Nvidia AI guide guide with CES 2025 ads

Nvidia The general director and founder of Jensen Huang went to the stage to get a speech in CES 2025 To outline the vision of a company regarding the future of artificial intelligence in games, autonomous vehicles (AVS), robotics and others.

“AI is developing at an amazing pace,” said Huang. “It started with AI perception – understanding of images, words and sounds. Then generative artificial intelligence – creating text, images and sound. Now we enter the era of” physical AI “, AI, which can see, reason, plan and act.”

Thanks to NVIDIA and GPU platforms, Huang explained how the company continues to fuel a breakthrough in many industries, while presenting innovations such as the Cosmos platform, the new generation of the new generation GeForce RTX 50 Series and compact supercomputer projects.

RTX 50 series: “GPU is a beast”

One of the most important ads during CES 2025 was the introduction of the GeForce RTX 50 series, powered by NVIDIA Blackwell architecture. Huang debuted the flagship GPU RTX 5090, proud of 92 billion transistors and achieving an impressive 3352 trillion activities of artificial intelligence per second (peaks).

“GeForce has enabled AI to reach the masses, and now AI is coming home to GeForce,” said Huang.

Huang, holding a darkened GPU, called him “beast”, emphasizing his advanced functions, including double cooling fans and the ability to use AI for revolutionary graphics in real time.

The RTX 50 series, set to spread at the beginning of 2025, contains the flagship RTX 5090 and RTX 5080 (available on January 30), followed by RTX 5070 TI and RTX 5070 (February). The laptop GPU join the composition in March.

In addition, NVIDIA has introduced DLSS 4-to-the-rising “Multi-Rame Generation” technology-who increases the eighth time of games to eight times, generating three additional frames for each rendered frame.

Other progress, such as neural RTX and mega RTX geometry, promise increased realism in video games, including precise face and hair rendering using generative artificial intelligence.

Space: AI physical damage

Nvidia took another step forward with the Cosmos platform on CES 2025, which Huang described as “changing the game” for robotics, and industrial and AVS. Just like the influence of large language models on generative artificial intelligence, Cosmos is a new limit of the AI ​​application in robotics and autonomous systems.

“The moment of chatgpt for general robotics is right around the corner,” said Huang.

Cosmos integrates generative models, tokenizers and video processing frames to enable robots and vehicles to simulate potential results and predict optimal activities. By consuming hints of text, image and video, Cosmos can generate “virtual states of the world”, adapted to complex robotics and cases of use of AV covering real environments and lighting.

The best robotics and automotive leaders – including Xpeng, Hyundai Motor Group and Uber – are one of the first who accepted Cosmos, which is available at GitHub through an open license.

Velagapudi presses, CTO on agility, comments: “Data deficiency and variability are key challenges for successful learning in robot environments. Text, images, image and video to the world allow us to generate and expand photorealistic scenarios for various tasks that we can use to train models, without having to capture data.”

Strengthening AI models programmers

NVIDIA also presented new models of the AI ​​Foundation for RTX computers, which are aimed at supplementing the creation of the content, performance and application of the company. These models, presented as NVIDIA micros services (model of neural interaction), have been designed to integrate with the RTX 50 series equipment.

Huang emphasized the availability of these tools: “These AI models operate in every cloud because the NVIDIA GPU is now available in every cloud.”

Nvidia doubles its pursuit of programmers with advanced tools for building AI solutions. The company introduced AI plans: pre -configured tools for creating agents tailored to specific needs of the company, such as content generation, fraud detection and video management.

“They are completely open source, so you can take and modify their plans,” explains Huang.

Huang also announced the release of Lama Nemotron, designed for programmers for the construction and implementation of powerful AI agents.

Ahmad al-Dahle, vice president and boss of Genai in Meta, said: “Agentic AI is another limit of the development of artificial intelligence, and ensuring this possibility requires optimization of a full stack in the LLM system to ensure efficient, accurate AI agents.

“Thanks to our cooperation with NVIDIA and our joint commitment to open models, the Nvidia Lama Nemotron family built on Llam can help enterprises quickly create their own AI agents.”

Philipp Herzig, director of AI at SAP, added: “AI agents who cooperate to solve complex tasks in many company lines, will unlock a completely new level of company performance outside today's AI scenarios.

“Through Joule SAP, hundreds of millions of company users will interact with these agents to achieve their goals faster than ever before. The new open model family of Llama Nemotron will be conducive to the development of many specialized AI agents to transform business processes.”

Safe and smarter autonomous vehicles

Nvidia's ads have expanded to the automotive industry, where its Drive Hyperion AV platform is conducive to AVS. Built on the new NVIDIA AGX Thor System-On-Chip (SOC), the platform allows vehicles to achieve functional safety and autonomous possibilities using generative AI models.

“The autonomous vehicle revolution is here,” said Huang. “Building autonomous vehicles, such as all robots, requires three computers: NVIDIA DGX for training AI models, omniverse for testing and generating synthetic data and running AGX, a supercomputer in the car.”

Huang explained that synthetic data are of key importance for AV development, because it dramatically improves real data sets. AI Nvidia data factories – powered by Omniverse and Cosmos platforms – generate synthetic driving scenarios, increasing the effectiveness of training data.

Toyota, the world's largest car manufacturer, is involved in the use of NVIDIA DRIVE AGX Orin and certified by the safety of NVIDIA DRIVEOS to develop new generation vehicles. Heavy, such as JLR, Mercedes-Benz and Volvo Cars, also accepted Hyperion.

Project numbers: compact supercomputer AI

Huang ended his nvidia speech at CES 2025 with the final announcement of “One More Thing”: Project Digits, the smallest but most powerful AI Nvidia supercomputer, powered by the most modern GB10 Grace Blackwell Superchip.

“This is the latest AI nvidia supercomputer,” Huang said, revealing his compact size, claiming that it is portable enough to “practically fit into your pocket.”

Number of projects enable programmers and engineers to train and implement AI models directly from their desks, ensuring the full power of AI Nvidia stack in a compact form.

In May, design numbers represent NVIDIA pressure to share AI supercomputer with both people and organizations.

Vision of tomorrow

Reflecting on Nvidia's journey since the invention of the programmable GPU in 1999, Huang defined the last 12 years of AI changes as a transformation.

“Each layer of technological stack has been fundamentally transformed,” he said.

Thanks to game progress, agents based on AI, robotics and autonomous vehicles, Huang predicts an exciting future.

“All the technologies of the contract I was talking about today will lead to surprising breakthroughs in general robotics and AI in the coming years,” sums up Huang.

(Image loan: Nvidia)

See also: Altman itself, Opennai: “Happy and humiliating” to work in the direction of superintelligence

Do you want to learn more about AI and Big Data from industry leaders? Check out AI and Big Data Expo It takes place in Amsterdam, California and London. A comprehensive event is located with other leading events, including Intelligent automation conferenceIN BlockxIN Digital Transformation WeekAND Cyber ​​security and clouds.

Browse other upcoming technological events and internet seminars driven by TechForge Here.