The executive director of Nvidia, Jensen Huangpresented on the stage of the Ultra Arena in Las Vegas a new range of chips for video games and outlined one of the company’s new objectives: make the artificial intelligence (AI) improve to robots y autos without having to invest a lot of money.
Huang, backed by unstoppable sales from 2022, insisted this year at the CES technology fair that the world is going through a new industrial revolution where AI will be the new commodity (basic merchandise), urging both companies and governments to “build” a global digital infrastructure.
A future full of intelligent robots
“We think that the era of robotics is just around the corner”said the 61-year-old Taiwanese, who dressed in his usual uniform, black clothes and a leather jacket.
In that sense, the CEO recognized that his company developed a revolutionary AI model, named Cosmoscapable of generate realistic videos that can be used to train robots and autonomous vehicles in a cheaper way, to “democratize physical AI and make general robotics available to all developers.”
Huang explained that Cosmos was trained with 20 million hours of video and that models can help robots and cars understand “the physical world.”
However, he warned that Cosmos will need much more data and announced that the models will be available under a open model license to accelerate the work of the robotics and autonomous vehicle community.
The executive director also opted for a future full of autonomous vehicles. This gave way to the announcement of the alliance with the Japanese automaker Toyota, who will use its technology to promote theadvanced driver assistance in several models, but did not give details about the models.
Lastly, the chipmaker announced a $3,000 personal AI supercomputer called Digits that is powered by its new GB10 Grace Blackwell superchip.
“Blackwell, the AI engine, is here for PC gamers, developers and creatives,” said Huang. “By fusing AI-powered neural rendering and ray tracing, Blackwell is the most significant innovation in computer graphics since we introduced programmable shading 25 years ago,” he added.
The new RTX GeForce Series 50 GPUs: what models were presented and how much they cost
On the other hand, the event also had the presentation of the new gaming chips they use AI technology “Blackwell” from Nvidia to provide hyper-realistic graphics, which manage to reproduce almost perfectly everything from small details in objects and textiles to the faces of human beings.
The chips, called RTX 50will go on sale between January and February in the United States with a price that fluctuates between $549 and $1,999.
In this regard, Huang admitted that his $549 mid-range gaming chips They will be the same as the company’s previous flagship chip -the RTX 4090 that sells for $1,600-, news that caused great applause among the 14,000 attendees.
For desktop users, the GPU GeForce RTX 5090 with 3,352 AI TOPS and the GPU GeForce RTX 5080 with 1,801 AI TOPS will be available on January 30 for $1,999 and $999, respectively.
GPU day GeForce RTX 5070 Ti with 1,406 AI TOPS and the GeForce RTX 5070 GPU with 988 AI TOPS will be available starting in February for $749 and $549, respectively.
It must be taken into account that the TOPS or Tera Operations Per Second They indicate the speed at which a Neural Processing Unit (NPU) can perform trillions of mathematical operations, which are the basis of the calculations necessary to train and run AI models. The higher the TOPS value, the faster you can execute AI tasks.
Nvidia, in addition to positioning itself as the main supplier of chips with AI capabilities, managed to reach a milestone in its history in June as the largest listed in the worldalthough its value fluctuates and rivals Apple and Microsoft at the top.
The degree of relevance of this technology worldwide is such that at least a dozen governments buy processors from it to develop “sovereign” AI infrastructures, including the US, France, India, Singapore, Japan and the United Arab Emirates. .