MemryX is an AI technology company founded in Ann Arbor in 2019, focused on efficient, scalable acceleration for AI inference. Its products serve deployments ranging from real-time edge applications to high-performance and data-center computing. The portfolio includes accelerator silicon such as the production-proven MX3, the next-generation MX3+, and planned MX4, alongside modules in M.2, PCIe, Raspberry Pi 5 HAT+ and USB formats. MemryX also offers production-ready edge and server systems that combine its accelerators with computing and connectivity components.

The company’s dataflow architecture is designed to keep computation moving efficiently, reduce data movement and support parallel inference. Its toolchain compiles and runs trained models while allowing developers to retain familiar models, frameworks and application workflows. Developers can access an SDK, compiler, model zoo, documentation, reference applications, open-source tools and a community forum.

MemryX supports industrial and robotics, video management, transportation intelligence and smart-device markets. Its offerings provide options for product teams at different stages, from chip integration to complete systems, with a shared architecture intended to scale across deployment needs.