Prev
Next
AIE-KT78/68
• Next-Gen AI Performance: Powered by NVIDIA® Jetson T5000™ / Jetson T4000™, delivering up to 2070 TFLOPS (FP4)
• High-Speed Sensor Fusion: Supports QSFP28 (up to 100 Gbps) and 8x GMSL2 cameras for advanced vision AI
• Industrial Connectivity: Equipped with dual 10G bE, 1GbE LAN support EtherCAT and 4x USB 3.2 Gen 2 ports
• Rugged Reliability: Built for harsh deployments with 9-48VDC wide-range input and -25°C to +55°C operating temperature
 |
Overview
Unleash Edge Supercomputing for the Next Era of Cobots and Humanoids
Supercharge Automation ROI by Deploying Multi-Modal Decision Models and Massive Sensor Fusion to the Edge
As next-generation Autonomous Mobile Robots (AMRs), humanoid robotics, and high-definition vision systems evolve, standard edge processors delivering only a few hundred TOPS are hitting their limits. Developers face critical bottlenecks when fusing high-resolution multi-sensor pipelines while maintaining deterministic, real-time control (such as EtherCAT Master communication) across external actuators.
The AIE-KT78/68 is Aetina's premium Leading Edge AI Flagship System, built to eliminate these barriers. Powered by the NVIDIA® Jetson Thor™ module (Jetson T5000 / Jetson T4000), it delivers server-class Blackwell AI performance directly to the field—enabling on-device multi-modal generative AI, zero cloud latency, and absolute data security.
By combining extreme computing with a high-bandwidth communication interface and a ruggedized industrial layout, the AIE-KT78/68 slashes integration cycles and dramatically reduces system TCO. This capability is particularly game-changing for Collaborative Robots (Cobots). By executing advanced Vision-Language-Action (VLA) models completely offline, it transitions cobots from rigid machines into intelligent cognitive partners that perceive, collaborate safely, and adapt to new tasks instantly without expensive reprogramming downtime.
Available configurations:
- AIE-KT78, powered by NVIDIA Jetson T5000, delivers up to 2070 FP4 TFLOPS
- AIE-KT68, powered by NVIDIA Jetson T4000, delivers up to 1200 FP4 TFLOPS
Product Benefits
| Product Benefits | Description |
|---|---|
| Server-Class NVIDIA Blackwell Architecture |
Powered by the NVIDIA Jetson Thor, delivering up to 2070 TFLOPS of FP4 compute power. This enables cobots and autonomous systems to run advanced Vision-Language-Action (VLA) models completely offline with zero latency. |
| Ultra-High Bandwidth Networking |
Features a built-in 100Gbps QSFP28 slot and dual 10GbE ports to seamlessly ingest and process massive real-time data streams from high-definition cameras and 3D LiDARs, eliminating multi-sensor fusion bottlenecks. |
| Deterministic EtherCAT Integration |
Equipped with a dedicated RJ45 1GbE EtherCAT port acting as an independent Master. It bridges high-level AI reasoning with low-level physical motors and joint actuators with microsecond-level synchronization. |
| Rugged Wide-Temp & Wide-Voltage | Built for harsh industrial environments, supporting 9 to 48V wide-range input and -25°C to +55°C operating temperatures. This guarantees 24/7 stable operation under full workloads regardless of voltage drops or extreme conditions. |
| Next-Gen I/O & Sensor Expansion |
Includes 4x USB 3.2 Gen2, 2x isolated CAN FD (exclusive to AIE-KT78), and mini-Fakra ports, with flexible M.2 slots for NVMe storage and 5G / LTE connectivity, creating a complete sensor fusion hub. |
Application Scenario
| Application | Description |
|---|---|
| Collaborative Robots (Cobots) | ▪ Instant Line Reconfiguration: Staff can instruct cobots via natural speech to adjust tasks. This slashes setup downtime from days to seconds, making high-mix, low-volume production highly profitable. ▪ Collision-Free Shared Workspaces: Microsecond-level reaction times ensure robotic joints stop or adapt before physical contact, protecting workers and eliminating safety liabilities. ▪ Reduced Hardware TCO: Combines AI computing and real-time motion control into one unit, saving thousands of dollars in external controllers and cabinet space. |
| Humanoid Robotics |
▪ Zero-Adaptation Workforce Integration: Unlike traditional wheeled AGVs or static arms, humanoids fit directly into existing factory and warehouse layouts designed for humans. You do not need to spend millions of dollars redesigning your facility, aisles, or workbenches.
▪ Bimanual Complex Assembly: With extreme Blackwell AI processing and EtherCAT synchronization, humanoids can perform complex, bimanual tasks—such as carrying bulky items, picking organic shapes, or installing delicate wiring harness components with human-like dexterity. ▪ Hazardous Environment Safety: Deploy humanoids in environments that are physically demanding for human workers (e.g., chemical plants, high-temperature foundries). You protect your workforce, reduce workplace injury liabilities, and maintain 24/7 operations. |
| Smart Factory |
▪ 100% On-Premise Data Privacy: Runs advanced AI models completely offline. High-value product designs and operational data never leave your facility, eliminating IP leak risks.
▪ Slashed Repair Downtime: Operators can photograph a machine failure to get instant, step-by-step troubleshooting guides from the local AI, bypassing long waits for external specialists. ▪ Immunity to Network Outages: Processes all AI decisions and controls locally. Your manufacturing lines remain fully operational even during total internet blackouts. |
| Smart City |
▪ Savings on Cloud & Data Bills: Processes video feeds locally at the roadside, transmitting only text alerts to city hall instead of massive raw footage—slashing cellular traffic and cloud storage fees.
▪ Disaster-Resilient Operations: Roadside traffic signals and safety monitors run independently. The system maintains public safety even when storms or earthquakes take municipal networks offline. ▪ Durable Design: Certified to operate continuously under extreme seasonal pavement temperatures (-25°C to +55°C) without hardware failures. |
Specifications
| Model Number | AIE-KT78 |
AIE-KT68 |
|
| AI Performance | 2070 FP4/TFLOPS / 1035 FP8 TFLOPS | 1200 FP4/TFLOPS | |
| Module Compatibility | NVIDIA Jetson Thor (Jetson T5000) | NVIDIA Jetson Thor (Jetson T4000) | |
| CPU | 14-Core ARM Neoverse V3AE (64-bit) | 12-Core ARM Neoverse V3AE (64-bit) | |
| Memory | 128GB 256-bit LPDDR5X | 64GB 256-bit LPDDR5X | |
| Storage | 1 x M.2 M-Key 2280 (NVMe 1TB built-in) | ||
| Display | 1 x HDMI 2.0b 1 x DP 1.4a |
||
| LAN | 2 x RJ-45 10GbE 1 x RJ45 1GbE EtherCAT |
||
| USB | 1 x USB-C for OTG only 4 x USB-A 3.2 Gen 2 |
||
| I/O Interfaces |
1× QSFP28 slot (T5000: 4× 25 GbE, T4000: 3× 25 GbE) |
||
| Expansion |
1 x M.2 M-Key 2280 (NVMe1TB built-in) |
||
| Camera | 2 x mini-Fakra (optional) | ||
| Power Input / Connector | 9V-48V | ||
| Dimension (WxDxH) | 220 x 143 x 80 mm | ||
| Net Weight | 3.171 kg | ||
| Temperature |
Operating Temp.: -25°C ~ +55°C (-13°F ~ +131°F) |
||
| Humidity | 95% (non-condensing) | ||
| Software Support | Linux | ||
| Certification | CE, FCC Class A, UKCA | ||
*Product specifications are subject to change without prior notice
Why Aetina
| Value-Added Points | Description |
|---|---|
| NVIDIA Elite Partner | Priority access to NVIDIA technologies and NVIDIA Jetson products, helping clients innovate faster and scale within the NVIDIA ecosystem. |
| End-to-End Customization and Integration | From chassis and thermal design to BSP integration, Aetina partners with customers throughout solution planning and system validation, providing tailored solutions with detailed test reports to meet every unique deployment requirement. |
| Collaborative AI Deployments | With over a decade of edge AI expertise and a robust ecosystem, Aetina provides comprehensive technical and team support to simplify integration and accelerate time to deployment.
|
| Rich Set of Validated Peripherals |
Backed by NVIDIA ecosystem partners and the Innodisk Group, Aetina offers a wide range of fully validated peripherals—including storage, cameras, Wi-Fi, and 4G / 5G—reducing integration time and overall system cost. |
FAQ
| Question | Answer |
|---|---|
| What’s the difference between the AIE-KT78 and AIE-KT68? |
AIE-KT78 is powered by NVIDIA Jetson T5000 and offers higher AI performance, more memory, and greater computing capability for demanding AI workloads. AIE-KT68 is powered by NVIDIA Jetson T4000 and is optimized for applications requiring a balance of performance, power efficiency, and cost. |
| What AI applications can the AIE-KT78/68 support? | The systems are designed for Physical AI and Edge AI applications, including Cobot, humanoid robots, industrial automation, smart city, smart factory, sensor fusion, and generative AI workloads such as Large Language Models (LLMs) and Vision-Language Models (VLMs). |
| Which AI software frameworks are supported? | AIE-KT78/68 supports NVIDIA JetPack™, NVIDIA® CUDA®, NVIDIA® TensorRT™, DeepStream, NVIDIA Holoscan, NVIDIA Isaac™ ROS, and other NVIDIA AI development tools, enabling accelerated development and deployment of AI applications. |
| What types of cameras and sensors are supported? | The systems support a wide range of sensors, including MIPI and GMSL2 cameras, LiDAR, radar, IMU, depth cameras, and other industrial sensors for AI perception and sensor fusion applications. |
| Does the system support remote management and deployment? | Yes. The AIE-KT78/68 supports remote monitoring, application deployment, and system management through Aetina Out-of-Band (OOB). OOB is an optional feature. Please contact Aetina for availability and configuration options. |
| Where can I purchase these products? | To request a quote or place an order, please contact Aetina Sales or click here directly. |
Specifications
Dimensions
Ordering Information
| Model Number | Description |
|---|---|
|
AIE-KT78-1-A1 |
THOR T5000 128G, 1 x DP, 1 x HDMI, 1 x GbE, 2 x 10GbE, 4 x USB, QSFP, B/E/M Key, 9 to 48 V, -25°C ~ +55°C |
|
AIE-KT68-1-A1 |
THOR T4000 64G, 1 x DP, 1 x HDMI, 1 x GbE, 2 x 10GbE, 4 x USB, QSFP, B/E/M Key, 9 to 48 V, -25°C ~ +55°C |
Downloads
| Model Name | File Name | OS | Version | Update Date | Download |
|---|---|---|---|---|---|
| AIE-KT78/68 | Aetina_AIE-KT78/68_Datasheet_Preliminary | v0.5 | 2026-07-30 | Download |