There was news that Nvidia is specializing in the robotics field to cut back its dependence on GPUs.
The Financial Times (FT) reported on the twenty ninth (local time) that NVIDIA plans to release ‘Jetson Thor’, a small computer for next-generation humanoid robots, in the primary half of 2025. Jetson Thor is an edge computer that performs AI tasks in fields akin to robots, drones, and autonomous driving.
NVIDIA said that a boom in robotics is imminent and that it’s competing to grow to be one of the best platform on this field. “A revolutionary moment like ChatGPT is coming soon in physical AI and robotics,” said Deeppu Tala, vice chairman of robotics at NVIDIA. “The market has reached a tipping point.”
This move is analyzed as a technique to cut back dependence on GPUs as competition intensifies, with not only chip manufacturing competitors akin to AMD, but additionally cloud giants akin to Amazon, Microsoft (MS), and Google developing their very own AI chips.
Moreover, starting this 12 months, we’re actively investing within the ‘Physic-AI’ field to support robot startups. Last February, it participated in an investment price $2.6 billion (roughly 3.8 trillion won) within the humanoid robot startup Figure together with Microsoft and Open AI.
Although robot-related sales figures aren’t disclosed, it is thought that they still account for a small portion of total sales. The information center sector accounted for about 88% of third quarter sales of $35.1 billion (roughly 51.7 trillion won).
NVIDIA provides an integrated full-stack solution starting from AI-based robot training software to chips, the core components of robots. It includes ▲foundation model training software using the ‘DGX’ system ▲real environment simulation on the ‘Omniverse’ platform ▲hardware that acts as a brain mounted contained in the robot. This helps robots learned in a virtual environment to operate easily in the actual environment.
“Over the past 12 months, the gap between simulations and the actual world has narrowed significantly, in order that experiments that were inconceivable two years ago can now be performed in simulations combined with generative AI,” said Tala. “We’re making these experiments and tasks possible. “We’re providing a platform to accomplish that,” he emphasized.
In keeping with U.S. market research firm BCC, the worldwide robot market is currently valued at roughly $78 billion (roughly KRW 115 trillion) and is anticipated to grow to $165 billion (roughly KRW 243 trillion) by 2029.
Reporter Park Chan cpark@aitimes.com
