Nvidia (NVDA) aims to extend its successful trajectory from 2024 into 2025 with several AI-focused announcements made at CES in Las Vegas on Monday. During the company’s keynote address, CEO Jensen Huang articulated his vision encompassing a range of applications, from AI software designed for robots and autonomous vehicles to a compact AI supercomputer suitable for desktop use.
Ahead of the keynote, Nvidia’s stock experienced a surge of up to 4.7%, as investors anticipated the latest innovations from this AI frontrunner. Over the past year, the company’s stock has soared by an impressive 205%, driven by its strategic investments in AI hardware and the CUDA software, which enables developers to utilize its chips for AI applications.
The recent announcements highlighted how developers can leverage Nvidia’s current hardware, specifically its Hopper and Blackwell platforms. The company is expected to unveil its next-generation chip at the upcoming GTC conference in March.
During the event, Huang introduced Nvidia’s latest Blackwell-based chip, the GB10 superchip. This compact version of the GB200 superchip integrates a Grace central processing unit (CPU) with two Blackwell graphics processing units (GPUs). The GB10 features a single Grace CPU paired with a Blackwell GPU.
Nvidia has indicated that this chip will be incorporated into a small desktop system known as Project DIGITS, which will offer 128GB of memory and 4TB of storage. The company asserts that this configuration is sufficiently robust for researchers engaged in “prototyping, fine-tuning, and running large AI models.”
Project DIGITS is set to launch at a starting price of $3,000 in May, available through Nvidia and its OEM partners.
In addition to the new chip and desktop system, Nvidia unveiled its open model license Cosmos platform, designed for the development of physical AI systems. This platform utilizes world foundation models (WFMs), which are AI models that replicate real-world conditions. Physical AI systems encompass technologies such as humanoid robots and autonomous trucks.
Companies are encouraged to utilize Cosmos to facilitate the development of software essential for operating robots and autonomous vehicles. This can be achieved by simulating various operational scenarios in a virtual environment, thereby eliminating the need for expensive robots or the risks associated with testing vehicles on public roads.
Huang remarked, “The ChatGPT moment for robotics is approaching.”
He elaborated, “Similar to large language models, foundational models for the world are crucial for the progress of robotics and autonomous vehicle technology. However, not all developers possess the necessary expertise and resources to train these models independently. We designed Cosmos to democratize physical AI, making general robotics accessible to every developer.”
In conjunction with Cosmos, Nvidia introduced its Isaac GROOT Blueprint, a tool for training humanoid robots. This software integrates with Apple’s Vision Pro headset, enabling developers to execute and record specific movements intended for robotic instruction. The Isaac GROOT Blueprint then synthesizes these movements, equipping the robot with a vast repertoire of actions derived from the developer's original gestures.
Typically, engineers must teach humanoid robots through repetitive actions that the robot can learn to replicate autonomously. This process is often labor-intensive; however, Nvidia asserts that the Isaac GROOT Blueprint will significantly reduce the time required to develop future humanoid robotic systems.
On the automotive side, Nvidia announced that Toyota will implement the company’s DRIVE AGX Orin chip and the Nvidia DriveOS operating system to enhance the advanced driver assistance capabilities in its upcoming vehicle models.
Additionally, Nvidia disclosed a partnership with automotive manufacturer Continental and self-driving truck company Aurora. This collaboration will enable these transportation firms to utilize Nvidia’s DRIVE hardware and DriveOS software in conjunction with Aurora’s level 4 autonomous driving system, known as Aurora Driver. Continental and Aurora aim to introduce autonomous trucks for freight transport on public roads by 2027.
Post a Comment