-
A new, open source AI model for humanoid robotics
Nvidia is releasing what it’s calling an AI foundation model for humanoid robotics.
Announced at GTC 2025 in San Jose, the model, dubbed Groot N1, is a “generalist” model, trained on both synthetic and real data. Nvidia claims that Groot N1 features a “dual system architecture” for “thinking fast and slow,” inspired by human cognitive processes.
Groot N1 is available in open source. Alongside the model, Nvidia is releasing simulation frameworks and blueprints for generating synthetic training data.
Image Credits:Justin Sullivan / Getty Images -
-
AI ‘agents’ is an ‘overused’ term, says Nvidia-backed agents startup founder
No one knows what the hell AI “agents” are, and even an Nvidia-backed AI agents startup founder admits the term is “overused.”
“AI agents really take the trophy” when it comes to “overused” AI jargon, said Misha Laskin, co-founder and CEO of Reflection AI, during a panel at GTC Tuesday.
Laskin is still bullish on AI agents, of course, noting their potential for building reasoning models and “world models.”
Reflection AI, which builds autonomous AI coding agents, emerged from stealth earlier this month with a $130 million fundraise, Bloomberg reported.
Both of Reflection AI’s co-founders are former Google DeepMind researchers, and the company won backing from investors, including Nvidia’s VC arm NVentures and Sequoia Capital.
-
-
Wall Street doesn’t seem terribly impressed with Nvidia’s reveals
Nvidia revealed a bunch of products during the first day of GTC. But Wall Street seemingly wasn’t impressed. The company’s stock price was down around 4% in after-hours trading.
Why the lukewarm reception? Well, it might be because there wasn’t much in the way of surprises. Nvidia’s GPU lineup was widely leaked, down to the release time frames of each chip. Investors might’ve been holding out hope for “one last thing” — or perhaps an accelerated launch window.
-
-
Going all in on autonomous vehicles
During the automotive portion of CEO Jensen Huang’s speech, he referred to AlexNet, a neural network architecture that gained widespread attention in 2012 when it won a computer image recognition contest. AlexNet achieved 84.7% accuracy in an academic competition called ImageNET.
The breakthrough result led to a resurgence of interest in deep learning, a subset of machine learning that leverages neural networks.
And as it turns out, AlexNet spurred Nvidia to go “all in” on autonomous vehicles, Huang said. “The moment I saw AlexNet — and we’ve been working on computer vision for a long time — the moment I saw AlexNet was such an inspiring moment, such an exciting moment,” he said.
-
-
‘AI personal supercomputers’
CEO Jensen Huang unveiled the two new machines: DGX Spark (previously called Project Digits) and DGX Station. The computers will allow users to prototype, fine-tune, and run AI models in a range of sizes at the edge.
“This is the computer of the age of AI,” Huang said during the presentation. “This is what computers should look like, and this is what computers will run in the future. And we have a whole lineup for enterprise now, from little, tiny ones to workstation ones.”
-
-
GM is turning to Nvidia to bring AI into the physical world
General Motors has been working with Nvidia for years now, but the two companies are expanding that collaboration in a bid to use AI to build more capable robots, better factories, and beef up automated driving efforts.
Nvidia CEO Jensen Huang noted that the company would be working with GM in three areas. “AI for manufacturing, so they can revolutionize the way they manufacture; AI for enterprise, so they can revolutionize the way they work to design cars and simulate cars; and then also AI for in the car.”
The announcement comes just a few months since GM stopped funding its commercial robotaxi program and shifted resources toward its hands-off advanced driver assistance system known as Super Cruise. GM plans to continue to improve its driver assistance features and eventually roll out fully autonomous personal vehicles.
-
-
Nvidia touts ‘incredible’ new reasoning model for enterprises that beats DeepSeek
Nvidia CEO Jensen Huang announced a new Llama-based reasoning model for enterprises during his keynote at GTC, describing it as an “incredible new model that anybody can run.”
The model is called Nvidia Llama Nemotron Reasoning. It’s part of Nvidia’s Nemotron family of models, which are intended to supercharge AI agents (although no one knows what the hell an AI agent is quite yet.)
Huang touted that the new reasoning model beats DeepSeek’s R1 substantially on both accuracy and speed. He also mentioned partnerships with Accenture, Blackrock, and others on the model.
-
-
‘Chief revenue destroyer’
Huang called himself the “chief revenue destroyer” during his GTC keynote today after talking up the company’s new Blackwell architecture and taking a jab at Nvidia’s Hopper architecture which was released in March 2022.
Huang said that there are circumstances where Hopper chips are “fine” — then added “but not many” because technology is moving too fast to justify buying or running the company’s old model anymore.
“[Blackwell] is better, OK?” Huang said. “The more you buy, the more you save. It’s even better than that. Now, the more you buy, the more you make.”
-
-
A slew of new GPUs are coming
During his keynote, CEO Jensen Huang announced a ton of new GPUs, but perhaps the most significant is the Vera Rubin, which will feature tens of terabytes of memory and a custom Nvidia-designed CPU called Vera. It will be released in the second half of 2026.
Rubin will be followed by Rubin Ultra in the second half of 2027, a collection of four GPUs in one package, Huang said. Later this year, Nvidia will release Blackwell Ultra, a more powerful incarnation of the company’s current Blackwell chips.
-
-
Nvidia versus DeepSeek
Jensen Huang has it out for Chinese AI lab DeepSeek. Onstage during his GTC keynote, Huang claimed DeepSeek’s “reasoning” AI model, R1, uses significantly more computing than a typical non-reasoning model.
Huang was perhaps spurred to do so because of DeepSeek’s impact on Nvidia’s stock price in recent months. The release of DeepSeek’s R1 and other models, which are widely perceived to be more efficient than most, led investors to assume that Nvidia’s chips would face less demand in the future — particularly if more efficient AI models became the norm.
Nvidia’s CEO has made the case that more efficient AI will actually increase demand for GPUs.
-
-
The AI ‘inflection point’
During his GTC 2025 keynote presentation, Nvidia CEO Jensen Huang said that “AI is going through an inflection point” as he compared the peak year of Hopper GPUs in 2024 to Blackwell so far in 2025.
Watch the full keynote with us below.
-
-
Jensen Huang says that practically the ‘entire world’ got AI scaling wrong
It didn’t take long for Nvidia CEO Jensen Huang to start talking about AI “scaling laws” during his keynote at GTC 2025. He claimed that nearly the “entire world got it wrong” on AI scaling slowing down and claimed that AI is, in fact, improving faster than ever thanks to new emerging scaling approaches and techniques.
Of course, it should be noted that the success of AI scaling is core to Nvidia’s business of selling huge quantities of GPUs to AI model developers and hosts.
-
Nvidia GTC 2025 live updates: Blackwell Ultra, GM partnerships, and two ‘personal AI supercomputers’
GTC, Nvidia’s biggest conference of the year, starts this week in San Jose. We’re on the ground covering all the major developments. CEO Jensen Huang will give a keynote address focusing on — what else? — AI and accelerating computing technologies, according to Nvidia.
We’re expecting Huang to reveal more about Nvidia’s next flagship GPU series, Blackwell Ultra, and the next-gen Rubin chip architecture. Also likely on the agenda: automotive, robotics, and lots and lots of AI updates.
Topics
June 5, 2025
Berkeley, California
LIMITED TIME: Save $300+ on your ticket, plus an additional 50% on a second for a full immersive day of AI! Hear from AI pioneers from Google DeepMind, OpenAI, Anthropic, and more on the main stage and in breakouts, and network like never before.