NVIDIA Introduces New Jetson Thor Computers to Advance Mainstream Robotics and Edge AI

2026-07-15 · Source: NVIDIA Blog · Field: Technology & Digital — Robotics & Autonomous Systems, Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Intermediate, short

Summary

NVIDIA has introduced the T3000 and T2000 modules, based on the NVIDIA Thor architecture, designed to power mass-market robotics and edge AI applications. The Jetson T3000 delivers 865 FP4 teraflops of AI compute, featuring an NVIDIA Blackwell GPU, an eight-core Neoverse Arm CPU, 32GB of LPDDR5X memory, 273GB/s bandwidth, and 25 GbE connectivity. The IGX T3000 offers similar performance with integrated functional safety. The Jetson T2000 provides 400 FP4 teraflops and 16GB of memory, serving as an entry point for visual AI agents and autonomous machines. Additionally, new Jetson agent skills automate memory optimization, reducing usage by up to 15GB for some applications, enabling deployment on lower-memory configurations like the 32GB module instead of 64GB. NVIDIA also expanded its Cosmos 3 frontier open world foundation model family with Cosmos 3 Edge, a 4-billion-parameter model for on-device inference. Developers can start building with emulation mode this month for T3000 and in Q1 2027 for module availability.

Key takeaway

For AI Engineers developing next-generation robotics or edge AI systems, NVIDIA's new Jetson T3000 and T2000 modules offer a scalable path to deployment. You can utilize the T3000's 865 FP4 teraflops for complex multimodal workloads or the T2000 for visual AI agents, reducing costs and power consumption. Employ Jetson agent skills to optimize memory, potentially moving to lower-memory configurations and accelerating your development timeline. Start development today using the available emulation modes.

Key insights

NVIDIA's new Thor-based modules and software tools enable scalable, power-efficient AI for mass-market robotics and edge applications.

Principles

• Compact AI supercomputers run foundation models at the edge.
• Software optimization reduces memory needs and system costs.
• Emulation modes accelerate development cycles.

Method

Developers can use Jetson agent skills to automate memory optimization and system configuration, then post-train Cosmos 3 Edge for specific embodiments and sensors in about a day.

In practice

• Deploy multimodal workloads on T3000 for cost reduction.
• Use T2000 for visual AI agents and industrial manipulators.
• Optimize memory with agent skills to use lower-memory Jetson SKUs.

Topics

• Jetson Thor
• Edge AI
• Robotics
• Foundation Models
• AI Agents
• Memory Optimization

Best for: Machine Learning Engineer, NLP Engineer, Computer Vision Engineer, Robotics Engineer, AI Engineer, MLOps Engineer

Related on AIssential

• NVIDIA Jetson Brings Agentic AI to the Physical World — NVIDIA Jetson now supports agentic AI via JetPack 7.2 and NemoClaw, enabling physical AI agents at the edge.
• Testing VLMs and LLMs for robotics w/ the Jetson Thor devkit — The Jetson Thor devkit offers high memory and low power for robotics inference, despite limited memory bandwidth.
• Deploying Open Source Vision Language Models (VLM) on Jetson — Deploying VLMs like Cosmos Reason 2B on NVIDIA Jetson devices enables real-time, interactive physical AI at the edge.
• Together AI at NVIDIA GTC 2026: Explore our latest innovations across research and products — Together AI and NVIDIA are advancing open, agentic, and production-ready AI systems through new model integrations and inference optimizations.
• Robots’ Holiday Wishes Come True: NVIDIA Jetson Platform Offers High-Performance Edge AI at Festive Prices — NVIDIA Jetson developer kits offer scalable edge AI compute for diverse robotics and physical AI applications.
• Nvidia Makes Its Biggest Robotics Push Yet at CES 2026 — Nvidia is building a full-stack ecosystem to power general-purpose robotics, from chips to open foundation models and simulation.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by NVIDIA Blog.