Aetina shows modular edge servers

Aetina said on May 22 that it would bring a full stack of edge AI hardware to COMPUTEX 2026, from embedded modules to enterprise servers built on Nvidia’s MGX reference architecture. The Taiwan-based company said the lineup spans four product families — SuperEdge, MegaEdge, DeviceEdge and CoreEdge — and will be shown with live demonstrations at Taipei Nangang Exhibition Center from June 2 to June 5. Nvidia says MGX is designed to let manufacturers build accelerated systems faster, with a modular architecture that can scale from single-node servers to rack-scale deployments. ### What exactly did Aetina put on show? Aetina said the COMPUTEX display centers on hardware for “Physical AI, Vision AI and Agentic AI,” with products ranging from Jetson-based edge devices to larger GPU servers. The company’s announcement names SuperEdge servers based on Nvidia MGX, MegaEdge inference platforms with GPU expandability, DeviceEdge systems powered by Nvidia Jetson Orin modules, and CoreEdge AI modules for vision and embedded workloads. (aetina.com) The company also described a demo that links a Jetson Thor-based DeviceEdge system with a SuperEdge AEX-2UA1 GPU server using an Nvidia RTX PRO 6000 Blackwell Server Edition GPU. Aetina said that setup runs a private large language model, Nvidia ACE, stereo-camera perception and robotic control on site rather than through a cloud service. ### Why does the Nvidia MGX piece matter? (aetina.com) Nvidia says MGX is a modular reference architecture for OEMs and ODMs building accelerated computing systems. The company says the design is meant to reduce engineering cost, shorten time to market and preserve compatibility across multiple generations of hardware. That matters because Aetina is not presenting one fixed server. (aetina.com) A modular MGX design lets a vendor assemble different system configurations around common building blocks, which is the practical appeal for operators that want inference capacity outside the largest cloud campuses. That reading is an inference from Nvidia’s description of MGX and Aetina’s use of it in its SuperEdge line. (nvidia.com) ### Where would this kind of hardware actually be deployed? Aetina’s own materials point to industrial automation, robotics, in-vehicle systems and enterprise automation. Those are all settings where workloads may need to run near cameras, robots, local users or private data rather than in a distant cloud region. (aetina.com) The hardware mix also fits smaller-footprint deployments. Embedded modules and Jetson systems can sit inside devices or compact edge boxes, while GPU-expandable inference platforms and MGX-based servers can be installed in on-premises racks or metro data centers. That deployment pattern is an inference from the product categories Aetina listed and Nvidia’s description of MGX system range. (roboticstomorrow.com) ### How does this connect to latency-sensitive media and newsroom workloads? Aetina did not frame the launch around newsrooms, but the product characteristics are relevant to media operators that need low-latency inference near users. Systems that handle preview generation, live assistance, search, transcription or clip selection often benefit from local processing when operators want lower delay, data control or resilience during network congestion. (aetina.com) That is an inference based on the edge deployment model Aetina described, not a claim the company made specifically about publishing. Nvidia’s description of MGX as a path from single-node servers to larger AI infrastructure also helps explain why vendors are pitching modular edge capacity rather than only hyperscale cloud usage. The same architecture can support a smaller starting footprint and a larger follow-on build, depending on workload and site constraints. ### What happens next in Taipei? (aetina.com) COMPUTEX 2026 is scheduled for June 2 through June 5 at Taipei Nangang Exhibition Center. Aetina said visitors will find its exhibit in Hall 1, booth K0106, where the company plans to show the SuperEdge, MegaEdge, DeviceEdge and CoreEdge portfolio with partner demonstrations. (roboticstomorrow.com) (nvidia.com)