Axelera, which is developing an AI accelerator chip, raises $12M

Axelera AI, a Dutch AI semiconductor startup, today emerged from stealth with $12 million in seed funding led by Bitfury, a security and infrastructure provider for the Bitcoin blockchain. Innovation Industries Fund participated in the round, which the company says will support developing of AI apps at the edge.

Edge AI, the processing of AI algorithms on devices like drones, digital signage, cameras, and sensors, derives from edge computing, which starts from the same premise: Data is processed and managed at internet of things endpoints. Edge AI is advantageous in industries like retail, robotics, and manufacturing, where it can deliver lower latency than sending data to remote servers — and avoid disruptions and costs in the process.

Incubated by Bitfury, Axelera says it’s developing a chipset to accelerate AI and machine learning algorithms at the edge. The startup claims that its product, which is scheduled to ship to select customers in 2022, will achieve a fraction of the power consumption and price of competing hardware.

“The market for AI at the edge is in its early days, with the current market leader being Nvidia. Recently, new startups such as Hailo, Mythic, and Blaize are coming to market with new products. As many products were released last year, there are no consolidated winners yet,” CEO and cofounder Fabrizio Del Maffeo, former head of AI at the Bitfury Group, told VentureBeat via email. “Our mission is to provide a ‘green’ (low power consumption) hardware and software platform that enables the industry to take full advantage of what AI can bring.”

Innovative design

The edge AI hardware market is projected to grow from 920 million units in 2021 to 2.08 billion units by 2026 — a lucrative uptick. According to one estimate, the AI chips market alone is poised to be worth $73.49 billion by 2025.

Del Maffeo asserts that Axelera’s technology is differentiated by a custom dataflow architecture with multicore in-memory computing. In-memory computing, or in-memory computation, is the technique of running computer calculations entirely in computer memory, which can deliver high performance at low wattage.

Usability will be a focus with the chip. Axelera customers will be able to tap off-the-shelf tools such as Facebook’s PyTorch or Google’s TensorFlow machine learning frameworks and won’t need expertise in new AI development environments. And Axelera will provide software utilities that leverage performance-boosting approaches like graph optimizations and quantization, which compresses AI models with only small losses in accuracy.

“Recent advances in deep learning have transformed the way computing devices process human-centric content such as images, video, speech, and audio,” Del Maffeo said. “Applying Axelera’s technology to IoT devices could enable a new generation of applications capable of performing complex sensing and recognition tasks.”

While Axelera is pre-revenue, the 30-person company has partnered with IMEC, a nanoelectronics R&D organization, to develop its chip architecture. (IMEC is also an investor.) Del Maffeo says that Axelera is in discussion with about 20 customers ahead of its planned ship date.