Announced, not yet shipping. NVIDIA introduced the Jetson Orin Nano 2 in September 2026 and expects the production module and the developer kit to be available in the first half of 2027. Pricing has not been published. Use Notify Me and we will contact you as soon as it can be ordered. Product images show the current Jetson Orin Nano Super Developer Kit, which shares the same form factor; we will update them at launch.
Overview
The NVIDIA Jetson Orin Nano 2 is the next generation of NVIDIA's entry-level robotics computer. Built around a new Orin-class system on chip with an NVIDIA Ampere GPU and upgraded Tensor Cores, an 8-core Arm Cortex-A78 CPU and 8GB of LPDDR5X memory, it delivers up to 78 TOPS of AI compute and, according to NVIDIA, twice the inference throughput of the Jetson Orin Nano Super in the same footprint.
The gains come as much from the memory subsystem as from raw compute: LPDDR5X raises bandwidth to 120 GB/s, which lets compact foundation models such as NVIDIA Cosmos, NVIDIA Nemotron, Gemma 4 and Qwen 3 run responsively on the device. In its 15W mode the Orin Nano 2 uses about 40 percent less power than its predecessor at equivalent performance, and a new 40W ceiling is available when peak throughput matters more than efficiency.
Orin Nano 2 runs NVIDIA's open software stack, including JetPack, Isaac and Metropolis, and supports the new Jetson agent skills for building agentic workflows at the edge. More than 20 ecosystem partners, among them AAEON, ADLINK, Advantech, Aetina, Antmicro, Aptiv, Auvidea and AVerMedia, are preparing carrier boards and reference systems, and early adopters such as Cognex, Doosan Bobcat, Matic, Wing and Amazon Robotics are already evaluating the platform.
Key Features
- 78 TOPS AI Performance: Ampere GPU with upgraded Tensor Cores, roughly 2x the real-world inference throughput of Jetson Orin Nano Super
- 8-core Arm Cortex-A78 CPU: Two more cores than the current Orin Nano for perception, planning and application logic
- 8GB LPDDR5X at 120 GB/s: The first Orin product with LPDDR5X, 18 percent more bandwidth than Orin Nano Super
- 15W to 40W: A more efficient 15W mode plus a new 40W peak mode
- Same Form Factor: Retains the Jetson Orin Nano footprint and connector, so existing carrier designs carry over
- Generative AI at the Edge: Runs compact LLMs and vision language models locally, without cloud dependency
- Open Software Stack: JetPack, Isaac, Metropolis, Cosmos and Jetson agent skills
- Module and Developer Kit: Both variants expected in the first half of 2027
Technical Specifications
Figures marked with an asterisk are reported by NVIDIA in the announcement; the remaining details are based on early press coverage and will be confirmed when the datasheet is published.
| Specification | Details |
|---|---|
| AI Performance | 78 TOPS (INT8, sparse)* |
| GPU | NVIDIA Ampere architecture with upgraded Tensor Cores*; 1,536 CUDA cores and 32 Tensor Cores (reported) |
| CPU | 8-core Arm Cortex-A78* |
| Memory | 8GB LPDDR5X*, 120 GB/s* |
| Power | 15W to 40W* |
| Efficiency | About 40 percent less power than Orin Nano Super at equivalent performance in 15W mode* |
| Inference Throughput | Up to 2x Jetson Orin Nano Super* |
| Form Factor | Same footprint as Jetson Orin Nano (260-pin SO-DIMM)* |
| Software | NVIDIA JetPack, Isaac, Metropolis, Cosmos, Jetson agent skills* |
| Availability | Module and developer kit, first half of 2027* |
Jetson Orin Nano 2 vs Orin Nano Super
| Specification | Jetson Orin Nano 2 | Jetson Orin Nano Super |
|---|---|---|
| AI Performance | 78 TOPS | 67 TOPS |
| Inference Throughput | Up to 2x | 1x (baseline) |
| CPU | 8-core Arm Cortex-A78 | 6-core Arm Cortex-A78AE |
| GPU | Ampere, upgraded Tensor Cores, 1,536 CUDA cores (reported) | Ampere, 1,024 CUDA cores, 32 Tensor Cores |
| Memory | 8GB LPDDR5X | 8GB LPDDR5 |
| Memory Bandwidth | 120 GB/s | 102 GB/s |
| Power | 15W to 40W | 7W to 25W |
| Availability | First half of 2027 | Shipping now |
Ideal Use Cases
- Home and service robots with conversational AI and gesture recognition
- Delivery and inspection drones that need real-time perception and obstacle avoidance
- Vision AI systems for industrial inspection, retail and smart spaces
- Autonomous mobile platforms running local language and vision language models
- Educational and research projects on the next entry-level Jetson
Why Choose This Product?
Jetson Orin Nano 2 doubles the useful inference performance of the most popular entry-level Jetson while keeping its form factor, so products designed around Orin Nano can move up without a redesign. If you are planning a 2027 launch, register your interest now and we will confirm pricing and lead time as soon as NVIDIA releases them. If you need hardware today, the Jetson Orin Nano Super Developer Kit is available from stock.
Click Notify Me to be contacted when Jetson Orin Nano 2 becomes available to order.
