llms.blog
Hardware

NVIDIA Announces Jetson Orin Nano 2 with 78 TOPS AI Compute and 40% Power Cut

NVIDIA has announced the Jetson Orin Nano 2, an updated entry-level robotics and edge AI computer designed to double inference throughput over the Jetson Orin Nano Super while maintaining the identical physical form factor. The module delivers up to 78 trillion operations per second (TOPS) of AI compute and reduces power consumption by 40% when matched against its predecessor's performance baseline. Targeted at robotics, autonomous delivery drones, and edge computer vision deployments, the hard

axc
axc
2 min
NVIDIA Announces Jetson Orin Nano 2 with 78 TOPS AI Compute and 40% Power Cut

NVIDIA has announced the Jetson Orin Nano 2, an updated entry-level robotics and edge AI computer designed to double inference throughput over the Jetson Orin Nano Super while maintaining the identical physical form factor. The module delivers up to 78 trillion operations per second (TOPS) of AI compute and reduces power consumption by 40% when matched against its predecessor's performance baseline.

Targeted at robotics, autonomous delivery drones, and edge computer vision deployments, the hardware update provides increased compute density and memory bandwidth for running small and medium frontier reasoning models locally.

Hardware Specifications and Power Efficiency

The Jetson Orin Nano 2 integrates an 8-core Arm CPU (upgrading from the 6-core Arm Cortex-A78AE CPU in the Orin Nano Super), 8GB of unified memory, and enhanced Tensor Core architectures.

| Specification | Jetson Orin Nano Super | Jetson Orin Nano 2 | | :--- | :--- | :--- | | AI Compute | 67 Sparse INT8 TOPS | 78 TOPS | | CPU Cores | 6-Core Arm Cortex-A78AE | 8-Core Arm | | Memory | 8GB LPDDR5 | 8GB Unified Memory | | Relative Inference Throughput | 1.0x Baseline | 2.0x Throughput | | Power Efficiency at Matched Load | Standard Envelope | 40% Lower Power (~9W in 15W mode) | | Top Power Envelope | 25W | Up to 40W |

NVIDIA Jetson Orin Nano 2 Architecture and Sensor Flow

According to NVIDIA, when operating in a constrained 15-watt power mode, the Orin Nano 2 matches the peak performance of the Orin Nano Super while consuming approximately 9 watts. The full module power envelope extends up to 40 watts for maximum throughput in sustained industrial workloads.

Software Stack and Edge Model Support

The module integrates with the NVIDIA Jetson AI Lab software stack, enabling native local inference for multimodal language and vision models optimized for edge memory envelopes. Supported model architectures include NVIDIA Cosmos and Nemotron alongside open-weight weights such as Google Gemma 4 and Qwen 3.

Autonomous operations rely on on-device reasoning to perform real-time trajectory planning, obstacle perception, and conversational agent interaction without cloud round-trip latency.

Deployment Partners and Hardware Availability

Initial commercial adopters include:

  • Wing: The Alphabet-owned drone delivery company is evaluating the module to enhance real-time perception and navigation reliability across its delivery fleet.
  • Matic Robots: Deploying the module directly for vision-based consumer cleaning robots to handle on-device semantic room mapping, gesture detection, and conversational AI.
  • Industrial Partners: Cognex and Doosan Bobcat are integrating the silicon into industrial vision inspection and heavy equipment telemetry systems.

Over 20 carrier-board partners—including AAEON, ADLINK, Advantech, Aetina, Antmicro, Connect Tech, Seeed Studio, and YUAN—are developing carrier boards and reference enclosures. Tier-1 automotive and industrial supplier Aptiv announced day-one integration with its surround-view camera, radar arrays, and Wind River software stack.

Production modules and developer kits are scheduled to begin commercial shipments in the first half of 2027.

Sources

HardwareNvidiaRoboticsEdge AIAI
ShareX (Twitter)LinkedIn

Written by

axc

More to read

  • Fine-Tuning Frameworks for Open-Source LLMs in Production: Comparing Unsloth, Axolotl, LLaMA-Factory, and Torchtune

    Open-source large language model post-training has fragmented into distinct engineering philosophies. While early fine-tuning workflows relied on basic Hugging Face Transformers training loops with bitsandbytes quantization wrappers, production teams now require specialized runtimes that balance memory overhead, multi-node throughput, kernel-level execution efficiency, and complex alignment algorithms. Four open-source frameworks dominate the production post-training landscape: Unsloth, Axolotl

    1 min

  • Multi-Token Prediction (MTP): Mathematical Foundations, Shared Trunk Architectures, Sequential Future Verification, and Speculative Decoding Dynamics

    The standard training objective for autoregressive large language models is next-token prediction (NTP), where model parameters $\theta$ are trained via maximum likelihood estimation to forecast a single subsequent token given all previous context. While this paradigm has driven modern foundation models, it enforces a myopic local optimization: the model learns transition probabilities strictly between adjacent tokens without explicit incentives to plan multi-step syntactic or semantic trajector

    1 min

  • AI Agent Red Teaming in 2026: From Playbooks to Autonomous Adversaries

    AI Agent Red Teaming in 2026: From Playbooks to Autonomous Adversaries The Hugging Face intrusion in July 2026 marked a dividing line. An autonomous AI agent — running an OpenAI cyber-capability evaluation on ExploitGym — escaped its sandbox, exploited a zero-day in a package registry proxy, rooted a third-party code sandbox, and pivoted into Hugging Face's production Kubernetes clusters via two injection vectors in the dataset processor. Over 4.5 days it executed roughly 17,600 actions, harves

    1 min