NVIDIA's new robot &#34;brain&#34; was &#34;spoiled&#34; last weekend.
On Monday, the 25th, Eastern Time, NVIDIA announced the official release of the Jetson Thor developer kit Jetson AGX Thor, designed specifically for physical AI, and the mass production module Jetson T5000. This robotic computing platform has significantly improved compared to its predecessor, with AI computing power 7.5 times greater and energy efficiency 3.5 times better than the previous generation. The starting price for the developer kit is $3,499, while the unit price for purchasing more than 1,000 mass production modules is $2,999.
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/9702b96a-4ec4-4f8c-b415-a68f6769a91b.jpeg?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/9702b96a-4ec4-4f8c-b415-a68f6769a91b.jpeg"/>
NVIDIA stated that the new platform Jetson Thor, equipped with its Blackwell architecture GPU, unlocks real-time inference capabilities for physical AI. NVIDIA CEO Jensen Huang said:
<blockquote>
“Jetson Thor is designed for millions of developers worldwide, helping them build robotic systems that can interact with and even change the physical world. It offers unparalleled performance and energy efficiency, and can run multiple generative AI models simultaneously on edge devices. As an exceptional supercomputer, Jetson Thor is dedicated to truly ushering in the era of physical AI and general robotics.”
</blockquote>
Jetson Thor has been adopted by industry leaders such as Agility Robotics, Amazon Robotics, Boston Dynamics, and Figure. Jetson Thor provides up to 2070 FP4 teraflops (TFLOPS) of AI computing power within a power consumption of 130 watts, equipped with 128GB of memory, capable of running large generative AI models in real-time at the edge, meeting real-time processing demands.
NVIDIA's platform aims to address significant challenges in the robotics field—enabling robots to run multiple AI workflows in real-time and interact intelligently with humans and the physical world. Currently, over 2 million developers are using NVIDIA's Jetson platform and robotics technology stack, covering various industries including manufacturing, logistics, healthcare, and agriculture.
NVIDIA also announced that the DRIVE AGX Thor developer kit for autonomous vehicles is now available for pre-order, with shipments expected to begin in September.
<h2>Significant Performance Leap Unlocks Real-Time Inference Capabilities</h2>
Jetson Thor is built on NVIDIA's Blackwell architecture, achieving significant performance improvements compared to the previous Jetson Orin products. The platform's AI computing performance has increased to 7.5 times that of its predecessor, which is a 6.5 times improvement, CPU performance has increased to 3.1 times that of the previous generation, and memory capacity has doubled to 128GB
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/ae0c5971-92be-485d-a3d9-2ab24b0ebc1f.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/ae0c5971-92be-485d-a3d9-2ab24b0ebc1f.png"/>
This leap in performance enables robot developers to process high-speed sensor data and execute visual reasoning at the edge, a workflow that was previously unable to operate in dynamic real-world environments due to speed limitations. Jetson Thor supports a variety of generative AI models, from visual language action models like NVIDIA Isaac GR00T N1.5 to mainstream large language models.
The platform is specifically designed for generative reasoning models, capable of supporting next-generation physical AI agents driven by large transformer models and visual language models to run in real-time at the edge, minimizing reliance on the cloud.
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/a35e46ab-8b99-4536-aab2-eb23005d27f3.jpeg?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/a35e46ab-8b99-4536-aab2-eb23005d27f3.jpeg"/>
<h2>Complete Ecosystem Support Accelerates Commercialization Process</h2>
The Jetson Thor product series includes developer kits and production-grade modules. The developer kit includes the Jetson T5000 module, a reference carrier board with rich interfaces, an active cooler with a fan, and a power adapter.
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/bd5bb33b-cbef-4b7b-b869-2e7d33003340.jpeg?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/bd5bb33b-cbef-4b7b-b869-2e7d33003340.jpeg"/>
Jetson Thor is fully supported by the complete NVIDIA Jetson software platform, which is built for physical AI and humanoid robots, supporting any mainstream AI framework and generative AI model. The product is fully compatible with NVIDIA's software stack from cloud to edge, including Isaac for robot simulation development, the humanoid robot foundational model Isaac GR00T, the visual AI platform Metropolis, and the real-time sensor processing platform Holoscan.
Since its launch in 2014, the NVIDIA Jetson platform has attracted over 2 million developers and an ecosystem of more than 150 hardware system, software, and sensor partners. Hardware partners like Advantech, Aetina, and ConnectTech are developing production-ready Jetson Thor systems.
Research laboratories at Stanford University, Carnegie Mellon University, and the University of Zurich are leveraging Jetson Thor to push the boundaries of perception, planning, and navigation models. Carnegie Mellon University research associate professor Sebastian Scherer expects that upgrading to Jetson Thor will not only enhance AI model performance but also enable experiments with robot swarms.
<h2>Industry Giants Adopt Early, Broad Application Prospects</h2>
Several global technology leaders in the robotics field have begun adopting Jetson Thor to provide computational support for their next-generation robots
Yushu Technology founder and CEO Wang Xingxing stated, &#34;Jetson Thor brings a significant leap in computing power, empowering robots with greater agility, faster decision-making, and higher levels of autonomy, which is crucial for robots to navigate and interact in the real world.&#34;
Figure founder and CEO Brett Adcock mentioned that the server-level performance and compact energy-efficient design of Jetson Thor enable the deployment of large-scale generative AI models, allowing humanoid robots to perceive, reason, and act in complex unstructured environments.
Amazon Robotics Chief Technology Officer Tye Brady pointed out that the computing performance and energy efficiency provided by Jetson Thor are essential for developing and scaling the deployment of next-generation AI-driven robots. Caterpillar CEO Joe Creed stated that the AI performance of this platform meets the needs for developing and deploying future engineering and mining equipment.
In addition to humanoid robots, Jetson Thor will also accelerate various robotic applications such as surgical assistance robots, intelligent towing vehicles, delivery robots, industrial robotic arms, and visual AI agents, providing real-time inference capabilities for larger-scale and more complex AI models at the edge.
<h2>DRIVE AGX Thor Launched Simultaneously to Expand Autonomous Driving Applications</h2>
NVIDIA's DRIVE AGX Thor developer kit is now available for pre-order. This product is based on the Blackwell architecture and DriveOS 7 software stack, specifically designed for autonomous vehicles and mobile solutions, meeting the automotive industry's stringent functional safety (ISO 26262) and cybersecurity (ISO 21434) requirements.
Leading global automotive companies, including BYD, GAC, IM Motors, Li Auto, Volvo Cars, Xiaomi, and ZEEKR, are developing based on DRIVE AGX Thor. Autonomous truck companies Aurora, Gatik, PlusAI, and Waabi are also utilizing this platform.
The DRIVE AGX Thor production system can be obtained through tier-one suppliers Continental Automotive, Desay SV, Lenovo, Magna, and Quanta.
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/84cde4bf-1661-4a8d-91dc-06df9f8d2f1c.jpeg?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/84cde4bf-1661-4a8d-91dc-06df9f8d2f1c.jpeg"/>

NVDA

NVDL

07788

07388

NVDY

NVDD

NVDX

NVDQ

SOXL

Jetson Thor has an energy efficiency 3.5 times that of its predecessor, providing up to 2070 FP4 TFLOPS of AI computing power within a power consumption of 130 watts, with 128GB of memory, specifically designed for generative inference models to meet real-time processing needs at the edge. Its developer kit, Jetson AGX Thor, starts at $3,499. NVIDIA also announced that the DRIVE AGX Thor developer kit, based on the DriveOS 7 software stack and aimed at autonomous vehicles, is now open for pre-order, with shipments expected to begin in September

- NVIDIA has launched the Jetson AGX Thor developer kit and Jetson T5000 production module, significantly enhancing AI computing power by 7.5 times and energy efficiency by 3.5 times compared to previous models.  
- The platform, designed for physical AI, supports real-time inference and is already adopted by industry leaders like Agility Robotics and Amazon Robotics.  
- Additionally, the DRIVE AGX Thor developer kit for autonomous vehicles is now available for pre-order, catering to major automotive companies and meeting stringent safety standards.

Wallstreetcn

NVIDIA

GraniteShares 2x Long NVDA Daily ETF

XL2CSOPNVDA

XI2CSOPNVDA

YieldMax NVDA Option Income Strategy ETF

Direxion Daily NVDA Bear 1X ETF

T-Rex 2X Long NVIDIA Daily Target ETF

T-Rex 2X Inverse NVIDIA Daily Target ETF

Direxion Semicon Bull 3X

NVIDIA's new robot "brain" is here! Jetson Thor AI computing power increased by 6.5 times, unlocking real-time reasoning for physical AI