Aetina is showing its latest integrated AI systems at GTC in San Jose, Calif. Source: Aetina Corp.
Aetina Corp., a leading global provider of edge AI systems, is demonstrating high-precision 3D vision technology and enterprise-grade generative AI agents this week at GTC 2026. The company said its systems use NVIDIA’s Blackwell architecture to enable robots and automated systems “to advance toward end-to-end autonomy.”
In physical AI application scenarios, robots and automation require real-time, reliable spatial perception. They also need a “decision-making brain” capable of executing complex task reasoning at the edge for dynamic orchestration amid shifting tasks, exceptions, and changing operational priorities, said Aetina.
The company claimed that its presence at GTC highlights its role as a key infrastructure enabler for both robotic and enterprise autonomy. Aetina said it provides the “eyes of perception” and the “brain of intelligence” required for autonomous systems, accelerating deployment of AI at the edge across industries.
The New Taipei City, Taiwan-based company offers a range of AI computing systems, platforms, hardware, and software tools for the creation of different types of vertical AI. Aetina said its integrated systems empower its Arm, x86 computers, GPUs, ASIC hardware, cloud management software, and development tools with AI.
With its ecosystem network, Aetina added that its global partners’ technologies make the creation and deployment of specific AI-powered applications easier.
Stereolabs GMSL2 3D Vision partners to advance perception
Aetina is showing the DeviceEdge AIB-AT78/68 series, which it claimed is one of the industry’s first production-ready platforms powered by NVIDIA Jetson T5000 and T4000 modules. The DeviceEdge series provides edge computing and real-time perception, as well as decision abilities to support increasingly complex robot behaviors and AI skills, according to the company.
It added that the AIB-AT78/68 has already been integrated into onboard computing platforms for humanoid robots across manufacturing, logistics, and service sectors.
The AIB-AT78/68 series delivers up to 2070 FP4 TFLOPS of AI performance with extensive I/O, including QSFP28 (100 Gbps), EtherCAT, and dual 10 GbE industrial-grade connectivity interfaces. This precisely addresses in-field demands for sensor fusion and high-speed motion control, said Aetina.
By combining NVIDIA Isaac, NVIDIA Isaac GR00T, and Vision AI workflows, the company said its platform can help close the gap between simulation and real-world deployment, driving scalable physical AI adoption.
In addition, through a collaboration with Stereolabs, Aetina plans to demonstrate the Humanoid Surround Vision Solution onsite. Built on the AIB-AT78/68 platform, the GMSL2 (Gigabit Multimedia Serial Link 2) system integrates two ZED X stereo cameras for frontal depth sensing and two ZED X One cameras for peripheral awareness.
Using GMSL2’s high bandwidth and low latency, the DeviceEdge system enables sub-millisecond 3D perception processing, which Aetina said gives robots human-like navigation capabilities. The platform supports an operating temperature range of -25° to 80°C (-13° to 176°F) and a 130W power-efficient design, engineered specifically for industrial environments.
Aetina builds a decision ‘brain’ with NVIDIA MGX edge servers
At GTC, Aetina also plans to present its innovations powered by NVIDIA Blackwell architecture, featuring the DeviceEdge AIB-AT78/68 series edge supercomputing platform powered by NVIDIA Jetson Thor and modular NVIDIA MGX edge servers with NVIDIA NIM microservices support.
“Aetina is deeply integrating GMSL2 3D visual perception technology with the NVIDIA Jetson Thor platform, while leveraging the outstanding inference capabilities of NVIDIA MGX modular edge servers to build diversified, deployable infrastructure for the physical AI era,” said Joe Lo, president of Aetina. “In addition to empowering next-generation robots with comprehensive environmental perception, we are helping construct an autonomous ‘brain’ that connects sharp sensing to intelligent decision-making—accelerating the journey from lab prototypes to real-world, multi-scenario deployments.”
Beyond robotics, Aetina will also showcase the SuperEdge AEX-2UA1 2U short-depth edge server based on the MGX architecture, accelerated by NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs.
By combining NIM microservices to run large language models (LLMs) at the edge, the company said it helps enterprises rapidly adopt agentic AI. It can perform on-premises reasoning, analyze operational data, and generate real-time insights and recommendations, while preserving data sovereignty and security.
This architecture is suitable for latency-sensitive scenarios such as warehouses and factories, where data cannot be easily moved to the cloud or where local real-time decision-making is required, noted Aetina. “It enables seamless edge-side integration of high-fidelity perception and workflow reasoning, propelling AI applications to a new level of autonomy,” the company said.
Aetina invited global partners and media to visit Booth 139 to see its live demonstrations. It said it expects the AIB-AT78/68 series to be commercially available in late March.
The post Aetina shows 3D vision and enterprise generative AI at GTC 2026 appeared first on The Robot Report.
