The Nvidia Blackwell-powered T3000 and T2000 modules will help users move advanced robotics, visual AI and edge workloads onto compact, power-efficient systems. (Image: NVIDIA)Nvidia on Wednesday, July 15, introduced its T3000 and T2000 robot compute modules based on the Jetson AGX Thor architecture. The announcement comes at a crucial time when robotics is moving beyond research labs to real-world applications, accelerating the demand for power-efficient AI supercomputers that are capable of running foundational AI models.Considering the rising demand, the chipmaker has introduced the T3000 and T2000, which will enable scaling of mass-market robotics and edge AI applications. The company said that the new Jetson AGX Thor is powering the next generation of humanoid and robotic systems. Industry leaders like Agile Robots, Amazon Robotics, 1X, Boston Dynamics, Hitachi, and Techman Robot are building onto this platform.According to the company, the hardware that bolsters these capabilities begins with Jetson and IGX T3000 modules that deliver 865 FP4 teraflops (a computer performance metric) of AI compute in a compact form factor, about half the size and capacity of the Jetson T5000.The new T3000 module combines the Nvidia Blackwell GPU, which is an eight-core Neoverse Arm CPU, with 32 GB of LPDDR5X memory and 237 GB/s of memory bandwidth with 25 GbE connectivity. When it comes to performance, the IGX T3000 maintains efficiency with integrated functional safety while seamlessly running Nvidia Halos for Robotics’ full-stack safety system for robots working alongside humans.Also Read | Nvidia launches GeForce NOW in India, bringing high-end PC gaming to cloudEven though the T3000 has a smaller footprint when compared to the T5000, it achieves comparable inference performance for multimodal workloads, including large language models, vision language models, vision language action models and world foundation models. According to Nvidia, migrating to T3000 can reduce costs, a key factor, amid the soaring memory prices.The Jetson T2000, backed by Thor architecture, comes with 400 FP4 teraflops of compute and 16GB of memory and offers an entry point for developers working on visual AI agents, autonomous mobile robots, etc.Story continues below this adWhile the T3000 is aimed at advancing robotics and humanoids, the T2000 is designed to accelerate Edge AI. The Jetson Thor T3000 will be available in emulation mode with JetPack 7.2.1 and the T2000 will be available in a future release. Reportedly, both modules are expected to be launched in the first quarter of 2027.