Source: Science and Technology Innovation Board Daily.
Author: Liu Rui.
At 10:30 AM Beijing time on January 7 (Tuesday), NVIDIA CEO Jensen Huang took the stage to deliver one of the most anticipated speeches at this year's CES.
In this approximately 90-minute presentation, Jensen Huang not only unveiled the latest generation of computer GPUs - the GeForce RTX 50 series but also brought many surprises regarding AI models, smart driving, AI robots, AI supercomputers, and more.
NVIDIA's first "big reveal": the super powerful GPU finally makes its debut.
At the beginning of his speech, Jensen Huang briefly reviewed the company's history and the development of AI before quickly throwing out the first "big reveal": the GeForce RTX 50 series GPUs.
Jensen Huang showcased the GeForce RTX 5090.
This series of GPUs includes the highly anticipated GeForce RTX 5090 - NVIDIA's most powerful graphics card to date, priced at $1999. The RTX 5090 claims to be the fastest graphics card in the world, with a speed that is twice that of its predecessor, the RTX 4090. It features 92 billion transistors and 4000 AI TOPS.
Additionally, NVIDIA also launched the GeForce RTX 5080, priced at $999; RTX 5070 Ti, priced at $749; RTX 5070, priced at $549. All of these graphics cards will start selling later this month.
NVIDIA stated that the RTX 50 series chips will support a feature called DLSS 4, which uses AI to improve game frame rates and can also display more details of characters' faces, providing users with better visuals and higher resolution.
Jensen Huang specifically mentioned that the performance of the RTX 5070 graphics card, priced at only $549, will be comparable to last year's RTX 4090 GPU, which was priced as high as $1600. Jensen Huang stated, "This would not be possible without artificial intelligence."
"We use GeForce to achieve artificial intelligence, and now AI is completely transforming GeForce. The new generation of DLSS can generate content beyond frames; it can predict the future," Jensen Huang stated during his keynote speech.
Jensen Huang also launched gaming laptops equipped with the aforementioned graphics cards, priced from $1299 to $2899. The laptops will start shipping in March this year.
Blackwell has fully entered production.
After showcasing the most significant GPU product launches, Jensen Huang used more of his speaking time to discuss broader aspects of artificial intelligence.
Jensen Huang first discussed the enormous demand for computing power in the era of AI development, which has driven the market's huge demand for NVIDIA's Blackwell.
He specifically held up an enlarged Blackwell GPU to explain - by the way, he made a gesture as if holding a shield while lifting the enlarged GPU, accompanied by uplifting background music, humorously for a brief moment.
Jensen Huang confirmed that despite some setbacks in 2024, the Blackwell chips have finally entered full production.
Blackwell has fully entered production.
Will AI become a digital workforce?
Next, Jensen Huang's topic shifted to AI agents. He played a video showcasing the tasks AI agents can assist with - from search assistants, factory operations to employee management, financial analysis, and more.
Jensen Huang bluntly stated, "This will be the next huge application of artificial intelligence, and AI agents 'could bring trillions of dollars in business opportunities.'"
He predicted that in the future, AI agents will become a digital workforce working alongside company employees, saying, "In many ways, the IT department of every company will resemble an HR department for AI agents!"
However, the audience's reaction to Jensen Huang's grand "vision" seemed lukewarm, and he jokingly remarked, "It seems this didn't leave a deep impression on you." He then shifted the topic to the next "vision" - Physical AI.
Jensen Huang officially announced the new Physical AI model, Cosmos.
Jensen Huang officially announced NVIDIA Cosmos, a world foundational model aimed at understanding the physical world.
Jensen Huang stated that the model has been trained on 20 million hours of video, aimed at "teaching AI to understand the physical world."
Jensen Huang stated that the Cosmos model will come in three sizes: Nano, Super, and Ultra.
He stated that developers can use Omniverse to create 3D scenes and then use Cosmos to convert them into realistic scenes. They can then generate multiple models simultaneously to help robots figure out the best ways to complete tasks.
Cosmos will be open-sourced on GitHub, and Jensen Huang expressed hope that it could be as influential as Llama 3.
Jensen Huang showcased the applications of Cosmos in factory management, autonomous driving, and more through video.
"In the future, every factory will have a digital twin, and each digital twin will operate exactly like the real factory."
AI empowers autonomous driving.
The topic then moved to autonomous driving. Jensen Huang revealed that NVIDIA's next-generation automotive processor product is named Thor, which has now entered full production. Its computing power is 20 times that of the previous generation automotive processor, Orin. Thor is also suitable for traditional robotics.
Jensen Huang showcased THOR.
Jensen Huang showcased his many automotive partners, including domestic giants BYD, Li Auto, Xiaomi, and Zeekr.
Jensen Huang also announced that Toyota will collaborate with NVIDIA to produce the next generation of autonomous vehicles. Jensen Huang stated:
"With the success of Waymo and Tesla, it is clear that autonomous vehicles have arrived."
NVIDIA's next-generation automotive chip product, Thor.
AI ROBOT.
Jensen Huang then discussed AI robots. He showcased over a dozen different companies' AI robots on the big screen and demonstrated how NVIDIA assists in training these AI robots:
"What are the key capabilities? How do you train these robots? It's easy for cars - just drive them - but demonstrating how humans work to robots is much more labor-intensive."
Jensen Huang unveiled Project Digits.
At the end of his speech, Jensen Huang showcased a super small NVIDIA AI supercomputer - Project Digits.
Jensen Huang showcased NVIDIA's AI supercomputer.
The computer is based on a new chip called GB110 - the smallest Blackwell GPU produced by NVIDIA. Jensen Huang revealed that GB110 is in production and being developed in collaboration with MediaTek.
Project Digits is also equipped with 20 Arm CPU cores. Jensen Huang expects Project Digits to be available around May. It can serve as a small workstation or be used in conjunction with existing PCs.
Usage scenarios for Project Digits.
