- 37 minuti fa
Categoria
📺
TVTrascrizione
00:04I'm at Nvidia's GTC conference here in Berlin and I have none other than Jensen Huang, the CEO of Nvidia
00:10with me.
00:10Jensen, thank you so much for making the time, you've had a busy morning.
00:13We're here in Europe, you've been travelling around London, Paris, all of these different cities around Europe.
00:19You've announced around $10 billion worth of investments across the continent into around 14 startups.
00:24What are the thematics behind those investments? What's getting you excited?
00:28Well, we're trying to create, kickstart the AI economy in Europe.
00:33Of course, you know, Europe has a tremendous amount of technology, tremendous amount of industrial know-how, incredible science and
00:42technology.
00:43And it's ready to engage and build the AI economy here.
00:47In order to build the AI economy, you need to invest in all five layers of that industry.
00:52Energy, chips, infrastructure, the models, and of course, all of the AI applications.
01:00Each one of the industries, every single company here in Europe is going to become an AI company someday.
01:05I'm thinking that in five years' time, whether you're Media Mark here in Germany, Mercedes, or Siemens, or Merck, all
01:15of these companies are all going to HSBC, they're all going to become AI companies.
01:21And so we're just accelerating that journey, a journey that we started a couple of years ago, and it's really
01:27picking up.
01:28Europe, like many other places in the world, has several challenges, whether it's access to capital, whether it's energy prices.
01:34There's often this discussion around, you know, who is strong in AI, the U.S. or China.
01:39And there's this view that Europe kind of gets sidelined in that discussion.
01:41Does Europe have what it takes, given some of those challenges, to compete with what we're seeing out of China,
01:47what we're seeing out of the U.S. in terms of AI capability, companies?
01:51Absolutely.
01:52The most important thing about leadership in AI is actually the application of the technology in your industry.
02:00And so whoever has whatever region, whatever country diffuses AI most and quickly will ultimately win.
02:09And the reason for that is because you want to lift the entire industry, every single industry, with AI, not
02:16just create a new industry of a few companies.
02:18And so you want every part of your economy to be accelerated by AI.
02:23Jensen, the market always loves to hear what you have to say about the state of play of the industry
02:27across the board.
02:28And at the moment, there feels to be a bit of nervousness in the market, and particularly around neoclouds.
02:34We've seen, you know, N-Scale looking at investors selling it to reduce its valuation, firmness pulling its IPO, concerns
02:40over the debt levels of some of these neoclouds.
02:42Is there a problem right now with the neocloud business model, or are these just teething issues?
02:48If you look at the fundamentals, the demand for the neoclouds is off the charts.
02:54Their capabilities are extraordinary.
02:57And so what is missing in the middle is the working capital to fuel that growth, to fund that growth.
03:03And because there's just so much demand across so many sectors of AI, there's a scarcity of capital.
03:11The demand of the capital is much greater than the availability of the capital.
03:15That's going to get solved.
03:16And that's going to get solved because fundamentally, the most important things are the values that these guys are providing,
03:23these companies are providing.
03:25The demand for it is so strong.
03:26Just given the number of players in this space, there's dozens of companies kind of described as neoclouds.
03:32Do you expect any kind of consolidation and M&A to take place in this particular sector of the industry?
03:38Ultimately, they need scale.
03:40They need regional excellence.
03:42Like, for example, here in Europe, the energy is scarce.
03:46It's scarce everywhere.
03:47But you want to allocate the energy to your own companies.
03:51And in the case of the U.K., they want to allocate their energy to N-Scale.
03:58In the case of France, they want to allocate their energy to Mistral.
04:02Around the rest of Europe, allocate your energy to Nebius.
04:06There's a fundamental reason why you want to make sure that your natural resources are dedicated, allocated to your own
04:14companies first.
04:15If the neoclouds in Europe are able to establish an infrastructure here, then you can build the AI ecosystem, the
04:24AI economy on top of them.
04:26And then the flywheel of the AI economy will start in Europe.
04:31It will terminate in Europe.
04:33You can bring investments into Europe.
04:35But you have to start with having your own infrastructure.
04:37Do you think that scale comes from some form of consolidation to grow those big plays?
04:42Right now, you have so much demand.
04:44I think you have more than several companies you could start up.
04:48Of course, each one of the countries probably don't need more than a couple.
04:55But across all of Europe, you're going to need a lot more neoclouds.
04:59The working capital part of the equation is an interesting one, Jensen.
05:03Because obviously earlier this year, you, along with a number of financial players, announced this sort of $500 billion financing
05:08mechanism.
05:09What is the demand at the moment for that?
05:11Is that something that can be tapped as these neoclouds look to scale?
05:16Well, that's ultimately what we need.
05:18One of the things that we're trying to do is to establish NVIDIA's compute as a financeable asset.
05:26The residual value of our compute, because it's fungible, it's so productive, it's so fungible, and it's so durable, it
05:33is now becoming an asset class.
05:35The more that the institutions recognize this, the more that the lenders recognize this, the more the capital will become
05:44available to the neoclouds.
05:46And so we're working to establish that NVIDIA compute as really a fungible, productive, durable, financeable asset.
05:56And I think that's going to be a game changer.
05:58And Jensen, what's your take on the current health of the Frontier Labs at this point?
06:02There was a lot made about Anthropix Q3 numbers, 45% growth.
06:06A lot of people saying, well, this is a significant slowdown from what we've seen.
06:09And at the same time, both OpenAI and Anthropix are aiming for these very highly valued IPOs.
06:15Just wanted to get your take on whether you're worried about any kind of slowdown at the Frontier Labs,
06:20and are they going to be able to justify those valuations in the end?
06:24We're looking at two of the most consequential technology companies in history.
06:28Incredible management team, incredible technical skills, incredible technical capability.
06:33They're also the two fastest growing technology companies, or any company, in human history.
06:40And so we could nitpick around the corners.
06:43I don't think it's, I don't think, I think it's maybe interesting, but it doesn't change the fundamentals.
06:50We're talking about AI, which is the most consequential technology in history,
06:55two of the most consequential technology companies in history,
06:58incredibly well built, and also the fastest growing ever.
07:02And so I think that's all that matters.
07:05Jensen, I know you've been talking a lot about safety and around social and societal impacts of this technology.
07:11One of the most interesting discussions has been around how do AI companies do good for society?
07:16And Anthropix came out and said, hey, we're going to commit 5% of our shares to a new public
07:20benefit fund,
07:21something that Elon Musk backed and said we're going to give kids in America some SpaceX shares.
07:26Have you considered any kind of initiative like this at NVIDIA?
07:30We do all kinds of things for social good, and everybody has their own way of doing it.
07:36The most important thing, you talked about AI safety.
07:40The most important thing that we need to do as an industry is to create an optimistic future, but do
07:46it responsibly.
07:47This technology has the potential to do incredible good.
07:52Whatever industry you choose, whatever economy you choose, whatever country you choose,
07:57and there is an extraordinarily optimistic and exciting future, but we have to build this technology responsibly.
08:05One of the most important things that we've done is created this technology called OpenShell.
08:09Anthropix is a very early adopter of it, very enthusiastic collaborator, really appreciate their collaboration.
08:16We all want to see this technology advance in the safest possible way, and we have the technology and the
08:23responsibility to do so.
08:25Yeah, I want to put on OpenShell in a second, Jensen, but just in terms of this kind of broader
08:29discussion about how to give back,
08:31do you think this sort of direct equity method is a good path to go down?
08:35The best way for those companies to give back is to contribute in the way that they fundamentally contribute.
08:42Elon's best way to give back is to inspire a whole generation to build things, heavy industries, launching rockets to
08:52space.
08:53What could be more consequential and inspirational than that way of giving back?
08:59Anthropix's way of giving back, number one, is to advance this technology, invent this technology, and bring it to the
09:05world safely.
09:05The way NVIDIA gives back is all of the things that we're doing here, supporting the global economy,
09:12helping every industry thrive, new companies to be created, invest in inspirational entrepreneurship.
09:19That's our way of giving back.
09:21The best way to give back is to help the economies grow.
09:25The best way to give back is to create new economies.
09:28Jensen, your view on the safety issue really is that if a company believes that a model is dangerous,
09:34they should just not release it, right?
09:35There doesn't need to be some kind of new broad regulation, et cetera.
09:38How can the general public trust the leaders of these companies to make that right decision?
09:48Because they're incentivized to do so.
09:51Regulations take a while to come.
09:54If there are concerns that exist today, it is your responsibility to take accountability and do something about it now,
10:05instead of waiting for laws and regulations to happen, which takes time, as you know.
10:09And so the reason why I trust them to do so is, one, I believe in the leadership.
10:15Two, I believe that the companies are incentivized to do the right things because there's no such thing as a
10:21good product that's not safe.
10:23And so it is absolutely in their perfectly aligned incentives for their company to advance the technology as responsibly as
10:32necessary.
10:33And part of the solution you have is open shell, right?
10:35Is being able to monitor these models because...
10:37To contain and monitor.
10:39Yeah.
10:39Because one of the issues we've seen is these models doing things, they're not supposed to.
10:43And so how do you control that?
10:45To contain it and to monitor it.
10:47And just continuously, look, here's what it's doing, here's what it's accessing.
10:51It's exactly the same way you think about a baby in a playpen.
10:55You contain the baby and you monitor it.
10:58Now, the technology, of course, is very sophisticated.
11:01And there's active monitoring and active controls that are part of this containment system.
11:06It's as if it's like a super smart force field around AI.
11:11And then you have multiple layers of it.
11:14And so it's not as if one layer is the end of the story.
11:27And you're confident this is going to stop some of the issues we've seen, such as these unintended hacks and
11:33accesses to websites, et cetera.
11:35We're confident that it would have handled the incidents we've seen.
11:38And we don't know what we don't know about the future.
11:40But the technology was built on great first principles.
11:44We also have the advantage of over 100 companies around the world,
11:47including Anthropik and OpenAI and SpaceX and Meta and Google and Microsoft and IBM and Red Hat and, you know,
11:57CrowdStrike and Palitone Networks.
11:59The world's leading operating system companies, cybersecurity companies and Frontier Lab companies are all part of this OpenShell initiative, contributing
12:08to it.
12:08And so we've got the best minds in the world working on this one concept, one issue, one initiative.
12:15And I can't imagine a more powerful coalition of bright minds around the world who want the same thing.
12:22Jansen, we've got a few minutes left, so I just want to get your take also on another big topic,
12:26which of course is China.
12:26A lot of focus on some of the innovation coming out of that market.
12:29For NVIDIA specifically, do you believe now this market is effectively closed for your chips?
12:37Well, no market is closed forever.
12:39And China wants foreign companies to invest in China, just as Germany wants foreign countries, companies and countries to invest
12:48in Germany.
12:49And so the market has to be open and has to be welcoming.
12:52I mean, there's no reason to believe that that market is going to be closed to the best competitors in
12:58the world.
12:59And NVIDIA is surely the best in the world doing what we do.
13:03And we have a lot to offer and we're going to go compete in China and around the world.
13:07And so to the extent that that market could be open to us, we'd really appreciate it.
13:12Otherwise, we're going to continue to be here and we'll keep looking for opportunities.
13:15There's been some really interesting strides made in the semiconductor industry in China in all different facets.
13:20CXMT, the memory company, is reportedly developing HBM3E now.
13:25Is there a world in which you could partner with CXMT, use their HBM3E and some of your older products
13:31to regain access to the market in a bigger way?
13:34Is that something you've considered?
13:35Of course, we're open-minded.
13:36We work with Chinese companies.
13:38Just about the whole world works with Chinese companies in the supply chain somewhere.
13:42And whether it's in power generation or cooling technology or solar technology or battery technology,
13:49China is incredible at the fundamental infrastructure, fundamental infrastructure technology.
13:55And no country, collectively, is better at manufacturing than China.
14:00And so we have to take their capabilities very seriously.
14:03And just in terms of the strides the Chinese semiconductor industry has made,
14:07how would you assess those given what we're seeing out of companies like Huawei?
14:10Well, we've definitely accelerated, because of all the regulations,
14:14we've accelerated their internal capabilities probably in 10 years and made it five.
14:21And so we've definitely accelerated China's initiatives,
14:26its nationalization of, if you will, semiconductor manufacturing capabilities,
14:33and all of the industrial support that has been put into that sector.
14:37And so there's no question we've accelerated that.
14:41But nonetheless, NVIDIA's American technology companies are mighty in this area.
14:47We're ready for competition from all over the place.
14:50And so we can take on competition from anybody.
14:53And just as you look at your supply chain, memory, of course, is one of the key bottlenecks,
14:58as there are looked bottlenecks across the supply chain.
15:00Well, one of the most challenging bottlenecks is substrates.
15:03Yeah, this week we announced that we're expanding our substrate capacity as large as it already is by two and
15:11a half times more in the next three years.
15:15And so we're expanding our substrate capacity as one of our most challenging limiters today.
15:20We're going to increase that by two and a half times.
15:27And so we're expanding our co-host packaging technology.
15:32And our work there with Amcor is going to increase our capacity tremendously.
15:37And so we've, this week, announced precisable expansion of our supply chain.
15:42And just on that memory story as well, Jensen, do you, there was a column, a columnist in South Korea
15:47who said, well, there could be an oversupply of memory in 2029.
15:51Are you seeing any signs of that?
15:53I don't think so.
15:54So, you know, one of the things that's really great about memory is that you could always use more of
15:59it.
16:00And, of course, through architecture ideas, and NVIDIA owns the architecture of our systems, through architecture ideas, we could shift
16:09around the resources as necessary.
16:12But HBM memory will be used for years to come.
16:16And LPDDR, we're the first company in the world to put LPDDR memory, what goes into your cell phones, into
16:23the world's data centers.
16:24And it's enabled us to deliver performance that is twice as energy efficient as anything in the world today.
16:31And so we're going to continue to use abundant amounts of that as well.
16:35Thank you.
16:35Jensen, it's been a pleasure speaking to you.
16:37Thank you so much for taking the time to speak to me.
16:40What a great event.
16:41What's going on in Germany, and don't, if I could just have a second more, this week we announced the
16:47expansion of our supply chain.
16:48We announced DGX Pod, which accelerates the build-out of the AI factories.
16:55We also announced OpenShell.
16:57I appreciate you giving me the chance to talk about AI safety.
17:01And then we announced all of our initiatives here, particularly in Germany, around physical AI, so robotics and cars and
17:07telecommunications.
17:08Great, Jensen, thank you for all of your insights.
17:10It's been a pleasure to speak to you.
17:11You've had a crazy schedule here.
17:13That was Jensen Hong, the CEO of NVIDIA.
17:15Jensen Hong, the CEO of NVIDIA.
17:16Jensen Hong, the CEO of NVIDIA.
17:20Jensen Hong, the CEO of NVIDIA, the CEO of NVIDIA, the CEO of NVIDIA, the CEO of NVIDIA, the CEO
17:20of NVIDIA, the CEO of NVIDIA, the CEO of NVIDIA, the CEO of NVIDIA, the CEO of NVIDIA, the CEO
17:20of NVIDIA, the CEO of NVIDIA, the CEO of NVIDIA, the CEO of NVIDIA, the CEO of NVIDIA, the CEO
17:20of NVIDIA, the CEO of NVIDIA, the CEO of NVIDIA, the CEO of NVIDIA, the CEO of NVIDIA, the CEO
17:20of NVIDIA, the CEO of NVIDIA.
17:20Grazie a tutti.