NVIDIA Jetson T5000 AI Computing Module with 2070 TOPS Performance
The NVIDIA Jetson T5000 is the ultimate platform for physical AI and robotics, delivering up to 2070 FP4 TFLOPS of AI compute power and 128 GB of LPDDR5X memory. Powered by the advanced Blackwell GPU architecture with 2560 CUDA cores and 96 fifth-generation Tensor Cores, this system-on-module (SoM) provides 7.5 times higher AI performance than the previous generation NVIDIA AGX Orin, with 3.5 times better energy efficiency. Designed for humanoid robotics, autonomous machines, edge AI deployment, and sensor-intensive applications, the Jetson T5000 runs generative AI models including vision language action (VLA) models, large language models (LLMs), and vision language models (VLMs) directly at the edge.
Key Features
-
Blackwell GPU Architecture: 2560 CUDA cores and 96 Tensor Cores deliver up to 2070 FP4 TFLOPS of AI compute, supporting the latest generative AI models with Multi-Instance GPU (MIG) capability
-
High-Performance CPU: 14-core ARM Neoverse V3AE processor at 2.6 times the performance of previous generation cores for real-time, low-latency multi-sensor processing
-
Massive Memory: 128 GB of 256-bit LPDDR5X memory with 273 GB/s bandwidth for handling complex AI workloads
-
Power Configurable: Operates between 40 W and 130 W, balancing performance with energy efficiency for diverse deployment scenarios
-
Advanced Networking: 4 × 25 GbE and 1 × 5 GbE connectivity for high-speed sensor fusion and distributed AI applications
-
PCIe Gen5 Support: 12 lanes of PCIe Gen5 (32 GT/s per lane) for ultra-fast storage and peripheral connectivity
-
Integrated Accelerators: Third-generation Programmable Vision Accelerator (PVA v3.0), dual video encoders and decoders supporting multiple 4K and 8K streams, and optical flow accelerator
-
Compact Form Factor: 100 mm × 87 mm module size, compatible with existing AGX carrier board ecosystem
Designed for Physical AI and Robotics
The Jetson T5000 belongs to a new class of robotic computers architected from the ground up to power next-generation humanoid robots and autonomous systems. It supports the NVIDIA Isaac platform for robotics, NVIDIA Metropolis for visual agentic AI, and NVIDIA Holoscan for sensor processing. The module effortlessly runs foundation models like NVIDIA Isaac GR00T N1.5, enabling developers to build, secure, and deploy autonomous agents directly at the edge without relying on cloud connectivity.
Enterprise-Grade AI Software Stack
Jetson T5000 runs the complete NVIDIA AI software stack for physical AI applications, providing seamless cloud-to-edge integration. The platform supports a wide range of popular open-source models through Jetson AI Lab, including Llama, Gemini, and Qwen. Advanced memory optimization unlocks multi-billion-parameter edge AI by reclaiming up to 12 GB of memory, powering real-time robots and physical AI on compact, energy-efficient devices.
High-Speed Connectivity and I/O
Built for sensor-heavy applications, the Jetson T5000 accelerates low-latency, real-time multi-sensor processing with extensive I/O options for sensor fusion. The module includes QSFP28 interfaces, MIPI camera inputs, HDMI 2.0b and DisplayPort 1.4a video outputs, multiple USB ports, and M.2 slots for NVMe storage expansion. WiFi 6E and Bluetooth support (when paired with appropriate carrier boards) enable wireless connectivity for modern IoT and edge AI deployments.
Applications Across Industries
-
Humanoid Robotics: Next-generation robots with real-time perception, decision-making, and motor control
-
Autonomous Vehicles: Centralized computing for autonomous driving, intelligent cockpit systems, and in-vehicle infotainment
-
Manufacturing and Logistics: Multiple concurrent AI inference pipelines for quality inspection, predictive maintenance, and warehouse automation
-
Healthcare: Medical imaging, surgical robotics, and real-time patient monitoring systems
-
Retail: Visual AI for inventory management, customer analytics, and autonomous checkout systems
-
Smart Cities: Edge AI for traffic management, surveillance, and infrastructure monitoring
What's in the Box
- NVIDIA Jetson T5000 Module (900-13834-0080-000)
- Documentation and quick start guide
Note: Developer kit, carrier board, cooling solution, and accessories sold separately.
Full Specifications
GPU
| Architecture |
NVIDIA Blackwell |
| CUDA Cores |
2560 |
| Tensor Cores |
96 (5th Generation) |
| AI Performance |
2070 FP4 TFLOPS (Sparse), 7.8 FP32 TFLOPS |
| Multi-Instance GPU |
Supported |
| Transformer Engine |
Yes |
CPU
| Processor |
14-core ARM Neoverse V3AE (ARMv9) |
| Performance Improvement |
2.6× vs. previous generation |
Memory
| Type |
LPDDR5X |
| Capacity |
128 GB |
| Interface |
256-bit |
| Bandwidth |
273 GB/s |
| Clock Speed |
1.07 GHz |
AI Accelerators
| Programmable Vision Accelerator |
PVA v3.0 |
| Video Encoders |
Dual encoders (4K/8K multi-stream support) |
| Video Decoders |
Dual decoders (4K/8K multi-stream support) |
| Optical Flow Accelerator |
Yes |
Networking
| High-Speed Ethernet |
4 × 25 GbE |
| Standard Ethernet |
1 × 5 GbE |
I/O Interfaces
| PCIe |
12 lanes Gen5 (32 GT/s per lane) |
| Display Outputs |
HDMI 2.0b, DisplayPort 1.4a |
| Camera Inputs |
MIPI CSI support |
| USB |
Multiple USB ports (configuration depends on carrier board) |
| Storage Interface |
NVMe via PCIe, M.2 expansion support |
Power
| Power Range |
40 W to 130 W (configurable) |
| Typical TDP |
40 W minimum, 130 W maximum |
Physical
| Module Dimensions |
100 mm × 87 mm |
| Weight |
0.28 kg (module only) |
| Manufacturing Process |
3 nm |
Software Support
| AI Frameworks |
NVIDIA Isaac, NVIDIA Metropolis, NVIDIA Holoscan, Jetson AI Lab |
| Supported Models |
VLA (Vision Language Action), LLM (Large Language Models), VLM (Vision Language Models) |
| Foundation Models |
NVIDIA Isaac GR00T N1.5, Llama, Gemini, Qwen |
| Operating System |
NVIDIA Jetson Linux (based on Ubuntu) |
What's in the Box
| Module |
NVIDIA Jetson T5000 (SKU: 900-13834-0080-000) |
| Documentation |
Quick start guide and reference materials |
Specifications are sourced from the manufacturer product page and may be updated. For complete system integration, a compatible carrier board is required (sold separately). The NVIDIA Jetson AGX Thor Developer Kit includes the T5000 module with reference carrier board, WiFi 6E module, 1TB NVMe SSD, and complete I/O connectivity.
