Today, NVIDIA CEO Jensen Huang unveiled a series of AI-centric products at the CES exhibition in Las Vegas, ranging from AI Software that powers robots and self-driving Autos to a new type of AI supercomputer that can be placed on a desk, outlining NVIDIA's future vision while highlighting its potential for business expansion.

Ahead of this keynote speech, NVIDIA's stock price had already risen by 4.7% as Wall Street was fully prepared and eagerly anticipating the release of these new products. In the past 12 months, NVIDIA's stock price has increased by 205%, mainly due to its forward-thinking investments in AI Hardware and CUDA Software, allowing developers to run AI programs using its chips.

AI robots and self-driving Autos.

In today's speech, Jensen Huang announced the 'World Foundation Model' NVIDIA Cosmos, which can generate realistic videos for training robots and self-driving Autos at a cost much lower than using traditional data. By creating what the Technology industry calls 'synthetic' training data, these models can help robots and Autos understand the physical world, just as large language models (LLM) help chatbots generate responses in natural language.

Users only need to provide a text description, and Cosmos can generate a video that adheres to the laws of physics. Clearly, this is much cheaper than the current methods of data collection. Currently, to train self-driving Autos, companies need fleets driving on the streets to collect data, while humanoid robots require real humans to repeat tasks again and again for training.

Jensen Huang said: 'Like large language models, the World Foundation Model is the foundation for advancing robot and self-driving development, but not all developers have the expertise and resources to train their own models. We created Cosmos to democratize Physical AI, allowing every developer to utilize universal robots.'

Jensen Huang stated that the "ChatGPT moment" for robots is approaching, but before that, the Cosmos model needs more data. Cosmos will be offered under an "open license," similar to the widely used Meta Platforms' Llama3 language model. Huang said, "We really hope that Cosmos can serve as the Llama3 for the Robotics and Industrial AI field."

In addition to Cosmos, NVIDIA also showcased the Isaac GROOT Blueprint used for training humanoid robots. This Software connects to Apple's Vision Pro headset, allowing developers to perform and record specific actions they want to teach the robots. Then, the Isaac GROOT Blueprint synthesizes those actions, providing the robots with a vast set of movements based on the developers' original actions.

Previously, engineers had to teach humanoid robots to move by performing repetitive actions, which was quite time-consuming. Now, with the Isaac GROOT Blueprint, developers will be able to reduce the time needed to create future humanoid robot systems.

In the Automotive field, Jensen Huang announced that Toyota Motor will start using the company's DRIVE AGX Orin chip and NVIDIA DriveOS operating system to power advanced driver assistance features in its next-generation Autos.

NVIDIA also stated that agreements have been reached with automotive company Continental and self-driving truck company Aurora, which will use NVIDIA's DRIVE hardware and DriveOS software, along with Aurora's level 4 self-driving system, Aurora Driver. Continental and Aurora expect that their self-driving trucks will be on the road delivering goods by 2027.

Currently, NVIDIA's Automotive and Robotics department (A&R) accounts for only a small portion of its total revenue. In the third quarter of 2024, the department's revenue was $0.449 billion, while its Datacenter business revenue was $30.8 billion, making up a substantial part of NVIDIA's total revenue of $35.1 billion. Nevertheless, NVIDIA's Automotive and Robotics business is growing, with third-quarter sales increasing by 72% year-on-year.

NVIDIA AI Agent

Jensen Huang also announced a series of AI Software and Hardware products, including the AI Blueprint, which is an agent AI application that allows developers to build and deploy their own custom AI agents.

While generative AI excels in creativity and automation, agent AI functions more like a digital assistant, capable of handling workflows, solving problems, and providing real-time support in industries such as Human Resources, Software Engineering, and Medicine.

Jensen Huang stated in his speech today: "We officially enter the era of 'Agent AI'. AI agents are the new digital workforce. In the future, every company's IT department will become the Human Resources department for AI agents."

Jensen Huang also predicted that different agents will transform the way we work in the future. He also indicated that AI could create a market opportunity of one trillion dollars.

Currently, companies like Google and Microsoft have high hopes for AI agents, believing that it represents the next significant shift in Business and Consumer AI, as they can automate some daily tasks, such as importing information from emails into spreadsheets.

Small AI Supercomputers.

Today, Jensen Huang also unveiled a new micro AI supercomputer called Project Digits, which is based on the new GB10 superchip (Grace Blackwell Superchip) and is the smallest Blackwell GPU produced by NVIDIA.

Project Digits is equipped with 20 Arm CPU cores and 128GB of LPDDR5X memory, resembling a contemporary SoC, except it lacks I/O.

Jensen Huang stated that GB10 is being developed in collaboration with MediaTek and is currently in production, expected to launch in May this year. It can be used as a small workstation or in conjunction with existing PCs.

Jensen Huang said: "AI will become mainstream in every industry and every application. Through Project Digits, the Grace Blackwell Superchip will attract millions of developers. Putting AI supercomputers on every data scientist's, AI researcher's, and student's desk allows them to participate in and shape the AI era."

RTX Blackwell series GPUs

Today, Jensen Huang unveiled the RTX Blackwell series GPUs, offering breakthrough performance improvements at lower prices. The most notable is the GeForce RTX 5070, which Huang stated matches the performance of the previous RTX 4090 but is priced at just $549, significantly lower than the RTX 4090's launch price of $1599.

Jensen Huang attributed this cost-performance breakthrough to AI-driven efficiency. He stated: "This would not be possible without AI."

Jensen Huang stated that the RTX Blackwell series GPUs contain 92 billion transistors and G7 memory, with a bandwidth of up to 1.8TB per second, doubling the performance of the previous generation. Huang described this engineering achievement as a "miracle," emphasizing how the GPU handles both graphics and AI workloads simultaneously.

In addition to the RTX 5070, Jensen Huang also announced the RTX 5070 Ti ($749), RTX 5080 ($999), and the top model RTX 5090 ($1999). Notebooks using the RTX Blackwell GPU are expected to have a 40% increase in battery life, double the performance, while consuming half the power, with prices ranging from $1299 to $2899.

NVIDIA stated that the new RTX GPU introduces technologies such as RTX Neural Shaders and RTX Neural Face, which use AI to enhance rendering and create more realistic faces; while RTX Mega Geometry enhances environmental details, Reflex 2 reduces latency.

Another significant leap showcased at CES is DLSS 4, NVIDIA's latest AI upgrade, capable of generating multiple frames at once. In the demonstration, DLSS 4 rendered scenes at a speed of 247 frames per second (FPS), more than eight times faster than without AI, while keeping latency at 34 milliseconds.

Continuing the glory into 2025.

Clearly, after a still strong 2024, NVIDIA hopes to continue its momentum in 2025.

Daniel Newman, CEO of research firm Futurum Group, stated: "The Cosmos and other products released by NVIDIA at CES could bring huge returns for the company. Cosmos is a key driver for clients developing the next generation of robots, especially humanoid robots, and presents a market opportunity worth trillions of dollars over the next decade."

Rev Lebaredian, Vice President of Simulation Technology at NVIDIA, told the media: "For those autonomous vehicle and robotics companies with hundreds of millions of hours of data to organize, Cosmos is definitely a game-changing technology." Lebaredian also stated that NVIDIA has developed an ecosystem over the past decade composed of AI Software, Hardware, Robots, and Autonomous Vehicles Simulation Computing, laying the foundation for Cosmos.

As NVIDIA actively explores new markets, the company is facing increasing pressure from its largest customers, including Amazon and Microsoft, who are eager to build their own internal AI datacenter chips.

Analyst from Bank of America stated: "NVIDIA's decision to increase investment in 'Physical AI' is a logical next step. The challenge lies in making products reliable enough, cheap enough, and widespread enough to create a credible business model."

Regarding the performance of the new generation RTX Blackwell series GPUs, Jensen Huang conducted a live demonstration showcasing a series of highly detailed graphics with textures and movements. Gary Yang, a robotics graduate student from Caltech, said: "It's amazing that they can do this in real-time. We used to think these graphics were pre-rendered."

Scott Epstein from the tech startup Agenovate AI said: "I think this is incredible. They are continuing to innovate."

On Monday local time, NVIDIA's stock closed at a historic high of $149.43, bringing the company's market cap to $3.66 trillion, making it the second-largest company after Apple.