NVIDIA Jetson Thor Module: A Next-Gen AI Powerhouse for Robotics [2025]

NVIDIA Jetson Thor Module: Redefining Physical AI and Robotics

Introduction to NVIDIA Jetson Thor Module

The NVIDIA Jetson Thor Module (T5000) is a system-on-module (SOM) that pushes the boundaries of physical AI and robotics computing. With 2070 FP4 TFLOPS of AI compute, a 14-core Arm Neoverse-V3AE CPU, and the revolutionary Blackwell GPU architecture, it represents the most advanced embedded AI module available for developers and enterprises.

Designed from the ground up for next-generation humanoid robots, AI agents, and industrial edge AI, the Jetson Thor module brings together high performance, efficiency, and scalability in a compact form factor.

Key Features of Jetson Thor T5000

Unmatched AI Compute: 2070 TFLOPS

With 2070 FP4 TFLOPS, Jetson Thor can handle large language models (LLMs), vision-language models (VLMs), and Vision Language Action (VLA) models with ease.

Powered by Blackwell GPU with Transformer Engine

The 2560-core Blackwell GPU and 96 fifth-gen Tensor Cores provide transformer acceleration and Multi-Instance GPU (MIG) support for simultaneous AI workloads.

High-Capacity 128 GB LPDDR5X Memory

Developers gain 128 GB memory with 273 GB/s bandwidth, ensuring smooth operation of multi-sensor robotics, AI video analytics, and generative AI pipelines.

Technical Specifications at a Glance

CPU and GPU Performance

  • 14-core Arm Neoverse-V3AE CPU, 2.6 GHz max frequency
  • Blackwell GPU with MIG support
  • 96 Tensor Cores, 10 TPCs

Vision Accelerator and Video Capabilities

  • Programmable Vision Accelerator (PVA v3)
  • Supports 6× 4Kp60 encoding and 4× 8Kp30 decoding

Camera, Connectivity, and Networking

  • Supports up to 32 cameras via MIPI CSI-2 & virtual channels
  • 4× 25 GbE networking for high-speed data transmission

Storage and Expansion Options

  • NVMe SSD via PCIe Gen5
  • USB 3.2/USB 2.0 expansion for external storage and devices
NVIDIA Jetson T5000 - Technical Specifications
AI Performance2070 TFLOPS (FP4—sparse)
GPU2560-core NVIDIA Blackwell architecture, GPU with 96 fifth-gen Tensor Cores,
Multi-Instance GPU with 10 TPCs, 
GPU Max Frequency1.57 GHz
CPU14-core Arm® Neoverse®-V3AE 64-bit CPU
64 KB I-Cache, 64 KB D-Cache
1 MB L2 Cache per core
16 MB shared system L3 cache
CPU Max Frequency2.6 GHz
Vision Accelerator1x PVA v3
Memory128 GB 256-bit LPDDR5X, 273 GB/s
StorageSupports NVMe through PCIe,
Supports SSD through USB3.2
Video Encode6x 4Kp60 (H.265)
12x 4Kp30 (H.265)
24x 1080p60 (H.265)
50x 1080p30 (H.265)
48x 1080p30 (H.264)
6x 4Kp60 (H.264)
Video Decode4x 8Kp30 (H.265)
10x 4Kp60 (H.265)
22x 4Kp30 (H.265)
46x 1080p60 (H.265)
92x 1080p30 (H.265)
82x 1080p30 (H.264)
4x 4Kp60 (H.264)
CameraUp to 20 cameras via HSB
Up to 6 cameras through 16x lanes MIPI CSI-2
Up to 32 cameras using virtual channels
C-PHY 2.1 (10.25 Gbps),
D-PHY 2.1 (40 Gbps)
PCIe*Up to 8 lanes—Gen5
Root port only—C1 (x1) and C3 (x2)
Root Point or Endpoint—C2 (x1), C4 (x8), and C5 (x4)
USB*xHCI host controller with integrated PHY
3x USB 3.2
4x USB 2.0
Networking4x 25 GbE
Display4x shared HDMI2.1
VESA DisplayPort 1.4a—HBR2, MST
Other I/O5x I2S / 2x Audio Hub (AHUB), 2x DMIS, 4x UART, 4x CAN, 3x SPI, 13x I2C, 6x PWM outputs
Power40 W–130 W
Mechanical100 mm x 87 mm,
699-pin B2B connector,
Integrated Thermal Transfer Plate (TTP) with heatpipe

Jetson Thor vs Jetson Orin: Performance Leap

AI Compute Power Comparison

  • Thor: 2070 FP4 TFLOPS
  • Orin: ~275 FP8 TFLOPS
  • Result: >7.5× higher AI compute

Energy Efficiency Breakthroughs

Thor delivers 3.5× better energy efficiency, operating between 40W–130W, making it ideal for high-performance edge deployments.

Applications and Use Cases

Humanoid Robotics and Automation

Jetson Thor is engineered for humanoid robotics, enabling lifelike movement, AI-driven decision-making, and sensor fusion for autonomous navigation.

Generative AI Models and Agentic AI Workflows

From LLMs and VLMs to agentic AI video summarization and search, Thor supports the latest generative AI applications.

Edge AI for Industry, Healthcare, and Smart Cities

Industries can deploy Jetson Thor for:

  • Smart spaces and retail AI analytics
  • Industrial automation and predictive maintenance
  • Medical imaging and real-time diagnostics

Software Ecosystem Support

NVIDIA Isaac for Robotics

Helps developers with robotics simulation, reinforcement learning, and robot autonomy.

NVIDIA Metropolis for Smart Vision AI

Ideal for video analytics, surveillance, and smart city infrastructure.

NVIDIA Holoscan for Sensor Processing

Optimized for real-time medical and industrial sensor fusion workloads.

Developer Benefits and Ecosystem Support

  • Production-ready SOM with compact 100×87 mm size
  • Partner ecosystem offering carrier boards, sensors, and design services
  • Backed by NVIDIA AI software stack for seamless development

Where to Buy the NVIDIA Jetson Thor Module

  • Production-ready SOM with compact 100×87 mm size
  • Partner ecosystem offering carrier boards, sensors, and design services
  • Backed by NVIDIA AI software stack for seamless development

FAQs

Q1: What is the price of the NVIDIA Jetson Thor Module?
The price is ₹284,710.00 (excluding taxes).

Q2: When will the Jetson Thor Module ship?
It is expected to ship by October 14, 2025.

Q3: How does Jetson Thor compare to Jetson Orin?
Thor provides 7.5× more compute and 3.5× greater energy efficiency.

Q4: Can Jetson Thor handle generative AI models?
Yes, it is designed for LLMs, VLMs, and VLA models.

Q5: Is it suitable for industrial edge AI applications?
Absolutely—it supports retail AI, smart spaces, industrial robotics, and healthcare AI.

Q6: What kind of ecosystem support is available?
NVIDIA provides a rich partner ecosystem with carrier boards, sensors, and design services

Conclusion

The NVIDIA Jetson Thor Module sets a new benchmark in AI robotics and physical AI computing. With Blackwell GPU acceleration, 128 GB memory, and powerful CPU performance, it’s an all-in-one solution for developers building the next generation of humanoid robots and AI-driven edge devices.