What’s next in AI is at GTC 2025. Not only the technology, but the people and ideas that are pushing AI forward — creating new opportunities, novel solutions and whole new ways of thinking. For all of that, this is the place.
Here’s where to find the news, hear the discussions, see the robots and ponder the just-plain mind-blowing. From the keynote to the final session, check back for live coverage from San Jose, California.
Research Giants Bill Dally, Yann LeCun Share the Stage at GTC 🔗
Two luminaries in AI research — Bill Dally, chief scientist at NVIDIA, and Yann LeCun, chief AI scientist at Meta and a professor at New York University — took the stage today at the San Jose Civic auditorium to share their vision for the field’s future.
Dally asked LeCun to share his thoughts on artificial general intelligence, the impact of AI on scientific research, his current work on world models and the qualities needed to foster innovation.
LeCun predicted that artificial general intelligence, or AGI — which he prefers to call advanced machine intelligence, because “human intelligence is superspecialized, so calling it general is a misnomer” — will be viable in three to five years.
He emphasized the importance of open-source projects to support the development of diverse AI assistants.
“We need assistants that are extremely diverse,” he said. “We need to speak all languages, understand all the cultures, all the value systems, all the sectors of interest. So we need a platform that anybody can use to build those assistants, a diverse population of assistants — and right now that can only be done through open-source platforms.”
LeCun also spoke about his work at Meta to develop world models that can understand, reason and plan around physical environments.
“What you need is a predictor that, given the state of the world and an action you imagine, can predict the next state of the world,” he said. “And if you have such a system, then you can plan a sequence of actions to arrive at a particular outcome.”
Dally noted that building world models like these will require significant AI infrastructure powered by NVIDIA GPUs.
“Keep them coming,” LeCun quipped. “We’re going to need all the computation we can get our hands on.”
Wrap-Up: Key Points From the GTC Keynote 🔗
Here are the major points NVIDIA founder and CEO Jensen Huang covered in his keynote:
- We’re at a $1 trillion computing inflection point. AI computing demand is accelerating rapidly, driven by the rise of reasoning AI and agentic AI. The scale and complexity of AI workloads are transforming data center investments worldwide.
- NVIDIA Blackwell is in full production, delivering 40x the performance of Hopper. The Blackwell architecture significantly enhances AI model training and inference, enabling more efficient and scalable AI applications. And the next evolution of the NVIDIA Blackwell AI factory platform, Blackwell Ultra, will be coming to systems in the second half of this year.
- NVIDIA will follow an annual rhythm for the buildout of AI infrastructure. Each year will bring new GPUs, CPUs and accelerated computing advancements, including the upcoming NVIDIA Vera Rubin architecture, designed to drive performance gains and efficiency improvements in AI data centers.
- AI infrastructure, including photonics and AI-optimized storage, is set to revolutionize the industry. Advanced networking and storage solutions will improve AI scalability, efficiency and energy consumption in large-scale data centers.
- Physical AI for industrial and robotics is a $50 trillion opportunity. AI-powered robotics and automation are set to transform manufacturing, logistics, healthcare and other industries, with the NVIDIA Isaac and Cosmos platforms leading the way.
Watch the replay:
Read more details about the keynote below and stay tuned for more insights and sessions throughout the conference, running through Friday, March 21.
Dive In: NVIDIA Founder and CEO Jensen Huang’s Keynote at GTC 🔗
Welcome to NVIDIA
Huang kicked off the keynote by bringing the audience “into NVIDIA’s headquarters,” with a stunning visual of its lobby that seemed to embrace the audience. He began by talking about where NVIDIA started 25 years ago, with GPUs. He described the growth of AI over the past decade, including the emergence of agentic AI — which can reason how to solve a problem, plan and take action.
AI at an Inflection Point
Huang then outlined the development of AI able to reason “step by step by step” and discussed how demand for inference and reinforcement learning is driving demand for AI computing. Demand for GPUs from the top four cloud service providers is surging, as AI is going through an “inflection point.” Huang said he expects the value of data center buildout to reach $1 trillion.
NVIDIA CUDA Ecosystem
NVIDIA CUDA-X GPU accelerated libraries and microservices are now serving every industry, Huang explained. In the future, Huang said, every company will have two factories: one for what they build and another for AI. Ticking through a sampling of NVIDIA’s role in a variety of endeavors, he announced that NVIDIA is going to open-source its cuOpt decision optimization platform. The installed base of CUDA is now “everywhere,” he said. “We have reached the tipping point of accelerated computing — CUDA has made it possible.”
General Motors, NVIDIA Collaborate on AI
AI needs infrastructure, Huang explained. AI is now going out “to the rest of the world,” in robotics and self-driving cars, factories and wireless networks. One of the earliest industries AI went into was autonomous vehicles, Huang said. “We build technology that almost every single self-driving car company uses,” he added, whether in the data center or the car. Huang announced the next step of that journey: GM, the largest U.S. automaker, is adopting NVIDIA AI, simulation and accelerated computing to develop next-generation vehicles, factories and robots. He also announced NVIDIA Halos, a comprehensive safety system bringing together NVIDIA’s lineup of automotive hardware and software safety solutions with its cutting-edge AI research in AV safety.
Data Centers and Inference
Next up: Huang spoke about data centers. He reported that the NVIDIA Blackwell platform is in full production, sharing a shot of systems from a broad range of industry partners. “How is this not beautiful,” Huang said.
He described how Blackwell supports extreme scale-up. “The reason why we wanted to do this is to solve an extreme problem,” Huang said, “and it’s called inference.”
Inference, Huang explained, is token generation, which will be critical to businesses. AI factories that generate these tokens have to be built with extreme efficiency and performance. And demand for tokens will only grow, with the latest generation of reasoning models able to think through and solve increasingly complex problems.
To further accelerate inference on a large scale, Huang announced NVIDIA Dynamo, open-source software for accelerating and scaling AI reasoning models in AI factories. “It is essentially the operating system of an AI factory,” Huang said.
NVIDIA Blackwell Ultra
Huang described how Blackwell delivers a “giant leap” in inference performance. “You want to make sure you have the most energy-efficient architecture you can possibly get,” Huang said, showing how Blackwell does more work, using less power than the previous generation. “The more you buy, the more you save,” Huang said, drawing laughter from the audience. “It’s even better than that — the more you buy, the more you make.”
Huang then rolled a video showing how a new NVIDIA Omniverse Blueprint can help plan a 1 gigawatt AI factory, letting engineers design, test and optimize a new generation of intelligence manufacturing data centers using digital twins.
Then, he announced the next evolution of the NVIDIA Blackwell AI factory platform, NVIDIA Blackwell Ultra, coming in systems the second half of this year. NVIDIA Blackwell Ultra boosts training and test-time scaling inference — the art of applying more compute during inference to improve accuracy — to enable organizations everywhere to accelerate applications such as AI reasoning, agentic AI and physical AI.
NVIDIA Vera Rubin
Paying tribute to astronomer Vera Rubin, Huang outlined a roadmap that will deliver data center performance gains for years to come, Huang offered new details on the next-generation NVIDIA Rubin Ultra GPU and NVIDIA Vera CPU architectures, packed with innovations. “Basically everything is brand new except for the chassis,” Huang said. Systems built on Rubin Ultra, including the Vera Rubin NVL 144, will arrive in the second half of next year. And due for the second half of 2027: systems built on Rubin Ultra. “You can see that Rubin is going to drive the cost down tremendously,” Huang said.
NVIDIA Photonics
Huang shifted to discussing how NVIDIA will help customers scale out to ever larger systems. Key to that is tightly integrating photonics — networking technologies that rely on transmitting data using light rather than electrical signals — into accelerated computing infrastructure. Enabling AI factories to connect millions of GPUs across sites while reducing energy consumption and operational costs, NVIDIA Spectrum-X and NVIDIA Quantum-X silicon photonics networking switches fuse electronic circuits and optical communications. “This is really crazy technology,” Huang said. NVIDIA photonics switches integrate optics innovations with 4x fewer lasers to deliver 3.5x more power efficiency, 63x greater signal integrity, 10x better network resiliency at scale and 1.3x faster deployment compared with traditional methods.
DGX Spark and DGX Station
Enabling AI developers, researchers, data scientists and students to prototype, fine-tune and inference large models on desktops, NVIDIA unveiled DGX personal AI supercomputers powered by the NVIDIA Grace Blackwell platform. Describing it as the perfect Christmas present, Huang announced DGX Spark — formerly Project DIGITS — and DGX Station, a new high-performance NVIDIA Grace Blackwell desktop supercomputer powered by the NVIDIA Blackwell Ultra platform, bringing the power of the Grace Blackwell architecture to the desktop. Users can run these models locally or deploy them on NVIDIA DGX Cloud or any other accelerated cloud or data center infrastructure. “This is the computer of the age of AI,” Huang said.
Agentic AI
Transitioning into a discussion of the future of agentic AI, Huang announced the open Llama Nemotron family of models with reasoning capabilities, designed to provide developers and enterprises a business-ready foundation for creating advanced AI agents that can work independently or as connected teams to solve complex tasks. Built on Llama models, the NVIDIA Llama Nemotron reasoning family delivers on-demand AI reasoning capabilities. NVIDIA enhanced the new reasoning model family during post-training to improve multistep math, coding, reasoning and complex decision-making.
Physical AI and Robotics
Describing robots as the next $10 trillion industry, Huang said that by the end of this decade, the world is going to be at least 50 million workers short. NVIDIA offers a complete suite of technologies for training, deploying, simulating and testing next-generation robotics.
In a video, Huang announced the availability of NVIDIA Isaac GR00T N1, the world’s first open, fully customizable foundation model for generalized humanoid reasoning and skills.
NVIDIA also announced a major release of new NVIDIA Cosmos world foundation models, introducing an open and fully customizable reasoning model for physical AI development and giving developers unprecedented control over world generation.
“Using Omniverse to condition Cosmos, and Cosmos to generate an infinite number of environments, allows us to create data that is grounded, controlled by us and yet systematically infinite at the same time,” Huang said.
He also introduced the Newton open-source physics engine for robotics simulation, under development with Google DeepMind and Disney Research, before he was joined on stage by a diminutive robot, “Blue,” that emerged from a hatch on the floor, beeping and booping at Huang.
‘Let’s Wrap Up’
Huang wound up his talk by emphasizing several major themes.
First, Blackwell is in full production — “and the ramp is incredible, customer demand is incredible,” Huang reported. “And for good reason, because there is an inflection point in AI, the amount of computation we have to do in AI is so much greater as a result of reasoning AI, and the training of reasoning AI systems and agentic systems.”
Second, Blackwell NVL72 with Dynamo offers 40x the AI factory performance of NVIDIA Hopper, and “inference is going to be one of the most important workloads in the next decade as we scale out AI.”
Third, NVIDIA has an “annual rhythm” of roadmaps, so the world can plan its AI infrastructure. NVIDIA is building three AI infrastructures: one for the cloud, a second for the enterprise and a third for robots.
Countdown to Keynote at GTC 🔗
The SAP Center is waking up, bathed in a soothing purple glow. A sea of attendees flows in — developers, researchers, business leaders — filling the space with a charge of anticipation.
Up on the big screen: the pre-show with Acquired podcast hosts Ben Gilbert and David Rosenthal, who are interviewing industry figures from the GTC show floor. Earlier today they even got a drop-in from NVIDIA founder and CEO Jensen Huang.
It’s almost time.
At 10:00 a.m. PT, Huang takes the stage, setting the tone for everything unfolding at GTC. Crowds in the arena will feel it — the shuffle of last-minute arrivals, the glow of smartphone screens.
Those tuning in remotely won’t miss a moment. Watch the keynote live. The future of AI, accelerated computing and everything in between is about to be unveiled.
Stay with us — we’re just getting started.
Keynote Today: Tune In Early for the Preshow With Acquired 🔗
The excitement starts long before NVIDIA founder and CEO Jensen Huang takes the stage. Catch the “Live at NVIDIA GTC With Acquired” broadcast now, featuring luminary speakers who’ll offer fascinating insights into NVIDIA’s remarkable journey over the past three decades.
State of AI Art: How Artists Bring AI, Robots Into Creative Process 🔗
“Consumption,” a piece by BREAKFAST, on display at GTC.
Artists over millennia have created with every medium available — fruit and vegetable dyes, blocks of marble, acrylic paints, cameras, rendering software, 3D printers. They’re now harnessing AI and robotics to build immersive experiences that reflect on the ways people respond to and interact with technology.
In a GTC panel discussion hosted today by Heather Schoell, creative director of AI strategy at NVIDIA, four artists shared how they use AI and robotics in their creative processes.
The integration of robotics in art “is a very logical continuation in what is a very long spectrum of humans and our relationship to our tools,” said panelist Catie Cuan, founder of consumer AI company Zenie and a postdoc at Stanford University, where she leads art and robotics efforts at the Stanford Robotics Center.
A former professional dancer, Cuan’s works include an eight-hour duet with a robot arm and a symphony of dancing robots that played sounds as they moved.
Another panelist, Alexander Reben, was OpenAI’s first artist in residence. He’s used large language models, visual generative AI models and NeRFs to create 3D sculptures.
Interactive art by two of the panelists — Zolty, a kinetic and robotics artist known as BREAKFAST, and Emanuel Gollob, a doctoral researcher at the University of Arts Linz, Austria — is on display at the main entrance of GTC.
BREAKFAST works with real-time data to create large kinetic art sculptures that are on display aboard a Royal Caribbean cruise ship and the Fontainebleau hotel in Las Vegas. The studio’s piece at GTC, titled “Consumption,” runs on an NVIDIA RTX GPU and features robotic arches that move in response to real-time data about water usage, including rainfall, reservoirs, groundwater levels and municipal supplies.
Gollob’s robotic installation, called “Doing Nothing With AI,” uses generative robot control, brainwave measurements and reinforcement learning to move in a way that guides a participant wearing an EEG headband to do nothing and enjoy a moment of inaction.
Panelists Discuss Energy Efficiency in the Age of Generative AI 🔗
To a packed room of attendees, panelists held an illuminating discussion on AI, energy and climate issues today at GTC. Josh Parker, senior director of corporate sustainability at NVIDIA, was joined by Lauren Risi from the Wilson Center, David Sandalow from Columbia University and Bernhard Lorentz of Deloitte.
“In the world of climate change, there’s an enormous amount of things that AI can do,” said Sandalow. “How can you mobilize AI tools to mitigate climate change?”
The industry experts addressed concerns about energy efficiency with the rise of generative AI and the uptick in data centers.
“NVIDIA GPUs are delivering 100,000x better energy efficiency over the past decade,” said Parker, summing up the energy efficiency gains from NVIDIA GPUs.
BMW Group Shows Off 3D Collaboration for Battery Assembly Factories 🔗
BMW Group presented the latest updates to its 3D AppStore today at GTC.
The 3D AppStore, BMW’s cloud streaming service for high-end visualization, is now making it possible to support the automaker’s battery assembly process across various factories worldwide. It allows engineers, designers and others to work in the digital environment collaboratively.
“With the avatars, they see the factory scene and can be in there together, and then you can basically just ask what is happening here, and the AI gives you a detailed description of what the station here is doing,” said Xaver Freiherr Loeffelholz von Colberg, product owner of 3D AppStore at BMW Group.
Offering a glimpse into the future of automotive design and manufacturing, BMW has two more speaker sessions on Thursday, March 20.
Explaining Tokens — the Language and Currency of AI 🔗
GTC attendees are likely to hear a lot about tokens — the language and currency of AI.
Tokens are units of data processed by AI models during training and inference, enabling prediction, generation and reasoning. The faster tokens can be processed, the faster AI models can learn and respond.
Get up to speed on tokens, tokenization and the ways enterprises can boost revenue by lowering cost per token in our explainer article.
GTC 2025: Real AI, Real Problems, Real Solutions 🔗
AI is confronting humanity’s toughest challenges head on. See it unfold next week at the NVIDIA GTC conference in San Jose, California.
From transformative healthcare sessions like “Revolutionizing Cardiac MRI Analysis and Diagnosis With AI” and “Designing the Future: Protein Engineering, AI and Responsible Innovation” to environmental breakthroughs in “Autonomous Systems and Remote Sensing for Better Earth Data,” “The Role of AI and Accelerated Computing in Understanding and Mitigating Urban Climate Change” and “Enhancing Photovoltaic Power Prediction With High-Resolution Weather Forecasting From NVIDIA Earth-2,” the impact is tangible and global.
The Future Rolls Into San Jose 🔗
Anyone who’s been in downtown San Jose lately has seen it happening. The banners are up. The streets are shifting. The whole city is getting a fresh coat of NVIDIA green.
From March 17-21, San Jose will become a crossroads for the thinkers, tinkerers and true enthusiasts of AI, robotics and accelerated computing. The conversations will be sharp, fast-moving and sometimes improbable — but that’s the point.
At the center of it all? NVIDIA founder and CEO Jensen Huang’s keynote, offering a glimpse into the future. It’ll take place at the SAP Center on Tuesday, March 18, at 10 a.m. PT. Expect big ideas, a few surprises, some roars of laughter and the occasional moment that leaves the room silent.
But GTC isn’t just what happens on stage. It’s a conference that refuses to stay inside its walls. It spills out into sessions at McEnery Convention Center, hands-on demos at the Tech Interactive Museum, late-night conversations at the Plaza de César Chávez night market and more. San Jose isn’t just hosting GTC. It’s becoming it.
The speakers are a mix of visionaries and builders — the kind of people who make you rethink what’s possible:
🧠 Yann LeCun – chief AI scientist at Meta, professor, New York University
🏆 Frances Arnold – Nobel Laureate, Caltech
🚗 RJ Scaringe – founder and CEO of Rivian
🤖 Pieter Abbeel – robotics pioneer, UC Berkeley
🌍 Arthur Mensch – CEO of Mistral AI
🌮 Joe Park – chief digital and technology officer of Yum! Brands
♟️ Noam Brown – research scientist at OpenAI
Some are pushing the limits of AI itself; others are weaving it into the world around us.
📢 Want in? Register now.
Check back here for what to watch, read and play — and what it all means. Tune in to all the big moments, the small surprises and the ideas that’ll stick for years to come.
See you in San Jose. #GTC25
