I don't think I really need to introduce Jensen Huang.
But a little backstory.
Jensen founded Nvidia in 1993.
He has served as its CEO since then,
leading the company to become a global leader
in accelerated computing.
Its invention of the GPU in 1999,
revolutionized computer graphics,
sparked the growth of PC gaming,
and now it's basically shaping the modern era
of artificial intelligence.
Let's give a warm welcome to Jensen Huang.
[audience cheers and applauds]
Hi. Hi, Jensen.
Can you hear me?
Hey Lauren. I wish I was there.
I wish you were here too,
because it's really hard to get the,
it's hard to get the vibes of that
leather jacket over Zoom.
I could do it.
[Lauren laughs]
Alright, Jensen, you are joining us today from Thailand,
and I know that you've been on
a bit of a world tour recently,
talking about Sovereign AI,
which is I think what you're doing in Thailand, correct?
Tell us about this.
Well, first of all, it's great to be with you.
I'm here in Bangkok.
In fact, it's really great to be back here,
because you probably don't know this,
but I lived here for about five years when I was a child,
and from about, from about five or so.
And I went to school,
an international school here.
It was called Rudi Internationals.
It's a couple of blocks away.
And so, it's really, it's really terrific to be here.
Jensen, I'll also note you're the only guest
who's technically joining us from the future,
because it's already December 4th there.
[Lauren and the audience laugh]
I think this is very fitting.
It it is. It is.
And I could, I could provide some insight
and some, you know, some warnings,
and keep you out harm's way.
Excellent. We need that.
Okay, so tell us about this.
Tell us about Sovereign AI.
Yeah, so what's going on is,
is countries are awakened to
the incredible capabilities of AI,
and the importance of AI for their own nations.
They realize that their data
is part of their natural resource.
Their data, of course, encodes their,
their society's knowledge and culture,
and common sense, and, you know,
their hopes and dreams.
And, they now realize that they really
should take a role in processing that data,
harvesting that data,
and refining it into intelligence,
and providing their native intelligence to their society.
And so that's one observation.
The other observation is that,
that people are starting to realize
that AI in a lot of ways is like
the energy infrastructure,
the communications infrastructure,
and now there's going to be an intelligence,
a digital intelligence infrastructure.
And these AI factories,
these data centers are part of that
infrastructure that they have to be mindful about,
and play a role in forming.
And this infrastructure is necessary for,
of course, their industry,
but it's also really important for their society,
and for their education, for their colleges,
for research and startups and so on and so forth.
And so I'm here in Thailand to celebrate
the partners that we have here.
We have an AI cloud company here
called SIAM AI that is starting.
We have some 56 or so AI startups
that I'm gonna be meeting,
I'll be meeting today,
a whole bunch of Professors- Just today, 56?
Yeah. Is there like a Jensen AI
that you deployed?
I mean, how do you fit those, how do you,
literally, how do you fit that many meetings in?
Lauren, you just gotta go fast, you know.
[audience laughs] You just do things fast.
[Lauren laughs] It's like our computers,
you gotta do things fast.
And so they're professors that I'm meeting,
there are companies that I'm meeting,
and of course had the benefit
of meeting the prime minister,
and she's really, really impressive.
And she's 38 years old, and leading our,
leading this country.
And really believes in technology,
and wants the country to lean forward in AI.
And so it's really terrific to be here.
The way that you're,
the way that you're describing this is,
I mean, I'm sure it's nice if you
get to sell more GPUs as well, but
it also sounds like you're, you know,
reclassifying this era of generative AI as infrastructure.
And I'm wondering what that means
for the development of AI models.
Well, let's see.
There are a couple of questions in there.
First of all, it is infrastructure in the sense that,
that so many different parts of society needs it.
You know, universities need it, researchers need it,
startup companies need it, large companies need it.
And, and so when every aspect
of a society needs something,
it's generally recognized as infrastructure.
And I think that artificial intelligence
will be a layer above the internet,
a layer above, above the way
we interact with computers today.
And, and your question about large language model
is it's essentially the new operating system,
you know, whereas the way that we used to use computers
is to program it and retrieve files,
and manage files, and so on, so forth in the past,
in the future, the way we'll use computers,
of course, is to, to prompt it,
to ask it a question,
to ask it to do something for us.
And at the core of that,
instead of the operating systems of the past,
the core of that is now large language models.
It's not one large language model,
but it's a system of large language models.
And, but at the core of it is the large language model.
Your question about, about large language models here
and around the world...
For example, I was, I was in Denmark,
and they're building,
they're building the large language model of Denmark.
And I was in Sweden,
and they're building the large language model of Sweden.
And I was in Japan,
they're building the large language model of Japan.
And of course Indonesia is building it.
India is building several.
And here in Thailand,
it's called ThaiGPT.
And so, there are these language models
that are, that are being created.
I think the availability of understanding
of how to build these large language models
is fairly well documented,
and the systems are available,
and people are doing a really good job.
People are also realizing that,
that it's not one giant model that needs to be built,
but in a lot of ways,
you could create very effective AIs
that are systems of large language models,
And the systems of large language models.
of course, some of it's related to reasoning,
some of it is related to tool use,
some of it's related to retrieval of information,
and guard railing, and you know,
the synthetic data generation
for reward models and reflection models,
and, you know, all kinds of different,
different types of models are involved in a,
in a system of AIs,
and the knowledge of how to do that,
and the frameworks and tools
of how to do that that is widely available.
And so I think every country should be able to do it.
Fascinating. One of the things
that you've been talking about lately,
and I think it's it's broadly
kind of a buzzword right now in AI, is agents.
You spoke about this a little,
a little bit on your most recent earnings call too.
I think this is probably one of those things
where people are like maybe even
saying it in conversation and not fully sure what it means.
So like, what is an AI agent to you?
What does it, what does it do?
And why are some folks saying that this is the next,
this is the next wave of generative AI?
Simple. So the first, the first,
the first wave was the development of-
[Lauren] And you don't have to go back to like
1993 by the way.
We're on a, we're on a timeframe here.
[audience laughs]
Okay. Alright. Alright.
Okay, 2012. The first wave started around-
[indistinct] Yeah, yeah, yeah.
And that first wave was perception AI, right?
Remember all of a sudden we,
we were good at computer vision,
and speed recognition,
and all kinds of perception related models.
The second generation with generative AI,
and that was really about three,
four or five years ago,
of course, GANs were the beginning of that,
GPT is an example of that,
diffusion models, so on and so forth.
Generative AI. Now agentic AI,
which is where we are today,
is really the combination of all that.
And it's really the fundamentals of intelligence,
which is perception, reasoning and planning.
And if you could say, if I replace planning with,
with generation, perception, of course, perception AI.
So really what's missing is the ability to do reasoning.
And we now have models that are,
are doing a better job with reasoning.
You could give it a basic mission,
and break it down to step-by-steps.
It could be a chain of thoughts,
it could be a chain of trees,
it could be all kinds of different ways
of coming up with a reason through a problem.
Maybe even inviting other,
inviting other Ais to help you do a job.
For example, an AI that could be good
at generating images,
or good at generating music,
or retrieving a file, you know,
generating citations, doing search,
so on and so forth so that you could,
you now have a system of AI's
that together are essentially agentic,
or otherwise robotic, or otherwise intelligent.
And so I think those words are somewhat interchangeable,
depending on their context.
But basically it's really about
perception, reasoning and planning.
And the idea too is that we're shifting from this world
where we used computers to do work,
but now the computers are doing the work for us.
You could use an agent to perform a task
for you on your computer, right?
Thus freeing up time.
Are you personally using AI agents
in that way right now?
Like, could you describe an example?
I use AI, I use several AIs,
and I use 'em all just because I enjoy them all.
I use Gemini, I use ChatGPT.
But is it actually doing your work
for you at this stage?
I use it to write a lot.
I give it a, I give it a basic outline,
give it some PDFs of my previous talks.
And I give it, get it to write my first draft.
It's really fantastic.
So, so anytime I write you an email,
and you actually respond, that's an agent,
you're saying, Jensen? [audience laughs]
No. Well, I have a filter,
so when Lauren writes me, I write back.
It goes right to the top.
Goes right to the top. Alright.
Let's talk about the future of NVIDIA's business.
You have a lot of partners around the world
as you're describing,
and I'm wondering how vulnerable, if at all,
you're feeling to political instability right now?
How are you thinking about this?
Well, the world is changing.
The world is changing, but, and, but,
but also simultaneously,
the fundamental technology trends are so incredible.
And so in our, in our industry, in our business,
this fundamental trend of artificial intelligence,
this new layer of computing,
this reinvention of the computer,
a reset of computing as we know it
over the last 60 years,
is just an extraordinary opportunity.
And the type of problems we're able to solve now,
the type of, the type of products and solutions,
and capabilities we can offer to the world
is just extraordinary.
And I think that's one of the reasons why
artificial intelligence is so central
to your conference today.
And so I think we all recognize that.
That's probably, that fundamental force,
that dynamic, that tailwind, if you will,
is just so incredibly strong,
that the technology industry
and all of the software industries that sit on top of it,
and all the industries that sit on top of that,
are gonna go through some extraordinary
transformations over the next decade.
Hmm. To specifically drill down to politics, though,
I'm thinking about something like,
for example, this Monday,
the US Department of Commerce
just expanded its export controls.
And that's something that has
the potential to affect NVIDIA's business
because the export controls apply
not just to semiconductors,
but also to more broadly, more componentry,
and that goes into your GPUs.
So, you know, when you look at something like that,
do you think the export controls are,
the reasons for them are justified?
How concerned are you about China, for example,
you know, rising up in AI,
and the competition that China poses?
Yeah, several questions there.
I think first of all, our job is to,
to do our best to inform, to address,
and make aware the dynamics of our industry,
and how it works,
and how our company works in
various markets around the world,
do our best to explain those things to the administration.
And it's up to them to come up
with the best po possible policies.
And then after that, our job is,
is to continue to do our best to innovate,
and drive our technology forward,
support our customers, and continue to be successful.
And so, so, you know,
those are things within our control,
and I think that our interactions
with the administration has been great.
In terms of the incoming administration,
president-elect, why are you guys laughing?
[audience laughs]
President-elect Donald Trump has said
in the recent past that he believes that
Taiwan has taken some of the chip business from us.
You work very closely with Partner TSMC based in Taiwan.
When you hear remarks like that, I mean,
how concerned are you about what that
could mean for your relationship with TSMC?
Well, TSMC, TSMC's role in the industry,
and role in the world is fairly unquestionable.
I I think they're,
they're obviously a very important technology innovator,
a very critical part of the world's supply chain,
an extraordinary company with excellent leadership,
and a culture that that was really built by hand
over the course of some 40 years,
and a relationship and a partnership
that is really critical to us.
And not just to us,
to companies that that everybody knows all over the world.
I think all of that is gonna remain true.
And if anything, you know,
TSMC as a company gets better and better
every single year that we work with 'em.
And, and so, so our relationship
with TSMC is gonna continue,
and our, the world's dependency on TSMC
is going to continue for quite a long time to come.
Have you spoken yet with
President-elect Donald Trump since the election?
I've reached out to President Trump
and congratulated him, and wished him success,
and told him that we'll do everything we can
to help the administration succeed.
Okay. Do you anticipate having
extended conversations about your business with him?
Well, artificial intelligence is
obviously really important.
[Lauren] Huge topic. And yeah, huge topic.
And it's important on a whole bunch of dimensions.
One dimension, just tech technological reasons
why it's important,
but it's also important because it's a brand new industry.
This is an industry that's never existed before.
You know, we're producing intelligence, and manufacturing,
manufacturing intelligence at scale for the very first time.
And the reason why I chose that word,
manufacturing is, is because it, usually,
an industry is formed,
when it manufactures something that didn't exist before.
And so, for example,
my generation was responsible for the,
for the manufacturing at scale of software.
Before my generation, there was no such thing as software.
And then after this generation of technologists,
and all of us working in the computer industry,
there's now a concept called software
that you could manufacture.
There's common sense about manufacturing at best practices
and all the technology necessary manufacture software.
And, and we do it at an incredible scale,
and a it's multi-trillion dollar industry.
Well, we have a brand new industry now,
and this industry is not about producing software,
it's about producing intelligence.
And this industry requires energy,
and requires a lot of factories,
and has great implications to
great social impact, industrial impact,
and the economic and technological success of a country.
And so I'm certain that
the new administration and President Trump
would be quite interested in this industry,
and I'll be more than delighted to provide any support,
and answer any questions that I can.
There's also been a lot of excitement about Blackwell,
which is your newest chip.
It is being delivered right now, right?
You've delivered it to some of the hyperscalers, correct.
Blackwell is a, is a whole system.
Yes. A bunch of switches,
and networking, and computing, and so and so forth,
and a whole mountain of software,
and it is in full production,
it's going quite well.
And we have Blackwell systems
installed all over the world now.
And what would you say is the,
like the number one biggest difference
from Blackwell that your customers are going to see?
I mean, it's assumed that it's faster,
but what are the, what are the downstream effects of that?
There are several things.
First, because in training,
because of its performance is so incredible,
it's a leap in performance.
Another way of thinking about that,
is instead of waiting months
to process the data necessary to train a model
and then train a model,
you could reduce that time,
you know, by, by three x, by four x.
And so it could have been six months,
and now it's month and a half.
And with everybody racing to the
next plateau of AI capability,
the difference between three months
from one company to another company could be a game changer.
And so everybody is really racing
to reach that next level.
In inference, this is one of the extraordinary things
that we've done with Blackwell.
Because we realize that inference is not
going to be just zero shot or one shot,
but inference is going to be,
it's gonna include long thinking.
And it's a, it's gonna be essentially
an AI coming up with all kinds
of different scenarios in its mind.
And then of course,
with more compute, offer a better answer,
which is a new way of scaling.
We call it test time scaling,
or inference time scaling.
And so because of that,
we invented new technologies so that,
so that inference can continue to be super fast,
super energy efficient.
It's like 30 times more energy efficient than,
and faster than previous inference capabilities.
So all these capabilities are now possible.
Jensen, we have less than five minutes left,
and I wish we could cover a million more topics,
but I wanna ask you about the future of Nvidia.
[Jensen] I got all day.
You have, do you? [audience laughs]
[Jensen] Yeah. You really are dispatching
your AI to take those 56 meetings.
I'm still gonna, I'm gonna think about that
for a long time, 56 meetings in a day.
You've had this incredible run.
I mean, truly incredible run.
Nvidia is at the top.
And I'm wondering how you think about maintaining that.
And I think this was probably on
a lot of people's minds this week,
as we saw that the, you know,
the CEO of Intel, Pat Gelsinger stepped down,
and some reported he was forced out.
And, you know, I don't think anyone would've
thought of Intel, you know,
in the same way 20 years ago.
And so how do you ensure that NVIDIA
doesn't end up in the same boat 10, 20 years from now?
What are the, what are the actual
strategic steps you're taking to ensure that?
Yeah, a really good question.
First of all,
I've worked with Pat for several decades,
and really appreciated his friendship and partnership.
And we've got all kinds of things
that we're working on with Intel.
And we've selected their CPUs
for many of our programs that are upcoming.
And this is a really important company.
And, and they're still just,
if I could put it out there,
just the teams that we work with,
on their CPU side, on their PC side,
on the data center side are really extraordinary.
And so we've got a lot of great friendships there,
and I wish 'em well.
At the core, at the core,
this is the challenge when,
when something fundamental happens to an industry,
and that force is so incredible.
And you have to keep remembering this,
that in fact, the central processing unit
was described in 1963, 1964.
It was literally called the central processing unit.
And the IBM system 360,
fairly clearly describes modern computing as we know it.
And it lasted about 60 years.
And so all of a sudden, not all of a sudden,
but over, because of deep learning,
and because of the incredible pace of innovation
around deep learning and machine learning,
the world went from coding,
and coding instructions that ran on CPUs
to machine learning of neural networks
that run on GPUs.
This force is so incredible.
It's not as if,
it's not as if you can compete against this.
You're either, you're either in this force,
you're either on this wave,
or you've missed that wave.
But no CPU could ever,
no matter how good it is,
could overcome machine learning.
This is no different than,
no disk drive can overcome
the power of cloud computing.
You just, you're not gonna build a better computer.
Because the fundamental trends and the benefits
of cloud computing is so incredible.
You're not gonna overcome the trend of mobile computing.
It's just too fundamental, mobile cloud.
And so, so machine learning is even more powerful than that.
The fundamental way that we do software has changed.
And the result of that has fundamentally changed.
Instead of creating tools, software tools,
we're now developing AIs.
And, and so the way you build it,
how you process it,
what you can build as a result of it
has all fundamentally changed.
And this all happened in 10 years.
So you're either preparing for that 10 year trend,
or you get caught off guard by that.
And this is, this is, you know,
because of, because of the success of CPUs
in the last 60 years, it's not,
it's not, it's probably not unexpected.
And it's probably even understandable
that an organization that is so focused
on building CPUs and and its leadership
in that area would be caught off guard.
And so I think first of all,
I think that that's really at the core of it.
They're, of course, a lot of other,
other people could criticize their circumstance,
but I think foundationally it's really about
a seismic change in technology,
and that the future of computing
is gonna fundamentally change.
And so how do you ensure that 30 years from now,
we're not saying that about GPUs,
and NVIDIA's position?
Well, there's no guarantee, Lauren.
First of all, we just gotta put it out there,
there's no guarantee.
I will say that because of the way
that NVIDIA was born and built by hand,
the company was really designed for agility.
You know, remember, this is a company
that started out with a fundamental
technology architecture that two years
after we were founded, we completely changed.
We admitted the fact that recognized,
and admitted the fact that the architecture that we had,
the technology we had was just, was just wrong.
We understood why we did it,
but it was fundamentally the wrong approach.
And completely changed the company
to go after a new way of doing computer graphics.
We also reinvented the company several times,
as, you know, people know well,
inventing Cuda, and then after that,
pursuing deep learning,
going from a chip company to a systems company,
systems company to an infrastructure company.
And so as we, as we make those, those changes,
it's in response to wanting to realize
our hopes and dreams of being an impactful company
in an industry that we care so deeply about,
the computer industry.
And so, so long as,
so long as the computer industry
continues to be important,
and we continue to be agile,
and be surrounded by amazing
computer scientists and talent,
we'll continue to, you know,
to shape shift ourselves,
to be impactful in the future.
And so that's my hope.
We'll see how it turns out.
Jensen, I can only speak for myself,
but I think I speak for others here when I say
I can't wait to see what you come up with next,
what your next pivot shift.
I mean, you are in the future right now.
So. I am.
Yeah. And we really appreciate you
taking the time to be here. A couple of hours.
You are. Yeah.
So save us all, please.
Lauren, I gotta tell you, the future is bright.
[audience laughs] Everything's fine.
The water's safe. That's Excellent. Thank you.
I think your stock just went up as you said that.
So, thank you so much, Jensen.
We really appreciate it. Thank you, Lauren.
Thank you. [audience applauds]
I wish I could be there with you.
Take care. Have a great conference.
[audience applause continues]
