AMD Unveils Versal AI Edge Gen 2, Aiming at AI-Powered Edge Solutions
Timothy Morano Sep 22, 2026 16:30
AMD launches Versal AI Edge Series Gen 2, combining AI inference, sensor processing, and embedded compute in a single chip for edge applications.
AMD has officially launched its Versal™ AI Edge Series Gen 2, a production-ready adaptive system-on-chip (SoC) aimed at simplifying AI-driven edge computing. This next-generation chip platform integrates real-time sensor processing, AI inference, and high-performance embedded compute into a single device, cutting through the complexity of multi-chip solutions. Paired with AMD's Vitis™ AI toolchain, the Versal AI Edge Gen 2 aims to accelerate deployment timelines for industries like automotive, robotics, and healthcare.
Edge AI deployments—such as advanced driver assistance systems (ADAS) and autonomous robotics—face stringent constraints on power, latency, and reliability. AMD's new platform targets these challenges by unifying the entire AI pipeline, from preprocessing to inference and postprocessing, on a single chip. This approach reduces system footprints, improves thermal efficiency, and enables faster decision-making in real-world scenarios.
Key Features of Versal AI Edge Gen 2
The Versal AI Edge Series Gen 2 incorporates heterogeneous compute capabilities tailored for end-to-end AI acceleration:
- Programmable logic: Handles sensor connectivity and real-time preprocessing.
- AI Engines: Accelerates power-efficient AI inference without consuming additional programmable resources.
- Scalar compute: Supports high-performance postprocessing and deterministic decision-making tasks.
- Single-chip integration: Reduces communication overhead and system complexity.
For developers, the Vitis AI software stack provides a unified environment to optimize AI inference and allocate resources efficiently. Vitis AI 6.2, released earlier this year, introduced out-of-the-box support for modern neural networks like convolutional neural networks (CNNs) and vision transformers, with support for multiple data precision formats, including INT8, FP16, and BF16.
Performance in Real-World Scenarios
AMD demonstrated the Versal AI Edge Gen 2's capabilities in a multi-camera setup processing four simultaneous 1080p video streams. Using the YOLOv8m object detection model, the system achieved an aggregate throughput of 120 frames per second (fps) while consuming just 64 out of 144 AI Engine tiles available on the chip. This highlights the platform’s ability to meet demanding workloads while conserving resources for additional tasks or power savings.
Market Context
AMD first announced the Versal AI Edge Gen 2 in April 2024, positioning it as a key product in its adaptive SoC portfolio. The platform builds on the first-generation architecture with up to 3x TOPS per watt efficiency and expanded support for advanced data types like FP8 and FP16. Early adoption by automotive players, such as Subaru, underscores its potential in safety-critical applications like ADAS.
The launch comes amid growing demand for edge AI solutions across industries, with edge computing projected to be a multi-billion-dollar market by 2030. AMD’s ability to integrate compute, inference, and safety features into a single chip could strengthen its competitive edge against rivals like NVIDIA and Intel in the embedded AI space.
What’s Next?
AMD plans to release Vitis AI 6.3 shortly, bringing further improvements in ease of use and deployment optimization. Enhancements include automated workflows, improved support for vision transformers, and expanded device compatibility across the Versal AI Edge Gen 2 portfolio.
For developers, AMD offers tools to jumpstart their design process, including evaluation boards and tutorials for building custom AI pipelines. With the Versal AI Edge Gen 2 now in production, AMD is positioned to play a pivotal role in enabling edge AI at scale.
Image source: Shutterstock.jpg)