Industrial-Grade AI in an M.2 Slot, No Cloud Required: USA-Made GW16168 AI Accelerator with NXP Ara240 Is Ready for Deployment

Industrial-Grade AI in an M.2 Slot, No Cloud Required: USA-Made GW16168 AI Accelerator with NXP Ara240 Is Ready for Deployment

PR Newswire

AKRON, Ohio, Nov. 19, 2026 /PRNewswire/ — As more embedded designs move AI workloads from the cloud to the device, Ezurio is launching the GW16168, its USA-made M.2 AI accelerator powered by the NXP Ara240 Discrete Neural Processing Unit (DNPU). The GW16168 delivers up to 40 equivalent tera operations per second (eTOPS) of on-device AI in a standard M.2 M-Key 2280 form factor and is available today.

Gateworks, an Ezurio company.

Designed and manufactured in the USA, the GW16168 gives industrial, defense, and other supply-chain-sensitive customers added transparency, quality control, and long-term availability. It is purpose-built for the VeniceFLEX and Catalina single board computers and works with other compatible M.2 M-Key 2280 platforms.

The GW16168 runs AI models entirely on the device, with no connection to the cloud. That matters for products that operate beyond reliable network coverage, such as drones, robots, and remote industrial equipment, and for applications where latency, privacy, or security rule out sending data off-site. It enables real-time machine vision, large language model (LLM) and vision language model (VLM) inference, and intelligent automation.

The GW11062 evaluation kit is the fastest route to testing it. Each kit pairs a Catalina SBC with a GW16168 accelerator (GW11062-1 with 16GB, GW11062-2 with 8GB) and comes preloaded with an Ubuntu image which reduces setup time and the learning curve for developers evaluating the platform.

“Edge AI is moving from pilot to production, and customers need hardware they can trust and source,” says Eric Singer, Ezurio Product Manager. “The GW16168 pairs NXP’s Ara240 DNPU with industrial-grade, USA-made design, so customers can scale AI performance without redesigning their entire platform. That brings flexibility, longevity, and cost efficiency to real-world deployments.”

GW16168 Key Features:

  • NXP Ara240 Discrete Neural Processing Unit (DNPU)
  • Up to 40 eTOPS of AI acceleration
  • 8GB or 16GB LPDDR4 DRAM
  • M.2 M-Key 2280 form factor (22.0 x 80 x 3.5 mm)
  • PCIe Gen4 x4 interface, single 3.3V supply
  • Industrial temperature range: -40°C to +85°C
  • Secure Boot and Root of Trust
  • TensorFlow, PyTorch, and ONNX model support via NXP Ara SDK
  • Made in the USA, with a long lifecycle commitment

The GW16168 embedded module and GW11062 development kits are available now. For product details, visit www.ezurio.com/gw16168.

Originally part of the Gateworks portfolio, the GW16168 is now fully supported by Ezurio. Ezurio acquired Gateworks in June 2026

About Ezurio

Ezurio engineers embedded wireless connectivity and compute solutions that power connected devices worldwide, backed by deep expertise in software, security, regulatory compliance, and long-term lifecycle support. Our portfolio includes RF modules, system-on-modules, single board computers, internal antennas, IoT devices, and custom solutions designed to accelerate development and reduce time to market.

For the latest news or more information, visit:

ezurio.com | linkedin.com/company/ezurio-llc | /x.com/ezurio_llc

For more information, please contact:
Jordan Manser, Sr. Marketing Manager
Email: jordan.manser@ezurio.com
Phone: +1-800-492-2320

Cision View original content to download multimedia:https://www.prnewswire.com/news-releases/industrial-grade-ai-in-an-m2-slot-no-cloud-required-usa-made-gw16168-ai-accelerator-with-nxp-ara240-is-ready-for-deployment-302945619.html

SOURCE Ezurio