MemryX Inc., the maker of artificial intelligence solutions for the edge, announced Thursday that it raised $44 in a Series B funding round.

Founded in 2019, MemryX designs AI accelerator chips that are used to power smart devices such as autonomous vehicles, robotics, industrial automation, internet of things devices, drones, healthcare equipment and more.

The company’s key products include an integrated chip called the MX3, which the company says outperforms most mainstream graphics processing unit solutions for vision-based AI inference using less than 10% of the power.

The chip has applications for AI-powered applications such as video management systems, industrial PCs, on-device large language models and robot brains. The company provides various software tools for developers, including benchmark measures for hundreds of AI models publicly available on the MemryX Developer Hub.

“Our software-first core architecture has enabled us to be extremely capital efficient while developing a robust software stack and support dozens of customers,” said Chief Executive Keith Kressin. “It also provides a very strong foundation to build our next-generation MX4 accelerator.”

Kressin added the funding would be used to ramp-up production on the M3 chip and complete the design of the iteration of the company’s chip, the MX4.

The MX3 can provide up to 6 TFLOPs, trillions of floating-point operations per second, depending on power availability. As for power, the chips are designed for very low power use because of the applications that they might be used in with an average power of 0.6 to two watts per chip.

The company also makes a M.2 module that houses four MX3 chips capable of delivering up to 24 TFLOPS. In this form factor, the accelerators can be plugged into a PC, laptop or other industrial device with a suitable M.2 slot with no hardware changes.

The module supports all common AI frameworks and up to 80 million weight parameters using on-chip storage that eliminates the need for extra memory. The company said it also supports concurrent models and end-user models can be deployed without the need for pruning or compression.

The increasing deployment of smaller AI models directly onto devices is driving a greater need for real-time data processing at the network edge. Applications such as self-driving cars and robotics heavily rely on vision models that must react swiftly to changing environments by processing data near its source. In healthcare, edge AI facilitates remote patient monitoring through on-device data analysis, enabling timely medical interventions.

Market analysis firm Grand View Research Inc. estimated the edge AI market size at $20.8 billion in 2024 with a forecasted $66.5 billion by 2030. The increasing demand for real-time data processing and analysis at the edge has been driving growth across the industry, especially in connection with 5G networks. North America dominated the edge AI industry with a revenue share of 37% in 2024 and is projected to have the highest growth given the penetration of 5G.

The funding round brings the total raised by MemryX to around $63 million. Although the company did not reveal the backers of its current round, its Series A in 2021 attracted investors including Innovation Partnerships, Arm IoT Capital, HarbourVest Partners, eLab Ventures and Hemi Ventures.

Image: MemryX