Building physical AI deployment infrastructure with AWS

Ultra-low-power AI semiconductor adoption expected to expand

The AWS Startup Partner Summit 2026, held in San Francisco. [Provided by DeepX]
The AWS Startup Partner Summit 2026, held in San Francisco. [Provided by DeepX]

DeepX CEO Kim Nok-won attended the Amazon Web Services Startup Partner Summit 2026 in San Francisco from Monday (local time) through Tuesday, seeking to expand the company's reach among global partners and customers.

DeepX's participation in the AWS startup partner ecosystem — which is dominated by cloud and software companies — is unusual for a physical AI chipmaker. The company is exploring new collaboration opportunities that would connect cloud infrastructure with edge hardware such as cameras, robots and industrial PCs.

DeepX has been building a physical AI deployment environment that links its low-power AI semiconductor, or neural processing unit (NPU), with AWS's cloud and Internet of Things (IoT) infrastructure as part of its push into global markets.

AWS highlighted the collaboration on Thursday through its official technology blog, introducing a method for deploying and managing NPU runtime environments across multiple edge devices by combining DeepX's DX-M1 NPU with AWS IoT Core and AWS IoT Greengrass.

Using the DeepX Greengrass Solution available on the AWS Marketplace, operators can apply a uniform AI runtime environment and continuously update software and models across multiple devices installed in factories, logistics centers, stores and robots — without visiting each location in person.

DeepX is considering expanding the cloud-to-edge deployment architecture built on the DX-M1 to customers in US robotics, industrial vision and smart factory sectors, and broadening its application to next-generation generative AI semiconductors such as the DX-M2.

"For physical AI to spread in earnest, what is needed is not only low-power AI semiconductors but also the infrastructure to easily deploy AI developed in the cloud to countless on-site devices and manage them continuously," a DeepX official said. "We will work to broaden the foundation for physical AI to reach actual industrial sites by connecting AWS's cloud and IoT ecosystem with DeepX's ultra-low-power AI semiconductor technology."


joze@heraldcorp.com