Acer used IFA 2026 to prove that serious AI horsepower doesn’t need a tower case anymore. The company introduced two mini desktops built for running AI agents locally instead of leaning on the cloud: the shipping Veriton RI110 AI Mini Workstation and a striking concept called the Acer SFF RTX Spark design. Both squeeze desktop-class compute into shells small enough to disappear on a shelf.
The Veriton RI110 fits an Intel Core Ultra X7 processor 358H and Intel Arc B390 graphics into a chassis measuring just 5.45 by 5.17 by 2.05 inches and weighing about 1.39 lbs. Despite that footprint, it supports AI models with up to 120 billion parameters, backed by up to 96 GB of memory and 4 TB of SSD storage, plenty for demanding local workloads.
Designer: Acer
That power feeds directly into Acer Qubi Claw, a new agentic AI assistant designed to run local workflows inside an isolated sandbox so sensitive data never has to leave the machine. Installation takes one click, and its starting skill set already covers in-depth online research, travel itinerary planning, and transcribing or translating audio into text notes, all schedulable on autonomous routines that work quietly in the background.
Acer also builds in OCuLink connectivity, a dedicated PCIe 4.0 x4 interface delivering up to 64 Gbps of bandwidth for connecting external GPUs or high-speed NVMe storage. That link skips the extra processing overhead of traditional expansion methods, so creators and developers can bolt on desktop-class rendering power only when a project demands it, then disconnect it again without carrying that bulk around every day.
Standing beside it, the Acer SFF RTX Spark design takes a different shape entirely, built around NVIDIA’s RTX Spark superchip rather than Intel silicon. Vented panels wrap a narrow vertical shell accented with a warm metallic stripe, giving the design a presence that reads more like a piece of audio equipment than a workstation, while a fold-out stand lets it stand upright on a desk instead of lying flat.
Underneath that shell, Acer packs up to a 6,144-core Blackwell RTX GPU, up to a 20-core NVIDIA Grace CPU, up to 128 GB of unified memory, and up to 1 petaflop of AI compute, numbers usually reserved for server racks. Paired with NVIDIA’s full-stack AI platform and its complete suite of RTX technologies, the design targets creators, developers, and gamers alike.
Acer hasn’t confirmed pricing or a release window for the SFF RTX Spark design, while the Veriton RI110 arrives in North America during the fourth quarter of 2026 and reaches EMEA in the first quarter of 2027. Together, the two machines make a case that agentic AI doesn’t need a data center to feel powerful, just a well-engineered box small enough to tuck beside a monitor and forget about.
