2025

Highlights
“NVIDIA is hosting the Super Bowl of AI ”
“NVIDIA mobilizes IT partners
to spread the AI gospel”
CIO Live

GTC used to be a gathering for a few thousand

developers. This year, we’re filling stadiums. It’s been
called the "Woodstock of AI"—this year, the "Super
Bowl of AI." The only difference? Everyone’s a winner.
More companies, more industries, more breakthroughs.
Every year, GTC just gets bigger because AI
solving more interesting problems for more industries
and more companies.
“NVIDIA Showcases AI Future at GTC”
Yahoo Finance

AI started with perception—seeing and hearing the

world. Then it learned to generate. Now it can
reason, plan, and act. The next wave is physical AI—
machines that understand the world and move
through it. Each step opens up new possibilities.
And we’re just getting started.
“NVIDIA’s big show hits its marks”
“NVIDIA's true strength
lies in its ecosystem”
ARC Advisory Group

This is what GTC is all about. A long time ago,

this was the only slide we had—one library after
another. Just as we needed AI frameworks to
create AIs, we need frameworks for physics,
biology, quantum, and more. We call them
CUDA-X libraries—acceleration frameworks for
every field of science.
“This is NVIDIA at its best”
Ben Thompson / Stratechery
“Three scaling laws are stacked
together and working in tandem”
SemiAnalysis

There are three fundamental questions in AI. provide, the smarter AI becomes? This is
How do you solve the data problem? How do where almost the entire world got it wrong.
you train without humans in the loop? And The scaling law of AI is more resilient—and
how do you scale—how do you find an in fact, hyper-accelerated.
algorithm where the more resources you
“NVIDIA’s strategy becomes clear:
taking on a much larger role than
ever before”
Technalysis

AI is going to go everywhere. The rest of the world has

different system configurations, different environments,
different needs—enterprise IT, manufacturing, robotics,
even new GPU clouds. Each one is building their own stack,
and NVIDIA is helping power them all.
“NVIDIA says every company Every company that builds something will have two

will become an AI factory” factories: one to make the product, and one to
manufacture the intelligence behind it. That second
Business Insider factory—the AI factory—will become essential.
“NVIDIA launches
Dynamo, the operating
system of an AI factory”
MarketWatch

We call it NVIDIA Dynamo. It’s the operating system of an AI factory. It

manages parallelism, batching, memory, context—everything needed to
orchestrate massive models across thousands of GPUs. Just like VMware
was for enterprise IT, Dynamo is for AI agents.
“NVIDIA hasn’t slowed the pace.
It is accelerating inference
throughput gains”
SemiAnalysis

Inference is the ultimate extreme computing problem. You're trying to

generate a whole bunch of tokens—but you're trying to do it as quickly as
possible. Smart AIs that are super fast. That's the trade-off. The perfect
answer is the upper right of the Pareto frontier—high interactivity, high
throughput. That’s the best AI factory you can build​.
“Tech giant details massive AI
opportunity ahead”
Investors Business Daily

We're building AI infrastructure for the cloud, for

enterprise IT, and for robots. And now, sovereign AI is
emerging as the fourth. Just as we’ve modernized the
cloud, we’re now reinventing the world’s $500 billion
enterprise IT industry for the age of AI​.
“Storage players ride the NVIDIA bus at GTC 2025”
Computer Weekly

Storage has to be completely reinvented for the age of AI. It won’t just retrieve
data—it will embed and interpret it. We’re working with the entire storage
industry to build systems that are GPU-accelerated and
semantics-driven. This is the enterprise AI
infrastructure of the future.
“NVIDIA unveils powerful new
Blackwell is in full production, and the ramp has been
incredible. Customer demand is strong—and later this
year, we’ll introduce Blackwell Ultra: more memory, more

Blackwell Ultra chips at Super Bowl of AI” bandwidth, and support for longer context and deeper
reasoning. It’s the next step in building AI infrastructure.

New York Post

“We saw NVIDIA up close and personal last week, and were blown away”
Jim Cramer / CNBC
“NVIDIA’s silicon
photonics-based switch
enable clusters with
millions of GPUs”
Tom’s Hardware

To scale to hundreds of thousands of GPUs, we need to move beyond copper.

That’s where silicon photonics comes in. We’ve built the world’s first 1.6
terabit-per-second co-packaged optics system—saving megawatts of power
and unlocking the next generation of AI factories.
“NVIDIA networking steals the show
for the second GTC in a row”
VentureBeat

To build AI factories at scale, we had to reinvent networking.

Spectrum-X brings AI performance to Ethernet, and with
our silicon photonics technology, we can now connect
hundreds of thousands of GPUs while saving megawatts of
power. This is the network fabric of the AI era.
“NVIDIA's Spark desktop
AI supercomputer
arrives this summer”
Engadget

This is what a PC should look like in the age of AI.

DGX Spark delivers 20 petaFLOPS of performance,
72 CPU cores, and high-bandwidth memory—all in a
workstation. It’s built for researchers and developers
powering the next wave of AI.
“What’s NVIDIA’s next big thing?
Superchip, named after astronomer
who discovered dark matter”
The Wall Street Journal

Rubin is our next-generation platform—brand new CPU,

GPU, memory, networking, everything. It’s designed for
the future of AI: trillion-parameter models, sovereign
infrastructure, and massive-scale AI factories.
“NVIDIA delivers
multiyear roadmap
at GTC”
Bloomberg

No company in history has ever laid out a roadmap like this—multiple

years, multiple architectures, every detail. But we do it because we’re
building infrastructure. The world is counting on us, and our partners
need trust, no surprises, and confidence in execution.
“NVIDIA GTC 2025:
AI Matures into Enterprise Infrastructure”
Bain & Company

We’re not a chip company—we’re an infrastructure company.

From RTX Pro to DGX Station, from B200 to GB300, we build the
full stack for AI: systems, networking, software, everything.

This is NVIDIA.
“NVIDIA’s new reasoning
models pave way for
advanced AI agents”
SiliconANGLE

Our Llama Nemotron model is built for reasoning—step-by-step thinking,

planning, and problem solving. It’s not just about answering questions. It’s
about understanding context, breaking down complex tasks, and taking action.
“NVIDIA launches blueprint to
revolutionize weather forecasts”
IT Brief

We’re seeing more extreme weather events than ever—threatening

lives and property. The NVIDIA Omniverse Blueprint for Earth-2 will
help industries prepare for and mitigate climate disasters with
faster, more accurate, energy-efficient forecasts.
“NVIDIA Dreams of a Robot Future”
The New York Times

The time for robots has come. Everything that moves—cars, train the AI, one to simulate it, and one to run it. We call that
forklifts, surgical arms, warehouse bots—is becoming physical AI—AI that understands the laws of physics, that can
autonomous. To build them, we need three computers: one to reason, plan, and act in the real world.
“Robots Invade NVIDIA’s GPU
Technology Conference”
CNET

In robotics, training is just the beginning. Post-training,

AI models need to reason, adapt, and simulate in real
time. Test-time scaling becomes critical—robots don’t
just recall, they think and act in dynamic environments.
That’s why we’re building physical AI infrastructure.
“NVIDIA’s Cosmos makes robot
training freakishly realistic—
and that changes everything”
VentureBeat

Omniverse is our operating system for physical AIs. We use it to

condition Cosmos, our generative model that understands the
physical world. Together, they create infinite training environments
that are grounded in physics and fully under our control.
“NVIDIA teams up with Disney and Today, we're announcing
something really, really special—a
Newton. It's a physics engine
designed for fine-grained control,

Google to build smarter robots”

partnership of three companies: soft bodies, tactile feedback—
Google DeepMind, Disney everything you need to train robots
Research, and NVIDIA. We call it with precision.
The Globe and Mail
“Cutest robots I’ve ever seen”
Monica White / Digital Trends
“Disney’s robotic droids are the
toast of Silicon Valley”
The Wall Street Journal

This is Besh, one of the BDX droids built by Disney

Research, powered by two NVIDIA computers and trained in
simulation. Tactile feedback, soft-body physics, fine motor
skills—Disney's BDX droids have got it all. This is what
happens when physical AI meets a little bit of magic.
“NVIDIA’s latest GROOT
AI model just took
another step closer to
fully humanoid robots”
TechRadar

We’re open-sourcing GR00T N1—our developers can teach robots skills

foundation model for humanoid robots. through demonstration. Companies like
It learns by watching humans, Boston Dynamics, Figure, and Unitree are
understanding language and video to already building on it. This is how we’ll
generate actions. With GR00T, bring physical AI to life.
“GM’s partnership with NVIDIA
could change driving forever”
Autoblog

We’re partnering with GM to build their future self-driving fleet. AI in the

car, AI in the factory, AI in the enterprise—every part of the company will
be powered by NVIDIA. The time for autonomous vehicles has arrived.
“Jensen Huang’s keynote was fantastic”
Jim Cramer / CNBC
 What the press said…

“NVIDIA is the undisputed tone-setter “Analysts roundly agree that

of the entire technology industry.” NVIDIA remains far ahead”
Creative Strategies The Wall Street Journal

“GTC felt more bullish “NVIDIA delivered

than ever” at GTC again”
TechCrunch Quartz
 What the financial analysts said…

“NVIDIA remains “NVIDIA continues to

the king of the hill” dominate the AI value chain”
Citi Bank of America

“From the “It was the best of “Blackwell ramp

Woodstock of AI times, it was the was the man of
to the Superbowl” best of times” the match”
Daiwa Cantor Itau
245K 27K 1,200 43M 58K 400
Conference In-Person Talks and Keynote Press Partner
Registrations Attendees Sessions Views Articles Sponsorships
We talked about several things.

Blackwell is in full production. We’re building AI

infrastructure for the cloud, for enterprise, and
for robots. We’re laying the blueprints not just
for our company—but for the industries that
are here, the ones that aren’t here, and the
entire supply chain.

I want to thank all of you for coming to GTC.

This is an extraordinary moment in time.

Jensen Huang