MemryX is an edge-AI semiconductor company designing accelerator chips that run neural-network inference efficiently on devices rather than in the cloud. Founded in 2019 and based in Ann Arbor, Michigan, MemryX targets smart cameras, autonomous vehicles, robotics, industrial automation, drones, healthcare equipment and IoT, where low power and on-device privacy matter. Its production MX3 accelerator claims more than 20x better performance per watt than mainstream GPUs for targeted inference, and the company is designing a next-generation MX4 plus a roadmap toward data-center dataflow architectures. MemryX pairs its silicon with a software stack that simplifies deploying standard AI models to the edge. In March 2025 MemryX raised a $44 million Series B, bringing total funding to roughly $63 million, with backing from HarbourVest, M Ventures, Arm IoT Fund, eLab Ventures, Motus Ventures and the NEOM Investment Fund. The capital funds MX3 production ramp-up and MX4 development as edge-AI demand accelerates.
MemryX
ActiveEnergy-efficient edge AI accelerator chips for on-device inference
Total raised
$44M
1 round
Stage
Series B
Team
51-200
since 2019
Pricing
Freemium
free plan
Founded
2019
Ann Arbor, United States
Agent-ready
—
MX3 edge AI accelerator chip
Up to 20x better performance per watt
On-device inference for privacy
Dataflow architecture
Software stack for model deployment
Support for standard AI frameworks
Scalable multi-chip configurations
Targeting cameras, robots, vehicles, IoT
25/100
Early
MCP server
Public API
Webhooks
OAuth 2.0
SDKs
No public agent surfaces detected yet.
— Series B $44M incl. Arm IoT Fund +5
Capital network
$63M raised ·6 backers·6 network links
- Backers6
- Shared portfoliocompanies these backers also fund
- Extended networkfunds that co-invest alongside them
Nebius
Full-stack AI cloud with large-scale GPU clusters for training and inference
Foundation ModelsAI Infrastructure
Celestial AI
Photonic Fabric optical interconnect for AI infrastructure
AI Infrastructure
d-Matrix
Digital in-memory compute (DIMC) chiplet-based hardware purpose-built for AI inference in the
AI Infrastructure
Chainguard
Secure, minimal container images for software and AI supply chains
AI InfrastructureAI for Cyber Defense
Alcatraz AI
AI facial authentication that replaces the access badge with your face
AI InfrastructureAI for Cyber Defense
Zilliz
Fully managed vector database for AI, built by the creators of Milvus
AI InfrastructureVector Databases
- What does MemryX build?
- MemryX designs energy-efficient AI accelerator chips for on-device inference at the edge.
- How efficient is the MX3?
- MemryX claims the MX3 delivers more than 20x better performance per watt than mainstream GPUs for targeted inference.
- How much has MemryX raised?
- About $63 million total, including a $44M Series B in March 2025.
Discussion
Sign in to join the discussion.
Sign inExplore more around MemryX
Contextual paths to related AI startups, deals and rankings.