SoftHetro™ by Axcelia Labs is a software-defined AI acceleration technology developed with Intel Corporation and designed to dynamically orchestrate AI workloads across heterogeneous computing resources. Rather than treating the CPU, GPU, and NPU as independent inference devices, SoftHetro coordinates workload execution across the platform based on application requirements and available compute resources
Dynamically coordinates AI workloads across CPU, GPU, and NPU, turning multiple compute engines into one intelligently managed acceleration platform
Distributes and schedules AI processing at a fine-grained level to improve latency, workload balance, and compute-resource utilization
Goes beyond inference by considering memory, cache, buffers, data movement, power, and compute availability to optimize AI execution across the entire platform
From AI inferencing and Edge AI to intelligent applications, real-time workloads, and next-generation gaming experiences, SoftHetro is designed to push heterogeneous platforms beyond conventional execution
Copyright © 2026 Axcelia - All Rights Reserved.
Axcelia Inc. reserves the right to make changes to this website, its services, and its Terms of Service, at any time and without notification.
All functionality, features, specifications, GUI and other information provided on this website including, but not limited to, the benefits, design, pricing, components, performance, availability and capabilities of the service are Axcelia Inc. intellectual property and are subject to change without notice or obligation.
HQ Address: 1625 The Alameda #700, San Jose, CA 95126 USA
We use cookies to analyze website traffic and optimize your website experience. By accepting our use of cookies, your data will be aggregated with all other user data.