OUR MISSION
Accelerate edge AI with a solution that is highly efficient, scalable, and simple to deploy and support.
The future of computing requires design focus be placed on the efficient movement of data. By jointly architecting a hardware and software solution from the ground up for dataflow-centric AI, we can quickly ramp Edge and make intelligent devices more pervasive in our everyday lives.
Dr. Wei Lu
Co-founder, MemryX
WHAT DEFINES US
The principles behind how we build.
//01
Rethink the fundamentals
We started by questioning the assumptions behind conventional AI compute. That same mindset continues to shape how we approach architecture, products, and the problems that come next.
//02
Efficiency by design
Efficiency isn’t an optimization we add later. It begins with the architecture—reducing unnecessary data movement and putting resources toward the work that matters.
//03
Complex technology, made usable
Advanced hardware only matters if people can build with it. We pair purpose-built silicon with software, tools, documentation, and support designed to make AI acceleration straightforward to adopt.
//04
Build for what comes next
AI is evolving quickly. We build technology with room to grow—across models, workloads, products, and generations—so customers can invest in MemryX with the future in mind.
HOW WE GOT HERE
From a Michigan lab to volume silicon
MemryX was co-founded in 2019 by Dr. Wei Lu, an IEEE Fellow and Professor of Electrical Engineering at the University of Michigan since 2005, and a highly influential expert in memory devices, neuromorphic computing and in-memory computing systems.
Dr. Lu and a small team at Michigan understood that the future of computing depended on the ability to scale and ramp dedicated AI accelerators across a large and diverse embedded market. The team set out to build a fundamental core architecture for edge AI with high flexibility and reconfigurability.
The result was a forward-looking design with dual priorities: minimise data movement, and ship software that is practical to deploy and scalable enough to support hardware generations that had not been designed yet.
That core architecture was proven and optimised across two internal generations before a single product shipped — and it is the same fabric running inside every MX3 accelerator today.
2019
MemryX founded
Co-founded in Ann Arbor by Dr. Wei Lu around a new at-memory dataflow architecture for AI inference.
2021
MX1
First-generation silicon used to validate the core MemryX architecture and software foundation.
2023
MX2
Second-generation hardware expanded the architecture and helped mature the compiler, runtime, and system-level design.
2025
MX3
Third-generation production silicon brought MemryX to market with efficient, scalable AI inference across the Cascade product family.
2026
Recognized in the field
MemryX gains broader industry recognition for efficient AI inference, including the 2025 Edge AI and Vision Product of the Year Award and growing deployment across multiple Cascade form factors.
Leadership
Proven semiconductor leadership
Built by people from Nvidia, Onsemi, Qualcomm, IBM and the University of Michigan.
Governance
Board of Directors
Investors and founders steering the company through production and scale.
Ross Jatou
President and Chief Executive Officer.
Dr. Wei Lu
Co-Founder and CTO. Professor at the University of Michigan and an IEEE Fellow.
Sebastian Schoefer
Representing Series A investors.
Charlie Talcott
Representing Series B investors.
Rick Bolander
Representing Series Seed investors. Co-founder and Managing Director at eLab Ventures; a venture investor since 1994 with more than 50 deals and over $100M led in early-stage financings.
Technical advisory committee
The people we pressure-test with
Architecture, silicon and machine-learning specialists who review where the technology is going.
Pat Moorhead
Founder and Chief Analyst; previously Corporate Vice President at AMD. Covers silicon, AI infrastructure and the semiconductor market.
Jem Davies
Arm Fellow and former General Manager of the Machine Learning group, previously leading GPU and media processing.
Onur Mutlu
Professor of Computer Science, previously Carnegie Mellon. Works on memory systems, processing-in-memory and computer architecture.
Guy Lemieux
Professor of Electrical and Computer Engineering and founder of VectorBlox Computing, working on reconfigurable and vector processing.
Proof
Independently Tested. Real-World Validated.
Independent labs, reviewers, and publications have tested MemryX for themselves. See what they found.
“… The 1st AI accelerator we’ve encountered for which both the hardware and the software just works … exceptionally easy to use while providing good performance and consuming little power.”
BDTi
Independent Benchmarking Evaluation
“It’s a low-power, high-efficiency workhorse … it integrates seamlessly into existing systems — bringing massive AI capability without the usual hardware overhaul.”
ip X change
Hardware Review
“When it comes to edge AI, the MX3 M.2 AI Accelerator Module punches way above its weight — literally.”
Elector {maker}
Hands-On Module Test
"The MemryX-powered system strengthens our ability to apply real-time AI on the factory floor while establishing a scalable architecture for our broader smart factory strategy."
Express LUCK
Sherman Chan, Vice President
“When compared to other AI flows, MemryX finished the crossing line in first place …”
Hackster.io
Community Build Review
Work with us.
Collaborating on a design, evaluating the MX3, or looking for a role — tell us what you need.
