Industrial robotics platform Axis.ONE supports inference on Intel CPUs and neural processing units without dedicated GPUs, with fully on-premise operation.

A press release filed by HeartCore Enterprises (HTCR) describes ONE AI’s Axis.ONE platform as supporting inference on Intel central processing units (CPUs), neural processing units (NPUs), and other edge hardware without dedicated GPUs. It is designed for 100% on-premise operation, including air-gapped and closed-network environments without cloud dependency.

Axis.ONE is designed to operate alongside existing robots, programmable logic controllers (PLCs), cameras, and industrial computers without modifying existing robot control systems. Its vendor-agnostic architecture supports Robot Operating System 2 (ROS 2), and the release cites the International Federation of Robotics’ figure of more than 4.6 million industrial robots operating worldwide as installed-base context.

The platform’s Axis.LOOP component uses field data for on-site model retraining, while Axis.MIMIC uses vision-language-action models designed to teach robots through demonstrations. Axis.CORE provides centralized AI operations management, model management, over-the-air updates, on-premise large language model services, and real-time dashboards across sites.

ONE AI estimates total cost reductions of up to 90% compared with conventional GPU-based deployments and reports per-site learning cycles falling from approximately 100 days to 10 days, or in some cases from 30 days to one hour. These are company claims in a filed press release, not independently verified benchmarks; the supplied excerpt provides no test methodology, workload definitions, or deployment sample size.