's been funded by the government. And oftentimes that's
• 18:08
  Speaker 8: the case in a case of in the case of Japan,
• 18:11
  Speaker 8: for example, the the Japanese government came out and UH
• 18:16
  Speaker 8: offered UH subsidies of a couple of billion dollars. I
• 18:21
  Speaker 8: think for several different internet companies and telcos to be
• 18:26
  Speaker 8: able to fund their AI infrastructure. UH India has a
• 18:32
  Speaker 8: sovereign AI initiative going and they're building their AI infrastructure. Canada,
• 18:39
  Speaker 8: the UK, France, Italy, missing somebody, Singapore, Malaysia, UH. You know,
• 18:48
  Speaker 8: a large number of countries are subsidizing their regional data
• 18:53
  Speaker 8: centers so that they could become able to build out
• 18:57
  Speaker 8: their AI infrastructure. They recognize that their countri's knowledge, their
• 19:03
  Speaker 8: countries data digital data is also their natural resource, not
• 19:09
  Speaker 8: just the land they're sitting on, not just the air
• 19:11
  Speaker 8: above them. But they realize now that their digital knowledge
• 19:15
  Speaker 8: is part of their natural and national resource, and they
• 19:19
  Speaker 8: are to harvest that and process that and transform it
• 19:23
  Speaker 8: into their national digital intelligence. And so this is what
• 19:29
  Speaker 8: we call sovereign AI. You could imagine almost every single
• 19:31
  Speaker 8: country in the world will eventually recognize this and build
• 19:36
  Speaker 8: out their AI infrastructure.
• 19:38
  Speaker 4: Jenson, you use the word resource, and that makes me
• 19:41
  Speaker 4: think about the energy requirements here. I think on the
• 19:44
  Speaker 4: cool you talk about how the next generation models will
• 19:47
  Speaker 4: have many orders of magnitude greater compute needs. But how
• 19:50
  Speaker 4: will the energy needs increase and what is the advantage
• 19:53
  Speaker 4: you feel in Vidia has in that sense, Well, the.
• 19:58
  Speaker 8: Most important thing that we do is is increase the
• 20:01
  Speaker 8: performance of and increase the performance and efficiency of our
• 20:05
  Speaker 8: next generation. So Blackwell is many times more performance than
• 20:10
  Speaker 8: Hopper at the same level of power used, and so
• 20:14
  Speaker 8: that's energy efficiency, more performance with the same amount of power,
• 20:18
  Speaker 8: or same performance at a lower power.
• 20:21
  Speaker 3: And that's number one.
• 20:22
  Speaker 8: And the second is using luca cooling. We support air
• 20:26
  Speaker 8: cool we support air cooling, we support liqual cooling, but
• 20:29
  Speaker 8: liqual cooling is a lot more energy efficient. And so
• 20:32
  Speaker 8: the combination of all of that, you're going to get
• 20:34
  Speaker 8: a pretty large, pretty large step up.
• 20:37
  Speaker 3: But the important thing to also.
• 20:39
  Speaker 8: Realize is that AI doesn't really care where it goes
• 20:41
  Speaker 8: to school, and so increasingly we're going to see AI
• 20:45
  Speaker 8: be trained somewhere else, have that model come back and
• 20:49
  Speaker 8: be used near the population, or even running on your
• 20:53
  Speaker 8: PC or your phone.
• 20:54
  Speaker 3: And so we're going.
• 20:55
  Speaker 8: To train large models, but the goal is not to
• 20:57
  Speaker 8: run the large models necessarily all the time. You can
• 21:00
  Speaker 8: surely do that for some of the premium services and
• 21:04
  Speaker 8: the very high value AIS, but it's very likely that
• 21:08
  Speaker 8: these large models would then help to train and teach
• 21:11
  Speaker 8: smaller models, and what we'll end up doing is have
• 21:14
  Speaker 8: one large, few large models that are able to train
• 21:18
  Speaker 8: a whole bunch of small models and they run everywhere.
• 21:22
  Speaker 4: Jensen, you explain clearly that demand to build generative AI
• 21:27
  Speaker 4: product on models or even at the GPU level is
• 21:31
  Speaker 4: greater than current supply. In black Vel's case in particular,
• 21:35
  Speaker 4: explain the supply dynamics to me for your products and
• 21:38
  Speaker 4: whether you see an improvement sequentially quarter on quarter or
• 21:41
  Speaker 4: at some point by the end of fiscal year into
• 21:43
  Speaker 4: next year.
• 21:45
  Speaker 8: Well, the fact that we're growing would suggest that our
• 21:48
  Speaker 8: supply is improving and our supply chain is quite large,
• 21:53
  Speaker 8: one of the largest supply chains in the world. We
• 21:56
  Speaker 8: have incredible partners and they're doing a great job supporting
• 22:00
  Speaker 8: us in our growth. As you know, we're one of
• 22:03
  Speaker 8: the fastest growing technology companies in history, and none of
• 22:07
  Speaker 8: that would have been possible without very strong demand but
• 22:10
  Speaker 8: also very strong supply. We're expecting Q three to have
• 22:13
  Speaker 8: more supply than Q two, We're expecting Q four to
• 22:15
  Speaker 8: have more supply than Q three, and we're expecting Q
• 22:17
  Speaker 8: one to have more supply than Q four, And so
• 22:19
  Speaker 8: I think our supply, our supply condition going into next
• 22:22
  Speaker 8: year will be in it will be a large improvement
• 22:25
  Speaker 8: over this last year with respect to demand. Blackwell is
• 22:31
  Speaker 8: just such a leap and there's several things that are happening,
• 22:35
  Speaker 8: you know, just the Foundation model makers themselves. The size
• 22:38
  Speaker 8: of the foundation models are growing from hundreds of billions
• 22:42
  Speaker 8: parameters to trillions of parameters. They're also learning more languages.
• 22:46
  Speaker 8: Instead of just learning human language, they're learning the language
• 22:49
  Speaker 8: of images and sounds and videos, and they're even learning
• 22:55
  Speaker 8: the language of three D graphics. And whenever they are
• 22:58
  Speaker 8: able to learn these languages, they can understand what they see,
• 23:01
  Speaker 8: but they can also generate what they're asked to generate,
• 23:05
  Speaker 8: and so they're learning the language of proteins and chemicals
• 23:08
  Speaker 8: and physics. You know, it could be fluids and it
• 23:11
  Speaker 8: could be particle physics, and so they're learning all kinds
• 23:14
  Speaker 8: of different languages, learn the meaning of what we call modalities,
• 23:18
  Speaker 8: but basically learning languages.
• 23:20
  Speaker 3: And so these models are growing in size.
• 23:24
  Speaker 8: They're learning from more data, and there are more model
• 23:27
  Speaker 8: makers then there was a year ago. And so the
• 23:30
  Speaker 8: number of model makers have grown substantially because of all
• 23:33
  Speaker 8: these different modalities. And so that's just one, just the
• 23:36
  Speaker 8: frontier model. The foundation model makers themselves haven't really grown tremendously.
• 23:41
  Speaker 8: And then the generative AI market has really diversified, you know,
• 23:46
  Speaker 8: beyond the Internet service makers to startups and now enterprises
• 23:50
  Speaker 8: are jumping in and different countries are jumping in, so
• 23:55
  Speaker 8: the demand is really growing.
• 23:57
  Speaker 4: Jensen, I'm sorry to cut you off. I will lose
• 23:59
  Speaker 4: your time soon. You've also diversified. And when I said
• 24:03
  Speaker 4: to our audience you were coming on, I got so
• 24:05
  Speaker 4: many questions. Probably the most common one is what is
• 24:09
  Speaker 4: in Nvidia. We talked about you as a systems vendor,
• 24:12
  Speaker 4: but so many points on in Vidia GPU cloud, and
• 24:16
  Speaker 4: I want to ask, finally, do you have plans to
• 24:19
  Speaker 4: become literally a cloud compute provider.
• 24:23
  Speaker 7: No.
• 24:25
  Speaker 8: Our GPU cloud was designed to be the best version
• 24:30
  Speaker 8: of Nvidia Cloud that's built within each cloud. Nvidio DGX
• 24:35
  Speaker 8: cloud is built inside GCP, inside Azure, inside AWS, inside OCI,
• 24:43
  Speaker 8: and so we build our clouds within THEIRS so that
• 24:46
  Speaker 8: we can implement our best version of our cloud, work
• 24:50
  Speaker 8: with them to make that cloud, that that infrastructure, that
• 24:54
  Speaker 8: AI infrastructure, in video infrastructure, as performance, as great TCO
• 24:58
  Speaker 8: as possible. And so that strategy has worked incredibly well.
• 25:05
  Speaker 8: And of course we are large consumers of it because
• 25:08
  Speaker 8: we create a lot of AI ourselves, because our chips
• 25:11
  Speaker 8: aren't possible to design without AI, our software is not
• 25:14
  Speaker 8: possible to write without AI, and so we use it
• 25:16
  Speaker 8: ourselves tremendous, you know, tremendous amount of it self driving cars,
• 25:19
  Speaker 8: the general robotics work that we're doing, the omniverse work
• 25:22
  Speaker 8: that we're doing. So we're using the DGX cloud for ourselves.
• 25:26
  Speaker 8: We also use it for an AI foundry. We make
• 25:30
  Speaker 8: AI models for companies that we'd like to have expertise
• 25:34
  Speaker 8: in doing so, and so we are AI.
• 25:36
  Speaker 3: We're an AI.
• 25:37
  Speaker 8: We're a foundry for AI like tsmcs a foundry for
• 25:40
  Speaker 8: our chips, and so there are three fundamental reasons why
• 25:44
  Speaker 8: we do it. One is to have the best version
• 25:47
  Speaker 8: of Nvidia inside all the clouds. Two because we're a
• 25:50
  Speaker 8: large consumer to ourselves, and third because we use it
• 25:53
  Speaker 8: for AI foundry for to help every other company.
• 25:56
  Speaker 4: Jensen one, CEO and Video. I want to thank you
• 25:59
  Speaker 4: for your time the extended conversation straight off the earnings call.
• 26:02
  Speaker 4: Thank you in videos earnings, Fiscal is good to see
• 26:06
  Speaker 4: you too. Fiscal second quarter done. I point out that
• 26:09
  Speaker 4: in after hours the stock is down still almost seven percent.
• 26:12
  Speaker 4: There'll be a lot of analysis to be done overnight
• 26:14
  Speaker 4: by the cell side, by the byside, and we'll bring
• 26:17
  Speaker 4: you the best from Bloomberg's reporters and editors from San Francisco.
• 26:20
  Speaker 4: This is Bloomberg Technology