Microchip Technology Inc. Unveils VectorBlox™ 3.0 Accelerator SDK for Enhanced AI Implementation

Microchip's VectorBlox 3.0 SDK is designed to streamline the development process for field-programmable gate array (FPGA)-based AI applications. The toolkit, which is available free of charge to developers, integrates optimization, compilation, and deployment of convolutional neural network (CNN) models tailored for PolarFire® FPGA and SoC-based platforms. The ability to efficiently scale across various model sizes and support multiple AI workloads on a single low-power FPGA significantly enhances the potential for consolidating different vision or sensor-based AI functionalities.

“As AI models continue to grow in complexity, compression is becoming essential for deploying intelligence at the edge,” stated Shakeel Peera, Corporate Vice President and General Manager of Microchip’s FPGA business unit. The VectorBlox 3.0 SDK utilizes sparsity-based model compression technologies acquired from Microchip’s Neuronix division, thereby reducing computational demands while maintaining output accuracy.

The primary advantage of the VectorBlox 3.0 SDK lies in its support for sparse neural networks. By enabling the execution of vision-based CNN models to skip zero-valued operations, developers can accelerate inference performance without compromising power consumption. This feature is crucial for always-on edge AI applications, which require a delicate balance between responsiveness and energy efficiency.

Vito Fortunato, SPACEDGE™ Services Line Manager at Planetek Italia, emphasized the practical applications of this technology, stating, “Leveraging VectorBlox acceleration on Microchip’s PolarFire SoC enabled us to efficiently deploy advanced onboard AI pipelines for low-latency payload operations in orbit.” He highlighted the successful validation of real-time Earth observation capabilities, such as object detection and semantic scene analysis, using the AI-eXpress-1 satellite, which was deployed in 2025.

In addition to CNN implementations, the announcement of the Spacecraft Pose Network v2 (SPNv2) offers significant advancements in autonomous navigation. This neural network utilizes vision data to estimate position and orientation, facilitating applications like autonomous rendezvous and docking, space debris removal, and satellite inspection. Built on the mid-range, power-efficient, single-event-upset (SEU) immune PolarFire FPGAs and SoCs, SPNv2 provides essential features including secure boot, anti-tamper protection, and high reliability, making it suitable for the rigorous demands of aerospace and defense applications.

Federico Fontana, Head of Hardware Engineering at AIKO, remarked on the synergy created by combining the PolarFire SoC with VectorBlox, stating, “We validated this through the deployment of our clear_CHARLES suite, which provides onboard cloud and ship detection for adaptive and autonomous payload operations.”

The VectorBlox SDK v3.0 and the associated CoreVectorBlox IP are accessible to customers at no charge, reflecting Microchip's commitment to supporting developers in the AI space. For further information, interested parties can engage with a Microchip sales representative or an authorized distributor.

For those looking to delve deeper into the capabilities of the VectorBlox SDK, Microchip is hosting a free webinar on July 16, 2026, titled "Achieve Two Times Faster CNN Inference with Sparsity-Aware AI Acceleration on PolarFire® SoC FPGAs." This session will be available on demand following the live presentation.

Microchip Technology Inc.'s release of the VectorBlox™ 3.0 Accelerator SDK marks a significant advancement in the realm of AI implementation, particularly for industries that require high efficiency and reliability in challenging environments. With its focus on simplifying development and enhancing performance, Microchip continues to solidify its position as a leader in semiconductor innovation, poised to meet the ever-evolving demands of technology in aerospace, defense, and beyond.