I do not think that it is an understatement to say that Nvidia wants to be the company that feeds the universe of artificial intelligence in which we can one day live permanently.
The opening speech of NVIDIA GTC 2025, was colloquially called the Woodstock of the AI, saw the CEO Jensen Huang praising the virtues of an AI ecosystem propelled by the hardware – and the software – of the second most precious company in the world.
We have been reintroduced to the concept of AI factories, which are essentially gigantic data centers and where, the world of Nvidia, the tokens generated are equivalent to income: the presentation mentioned a price, 1 billion of dollars.
The cost of building data centers, such as mega-aggressions, increases exponentially in each generation as they become more complex and require more energy to shelter even more calculation.
The Stargate noble project has a budget of $ 500 billion and Apple has already undertaken to spend $ 500 billion in the coming years, a large part of which will be in data centers.
According to analysts you are talking about, capital expenses in the world data center should reach 1 dollars billion in 2027 (PWC) or 2029 (Dell’oro).
This therefore gives you a measure of the extremely ambitious targets of Nvidia for herself and for humanity.
A large piece of this puzzle that Nvidia introduced at the disposal is its photonic switching systems which will allow these AI factories to “develop millions of GPUs”, apparently in the same physical perimeter.
Large number
The greatest number of them in this presentation was when the vice-president, omniverse and the technology of simulation of the reverend nvidia, omniverse and simulation, presented a slide on the way in which physical AI transforms 50 billions of dollars into industries.
These extend over two billion cameras, 10 million factories, two billion vehicles and, Lebareian added “future billion humanoid robots”.
Yes, these are not only robots, but humanoid robots, millions of them. Robots, in general, have existed for decades in many forms and forms.
NVIDIA, however, envisages a world where billions of robots are built on what he calls, his “three computers”: pre and post-training (fact via DGX), simulation and generation of synthetic data (made via omaverse) and execution (made using AGX).
The world of the future could house “several types of humanoid robots” which can be post -train – while being, so to speak – with real or synthetic data, which allows them to learn new skills or improve their knowledge base on the fly, literally to (or near) the speed of light.
And the first could well be based on its new GR00T N1 Humanoid Robot Foundation model, developed in partnership with Google Deepmind and… Disney Research.
(ED: We cannot see the link with GR00T by Guardians of the Galaxy and Wall-E, a potentially premonitory film on the impact of robotics and AI on humanity).
Where does that leave us? With a clear roadmap towards a world where AI becomes an essential part of what a human being means, perhaps the most important utility of all. Never, in the history of humanity, if many depend so little.
Oh and the other acronym (AG) was not mentioned during the presentation, not once.
