Latest / Bloomberg Intelligence / Special Edition: Nvidia CEO Jensen Huang on Investing in South Korea

Transcript

Special Edition: Nvidia CEO Jensen Huang on Investing in South Korea · Bloomberg Intelligence

• 0:00
  Speaker 1: Bloomberg Audio Studios, Podcasts, radio News. I'm Monica Ricks in
• 0:08
  Speaker 1: the Bloomberg newsroom in New York with a special conversation.
• 0:12
  Speaker 1: Bloomberg Tech host ed Ludlow sat down within Video CEO
• 0:15
  Speaker 1: Jensen Huang to discuss the tech giant's latest investments in
• 0:19
  Speaker 1: South Korea, which includes teaming up with sk Group to
• 0:22
  Speaker 1: build new data centers and investing a billion dollars in
• 0:25
  Speaker 1: internet and cloud service provider Neighbor. Let's listen into a
• 0:29
  Speaker 1: portion of their conversation, coming not long after the Nvidia
• 0:32
  Speaker 1: CEO appeared at a Korean AI summit.
• 0:35
  Speaker 2: Jens and I think we just start with the basics,
• 0:37
  Speaker 2: like career is incredibly important to the AI built out globally,
• 0:42
  Speaker 2: we'll get intohigh bandwidth memory. But just from this summit,
• 0:45
  Speaker 2: from the president being here, what is the takeaway? What
• 0:47
  Speaker 2: is it you're trying to achieve?
• 0:49
  Speaker 3: Well, we're announcing a whole bunch of partnerships with them.
• 0:51
  Speaker 3: This is the golden ages for Korea. As you know,
• 0:54
  Speaker 3: there's semiconductor businesses booming, their industrial businesses booming. You know,
• 0:59
  Speaker 3: this is a country that has the ability to help
• 1:01
  Speaker 3: the world build out the AI infrastructure that they're incredibly
• 1:05
  Speaker 3: adapt at adopting new technologies, and it's a really technologically
• 1:11
  Speaker 3: forward leaning society and they love using AI. AI has
• 1:15
  Speaker 3: has really diffused throughout their society and and their industry,
• 1:19
  Speaker 3: and so so this is a great time for them.
• 1:22
  Speaker 3: We're announcing several things. We announce a big partnership with
• 1:27
  Speaker 3: SK Group where our companies are going to enter into
• 1:31
  Speaker 3: a business partnership where we do over five hundred billion
• 1:36
  Speaker 3: dollars of business with each other, whether it's consump consumption
• 1:39
  Speaker 3: and purchasing of memories or selling AI supercomputers to them
• 1:44
  Speaker 3: as they scale out to gigawatts of AI factories. There's
• 1:48
  Speaker 3: a whole bunch of other announcements. We're investing a billion
• 1:51
  Speaker 3: dollars in neighbor to help the they're the Korea's leading
• 1:55
  Speaker 3: AI cloud. They're going to scale up in Korea up
• 1:58
  Speaker 3: to two they're going to scale up to in mega
• 2:00
  Speaker 3: watts I think it is, and they're going to expand
• 2:03
  Speaker 3: across the world. And so we have a whole bunch
• 2:06
  Speaker 3: of announcements that we're making today.
• 2:07
  Speaker 2: With the expanded SK relationship. There's also sort of more
• 2:11
  Speaker 2: direct involvement with video on the roadmap for HBM future
• 2:14
  Speaker 2: generations of HBM talk about that. You know, I remember
• 2:19
  Speaker 2: you being on stage earlier this year saying five years
• 2:21
  Speaker 2: ago we told our supply chain what was going to happen,
• 2:24
  Speaker 2: and it did happen, and you gave some credit to
• 2:27
  Speaker 2: the memory makers in going with you on that journey.
• 2:30
  Speaker 2: But clearly you want to be involved in the direction
• 2:33
  Speaker 2: of travel for future generations of HBM.
• 2:36
  Speaker 3: Yeah, we're working together on of course, we started with
• 2:38
  Speaker 3: HBM two, worked on HBM three, three E, four, four E,
• 2:42
  Speaker 3: and then beyond, and so we've got a whole roadmap
• 2:46
  Speaker 3: of memories that we're working on together. It is also
• 2:49
  Speaker 3: the case that the semiconductor industry has really changed, and
• 2:53
  Speaker 3: the reason for that because we used to build computers
• 2:57
  Speaker 3: for people to use, and we're going to still continue
• 2:59
  Speaker 3: to build incredible computers. These are now eight processing ais
• 3:03
  Speaker 3: for humans to collaborate with. But in the future we
• 3:07
  Speaker 3: also have AI agents and robots and they're going to
• 3:10
  Speaker 3: be using computers. So instead of just a billion people
• 3:13
  Speaker 3: using computers, we're going to have one hundred billion agents
• 3:16
  Speaker 3: and billions of robots all using computers. The computer industry,
• 3:20
  Speaker 3: the chip that's built on top of the chip industry
• 3:23
  Speaker 3: surely is not big enough. And so this is one
• 3:25
  Speaker 3: of the realizations of the semiconductor industry that now computers
• 3:30
  Speaker 3: are built not just for people to use, but computers
• 3:32
  Speaker 3: are being built for computers to use. My guess is
• 3:36
  Speaker 3: that that the semiconductor industry is probably going to have
• 3:40
  Speaker 3: to be ten times larger than it is today over
• 3:43
  Speaker 3: the next decade or so. And so working with our
• 3:47
  Speaker 3: partners in Korea and around the world to scale up
• 3:50
  Speaker 3: the supply chain and the semiconductors so that we're prepared
• 3:52
  Speaker 3: for this AI future. It's really important.
• 3:54
  Speaker 2: I've had the obvious need to ask you about this
• 3:55
  Speaker 2: more than once this year, but how much do you
• 3:57
  Speaker 2: need the Korean economy to kind of get going to
• 4:00
  Speaker 2: increase the supply of HBM bits for Nvidia based systems
• 4:04
  Speaker 2: wherever they are.
• 4:05
  Speaker 3: Well, we don't have enough bits. We're constrained in HBO memories,
• 4:10
  Speaker 3: LPDD our memories. We're constrained, and just about every part
• 4:13
  Speaker 3: of the supply chain. We're even constrained now with land
• 4:16
  Speaker 3: and power and construction workers to set up the data centers.
• 4:20
  Speaker 3: I think this is one of the areas that is
• 4:23
  Speaker 3: going to make sure that we continued to build out
• 4:25
  Speaker 3: in a throttled way, you know, for a decade. And
• 4:28
  Speaker 3: the reason for that is because of these infrastructure Unlike electronics,
• 4:32
  Speaker 3: electronic devices like PCs and phones and things like that,
• 4:35
  Speaker 3: it's really really hard to scale up land power and
• 4:38
  Speaker 3: shell and so all of the supply chain just really
• 4:41
  Speaker 3: needs to get built out over the years. I think
• 4:43
  Speaker 3: we have the ability as an industry to double each year,
• 4:46
  Speaker 3: but we're going to have a hard time growing much faster.
• 4:48
  Speaker 2: In that the five hundred billions a number is large.
• 4:51
  Speaker 2: Would you just took a little bit more about what
• 4:53
  Speaker 2: it encompasses. We've gone over a lot on your commitment
• 4:55
  Speaker 2: to the US in terms of spending. Is that in
• 4:58
  Speaker 2: Vidia spending in the in economy or it's it's s
• 5:02
  Speaker 2: K fronting capital expenditures just a little bit more detail.
• 5:06
  Speaker 3: We're gonna we're gonna be purchasing memories from them for
• 5:09
  Speaker 3: many years to come. And as you know, we buy,
• 5:12
  Speaker 3: we build a lot of computers. In order to build
• 5:14
  Speaker 3: a trillion dollars worth of a roommate systems, you're gonna
• 5:17
  Speaker 3: have to buy a lot of system memories to go
• 5:18
  Speaker 3: with it. And so we have we have large purchase
• 5:22
  Speaker 3: agreements and large purchase intentions with s K Heinex. Meanwhile,
• 5:27
  Speaker 3: s K Telecom is going to become an AI cloud.
• 5:30
  Speaker 3: They're going, we're starting to build already, They're going they're
• 5:33
  Speaker 3: intending to build up to two gigawatts in the near future.
• 5:36
  Speaker 3: And uh and in that in that agreement, we will
• 5:40
  Speaker 3: be selling AI supercomputers to them. So between us, we're gonna,
• 5:43
  Speaker 3: we're gonna, we're gonna do half a trillion dollars worth
• 5:46
  Speaker 3: of business. Over half a trillion dollars worth a business.
• 5:48
  Speaker 2: I was able to sit down with SK Group chair
• 5:50
  Speaker 2: and JK One very recently for about forty minutes, and
• 5:54
  Speaker 2: at the end of the conversation we got to what
• 5:55
  Speaker 2: is the difference in approach, the academic difference and approach
• 5:58
  Speaker 2: on AI between the United States in China. And his
• 6:01
  Speaker 2: view on it was that China is very focused on
• 6:04
  Speaker 2: lowering the dollar PA token. In America were still focused
• 6:08
  Speaker 2: on the quality of tokens. I wonder what you think
• 6:11
  Speaker 2: of that.
• 6:13
  Speaker 3: The goal of the goal of AI is to produce
• 6:16
  Speaker 3: an intelligent smart answer. Now you could approach it in
• 6:20
  Speaker 3: a couple of different ways. You could, of course make
• 6:22
  Speaker 3: all of the token smart and smarter, and as a
• 6:24
  Speaker 3: result result in using less tokens to do so, you
• 6:27
  Speaker 3: could also produce AIS that are much more efficient, and
• 6:31
  Speaker 3: maybe you can think longer, explore more options and as
• 6:35
  Speaker 3: a result result as a result, produces a smart answer.
• 6:38
  Speaker 3: There are a couple, there are many different ways to
• 6:41
  Speaker 3: reach intelligence and deliver smart answers. In the end, really,
• 6:46
  Speaker 3: I think you have to take a step back and
• 6:47
  Speaker 3: just realize that both countries has extraordinary AI researchers and
• 6:52
  Speaker 3: whatever whatever conditions and whatever resources that they have, amazing
• 6:58
  Speaker 3: people will find great answers. And so you're going to
• 7:00
  Speaker 3: find You're going to you know, my expectation is that
• 7:03
  Speaker 3: China and the United States will continue to advance AI.
• 7:06
  Speaker 3: The conditions are different, the resources are different, their constraints
• 7:09
  Speaker 3: are different, but they're all they're you know, these amazing
• 7:11
  Speaker 3: researchers will find answers and and I think that that
• 7:16
  Speaker 3: in the case of China, they're producing more AI researchers
• 7:20
  Speaker 3: than probably all of the world has, you know, in
• 7:23
  Speaker 3: any given year, and so they're producing. If manufacturing intelligence
• 7:29
  Speaker 3: is important, they manufacture the most most important version of it,
• 7:32
  Speaker 3: which is the researchers. And so so this is a
• 7:35
  Speaker 3: this is an area a country that's going to produce
• 7:38
  Speaker 3: excellent AI technology. We how to keep and keep continue
• 7:42
  Speaker 3: to learn from them, work with them. As you know,
• 7:45
  Speaker 3: you're here in Silicon Valley, right here in San Francisco.
• 7:47
  Speaker 3: The number of AI researchers here that came from China,
• 7:50
  Speaker 3: that are Chinese is really quite significant. And so, uh,
• 7:53
  Speaker 3: you know, we're really fortunate to have them here, and
• 7:56
  Speaker 3: and uh, you know, we just got to keep on racing.
• 7:58
  Speaker 2: You made your first post on x I did and
• 8:02
  Speaker 2: you did so by sharing a letter signed by many
• 8:06
  Speaker 2: of your peers American companies to talk about the importance
• 8:10
  Speaker 2: of open models to America, to to the industry, to
• 8:14
  Speaker 2: the development of AI. And in the letter it's pretty
• 8:16
  Speaker 2: well explained. You know your rationale, But what what was
• 8:19
  Speaker 2: the catalyst for now? Why did you and saty Inodella
• 8:23
  Speaker 2: and others need to do that in this moment?
• 8:26
  Speaker 3: What we sense we sense that there's there's a growing
• 8:34
  Speaker 3: sentiment that that and the wrong sentiment for open models.
• 8:41
  Speaker 3: It's really important to realize that open models is essential
• 8:45
  Speaker 3: for safety. Open models is essential for security, for cybersecurity.
• 8:49
  Speaker 3: Open models are essential for innovation. It's and as a
• 8:51
  Speaker 3: server for startups. It's necessary for sovereignty company sovereignty. I
• 8:58
  Speaker 3: see a future where the world uses tons of closed models,
• 9:02
  Speaker 3: and I encourage everybody, including my company, to use open
• 9:06
  Speaker 3: AI and cloud and cursor and cognition and perplexity, use
• 9:12
  Speaker 3: everything that you can because it's out of the cloud,
• 9:16
  Speaker 3: because it's just easier, and you build only what you must,
• 9:21
  Speaker 3: and so in order to build what you must, you
• 9:23
  Speaker 3: need to have open models to do that with. And
• 9:25
  Speaker 3: the area is where we must. Maybe it's because we
• 9:27
  Speaker 3: have expertise that we simply cannot afford to share. This
• 9:31
  Speaker 3: is our company's alpha, our company's intelligence, and we have
• 9:34
  Speaker 3: to make sure we keep that proprietary. Maybe it's because
• 9:36
  Speaker 3: our company works in an industry that's regulated, and therefore
• 9:40
  Speaker 3: we simply can't pass along the service level agreement and
• 9:45
  Speaker 3: we have to make sure that we can deliver fully
• 9:47
  Speaker 3: on the service and the promise that we sign up for.
• 9:50
  Speaker 3: Maybe it's something to do with sovereignty that you simply
• 9:54
  Speaker 3: in a particular country you have to have your own AI,
• 9:57
  Speaker 3: you have to control your own AI. Whatever those reasons are,
• 10:00
  Speaker 3: it could be cost reasons, but I think I think
• 10:03
  Speaker 3: that largely I would recommend people build their own AIS,
• 10:07
  Speaker 3: especially when they need to control it for whatever reason.
• 10:10
  Speaker 3: And so I think the future is going to have
• 10:12
  Speaker 3: lots and lots of use of AI that's closed and
• 10:15
  Speaker 3: AI that's open that you can build your own AI. Now,
• 10:19
  Speaker 3: one of the things that people, you know, misunderstand about
• 10:24
  Speaker 3: these open models is yes, you can host it yourself,
• 10:28
  Speaker 3: but you can build your own computer, but most people
• 10:30
  Speaker 3: use computers in a cloud. Frankly, I think close models
• 10:34
  Speaker 3: are cheaper. You know, if you don't have to build yourself,
• 10:36
  Speaker 3: if you don't have to train it yourself. It costs
• 10:38
  Speaker 3: a lot of expertise to fine tune and maintain and
• 10:42
  Speaker 3: guard rail and keep it safe and evaluate it and
• 10:45
  Speaker 3: of course even build computers to host it. So there's
• 10:47
  Speaker 3: nothing cheap about doing that. The reason why you need
• 10:50
  Speaker 3: open open models is because you need to have control,
• 10:54
  Speaker 3: because you need to adapt something for your own very
• 10:57
  Speaker 3: specialized use cases. And so I I think there's a
• 11:00
  Speaker 3: lot of misunderstanding about about closed versus open. We felt
• 11:03
  Speaker 3: that it was important for people to understand that there's
• 11:06
  Speaker 3: a world for.
• 11:06
  Speaker 2: Both weighted versus open source as well. There is a distinction.
• 11:12
  Speaker 3: I open open weighted as much as much as open
• 11:16
  Speaker 3: as that you can the more open. It is in
• 11:19
  Speaker 3: the way that we work. We put the weights out there,
• 11:22
  Speaker 3: we also teach people how to train the model from
• 11:24
  Speaker 3: the data that we also open source. And the reason
• 11:28
  Speaker 3: that is we want to enable you to completely reproduce
• 11:32
  Speaker 3: the AI model that we've open weighted, and so that
• 11:35
  Speaker 3: ability by us teaching you how to do that, you
• 11:37
  Speaker 3: can then do it for yourself. You know, I think,
• 11:40
  Speaker 3: I think the the idea that the world is going
• 11:44
  Speaker 3: to be one or the other is just completely wrong,
• 11:46
  Speaker 3: and and the idea that open models is somehow unsafe
• 11:50
  Speaker 3: is also fundamentally wrong, and so we just want to
• 11:53
  Speaker 3: make sure that people understand.
• 11:54
  Speaker 2: To finish our conversation, you know, the two big case
• 11:57
  Speaker 2: studies where the release of can communicate three which on
• 12:00
  Speaker 2: an open way to basis releases fully July twenty seventh,
• 12:03
  Speaker 2: and then the case study of two open AI models
• 12:07
  Speaker 2: mistakenly accessing hugging faces systems and hugging Face trying to
• 12:12
  Speaker 2: use an open model in its defense where the guard
• 12:15
  Speaker 2: rails were a factor. Would you just reflect on those
• 12:18
  Speaker 2: two I know that you've been asked about them. They
• 12:21
  Speaker 2: seem to be like really big moments in AI.
• 12:22
  Speaker 3: Over those are perfectly perfect canonical examples. Just because something
• 12:26
  Speaker 3: is closed doesn't necessarily therefore make it safe or secure.
• 12:31
  Speaker 3: It is possible for a model to be jailbroken, as
• 12:35
  Speaker 3: possible for a model to be if you will, stolen.
• 12:37
  Speaker 3: It could be possible that that somehow has leaked from
• 12:40
  Speaker 3: the inside. It's possible that the guardrails or the sandboxes
• 12:44
  Speaker 3: of an AI closed AI model wasn't properly engineered, and
• 12:48
  Speaker 3: as a result, it was able to attack another company
• 12:54
  Speaker 3: in some way. And so just because something is closed,
• 12:57
  Speaker 3: and just because something is proprietary, doesn't necessary really make
• 13:00
  Speaker 3: it secure and safe. Of course, thank goodness, we have
• 13:04
  Speaker 3: two companies, well, I guess more than that several companies
• 13:08
  Speaker 3: that build closed AI models, and these are extraordinary technology companies,
• 13:12
  Speaker 3: and they're doing their best to keep it safe and
• 13:14
  Speaker 3: keep it secure. But it is also the cainetical case
• 13:18
  Speaker 3: that single points of failure is where we have the
• 13:22
  Speaker 3: greatest vulnerability. We cannot have single points of failure as
• 13:25
  Speaker 3: an industry, as a world, we should have distributed, massively
• 13:30
  Speaker 3: distributed self defense. And so in the case of in
• 13:33
  Speaker 3: the case of the example you just mentioned, hugging Face
• 13:36
  Speaker 3: thankfully was able to access an open model and I
• 13:40
  Speaker 3: think they used GLM five point two. As my understanding,
• 13:44
  Speaker 3: they couldn't get a proprietary model, they could not get
• 13:47
  Speaker 3: a closed model to help them figure out what happened.
• 13:51
  Speaker 3: But this is exactly the reason why you want to
• 13:54
  Speaker 3: have open models, because in that case, they use GLM
• 13:56
  Speaker 3: five point two to identify where the vulnerability was, where
• 14:00
  Speaker 3: the penetration was, and we're able to quickly identify them
• 14:03
  Speaker 3: and patch it up. And so this is a perfect
• 14:06
  Speaker 3: example of self defense that's necessary, as a perfect example
• 14:11
  Speaker 3: of diversity of AI technology being necessary, and so a
• 14:15
  Speaker 3: perfect example of why open models and open capabilities for
• 14:19
  Speaker 3: self defense is really important.
• 14:22
  Speaker 1: That's in Nvidia CEO Jensen Huang in a special conversation
• 14:25
  Speaker 1: with Bloomberg Tech host ed Ludlow. You can watch the
• 14:28
  Speaker 1: full interview now at Bloomberg dot com, slash Videos and
• 14:31
  Speaker 1: on the Bloomberg Business app. You can also listen by
• 14:34
  Speaker 1: subscribing to the Bloomberg Tech podcast feed. I'm Monica Ricks,
• 14:38
  Speaker 1: thanks for listening. This is Bloomberg