Microchip Technology has introduced the VectorBlox™ 3.0 Accelerator Software Development Kit (SDK), a new solution designed to simplify the deployment of artificial intelligence applications on PolarFire® FPGA and System-on-Chip (SoC) platforms. The latest release is aimed at accelerating AI inference for power-constrained and mission-critical applications, including aerospace, defense and industrial systems, while improving performance, energy efficiency and development speed.
Available free of charge, the VectorBlox 3.0 SDK integrates optimization, compilation and deployment tools into a single development environment for convolutional neural network (CNN) models. Combined with the CoreVectorBlox intellectual property (IP), the platform enables developers to efficiently implement AI workloads on PolarFire FPGA and SoC devices. The scalable architecture also allows multiple vision and sensor-based AI applications to run on a single low-power device, reducing hardware complexity and overall system costs.
A key enhancement in the latest release is support for sparse neural networks, enabling the software to bypass zero-value operations during AI inference. This approach reduces computational and memory requirements while maintaining model accuracy, resulting in faster processing and lower power consumption. The technology is particularly suited to always-on edge AI applications that require real-time responsiveness while operating within strict power budgets.
“As AI models continue to grow in complexity, compression is becoming essential for deploying intelligence at the edge,” said Shakeel Peera, Corporate Vice President and General Manager of Microchip’s FPGA Business Unit. “VectorBlox 3.0 leverages sparsity-based model compression technology to reduce compute demands while preserving accuracy, helping developers deploy advanced AI solutions more efficiently.”
The platform has already demonstrated its capabilities in space-based applications. Planetek Italia successfully used VectorBlox acceleration on Microchip’s PolarFire SoC to implement onboard AI pipelines for low-latency Earth observation processing aboard the AI-eXpress-1 satellite. The solution enabled real-time object detection, scene analysis and actionable data generation while delivering the radiation tolerance and reliability required for continuous Low Earth Orbit operations.
Microchip also highlighted support for Spacecraft Pose Network v2 (SPNv2), an AI model designed to estimate spacecraft position and orientation using vision data. The technology enables autonomous navigation, rendezvous and docking, satellite inspection, debris removal and formation flying.
Running on PolarFire FPGAs and SoCs, the platform combines power efficiency with secure boot, anti-tamper protection and high reliability for demanding aerospace, defense and industrial environments. VectorBlox™ SDK v3.0 is fully supported by Microchip’s Libero® SoC Design Suite and integrates seamlessly with CoreVectorBlox IP, providing developers with a comprehensive framework for building advanced edge AI applications while reducing development time and accelerating deployment.
Microchip Technology is widely recognized for its semiconductor, embedded systems, and timing solutions used across automotive, industrial, aerospace, communications, and consumer electronics applications. Through ongoing innovation in frequency control and precision timing technologies, the company aims to support next-generation electronic systems that demand higher performance, lower power consumption, and improved operational resilience.
