MemryX
Company Information
Headquarters
Ann Arbor, Michigan, USA
Year Founded
2019
Industries
Semiconductors, Industrial Automation, Robotics, Automotive, Internet of Things, Computer Hardware, Intelligent Infrastructure
What Is MemryX?
MemryX is a fabless semiconductor company developing AI inference accelerators for running artificial intelligence workloads at the edge and in server environments. Its technology combines specialized AI hardware with a proprietary at-memory dataflow architecture designed to process neural-network models efficiently while reducing data movement and power requirements. The company’s current production technology centers on the MX3 accelerator, which can be deployed as individual chips or integrated into modules and systems for computer vision and other inference workloads.
What Does MemryX Do?
MemryX develops hardware and software that allow organizations to run AI inference locally rather than relying exclusively on cloud-based processing. Its accelerators offload deep neural-network workloads from a host CPU and can work with x86, Arm, and RISC-V processors across supported operating systems.
The technology is particularly suited to real-time computer vision and sensor processing, including object detection, segmentation, pose estimation, depth estimation, and OCR. MemryX provides chips, M.2 and PCIe modules, USB-connected products, server-oriented configurations, and development software for integrating its acceleration technology into new or existing computing systems.
MemryX Technology
At the center of MemryX’s technology is the MX3 accelerator and its proprietary at-memory dataflow architecture. Instead of repeatedly moving information between separate compute and memory resources, the architecture keeps data close to computation. MemryX says this approach reduces memory bottlenecks and unnecessary data movement during AI inference.
The architecture uses pipelined execution optimized for streaming inputs, making it particularly applicable to camera feeds and other continuous data sources. Multiple MX3 chips can operate together as a logical unit, allowing the architecture to scale across different hardware configurations.
MemryX also provides a software development environment that includes a neural compiler, bit-accurate simulator, runtime tools, APIs, performance analysis tools, and application examples. It supports common frameworks including PyTorch, TensorFlow, TensorFlow Lite, ONNX, and Keras. MemryX says models can be deployed without requiring users to retrain, prune, or compress them for its hardware.
Why MemryX Matters
MemryX addresses an important challenge created by the movement of AI from centralized cloud infrastructure into edge devices and physical systems: performing increasingly complex inference while controlling power consumption, latency, memory bandwidth, and host-computing requirements.
Its architecture is especially relevant to the growth of edge AI and physical AI, where cameras, robots, industrial equipment, vehicles, and intelligent infrastructure need to process information close to where it is generated. MemryX is also expanding the same underlying acceleration architecture from embedded and edge applications toward higher-density server infrastructure, illustrating how specialized AI inference hardware is evolving across computing environments.
MemryX Products & Platforms
MX3 Edge AI Accelerator
The MX3 is MemryX’s production AI inference accelerator and the foundation of its current product portfolio. It uses the company’s at-memory dataflow architecture and is designed for neural-network inference involving computer vision and sensor-processing workloads. The chip supports multiple model frameworks and can be scaled by combining multiple MX3 accelerators.
MX3 M.2 AI Accelerator Module
The MX3 M.2 module integrates multiple MX3 accelerators into a compact M.2 2280 form factor for adding edge AI acceleration to existing systems. The module offloads deep neural-network processing from the host CPU and connects through PCIe, allowing developers to add acceleration to compatible embedded computers and PCs without redesigning the underlying computing platform.
Cascade Platform
Cascade is MemryX’s broader AI deployment platform family, extending the MX3 architecture across embedded devices, edge systems, and higher-density server infrastructure. MemryX expanded the portfolio in 2026 as part of its effort to provide a common acceleration architecture across different deployment scales.
MemryX Developer Hub / SDK
MemryX provides development software around its hardware, including a neural compiler and runtime tools for preparing and deploying trained AI models. The environment also includes a simulator, model exploration and performance tools, APIs, application templates, tutorials, and code samples designed to support development across multiple host platforms.
MemryX Applications
MemryX technology is primarily designed for applications requiring low-latency AI inference close to the data source. Computer vision is a major use case, including video analytics, object detection, segmentation, pose estimation, smart cameras, access control, traffic monitoring, and other vision-based systems.
The company identifies applications across industrial automation and robotics, automotive systems, IoT, agriculture, healthcare and medical systems, video management, and smart vision. Its accelerators can also be paired with embedded computing platforms to process camera and sensor data while leaving the host processor available for application logic, communications, and other workloads.
Latest Companies
Frequently Asked Questions
What is the MemryX MX3 accelerator?
The MX3 is MemryX’s production AI inference accelerator chip. It uses the company’s at-memory dataflow architecture to execute neural-network workloads, particularly computer vision and sensor-processing models, while emphasizing low power consumption and efficient data movement.
Can MemryX run existing AI models?
Yes. MemryX supports models developed with frameworks including TensorFlow, TensorFlow Lite, PyTorch, ONNX, and Keras. Its development tools are designed to compile trained models for MX3 hardware without requiring model retraining, pruning, or compression in supported workflows.
Does MemryX require a specific CPU architecture?
No. MemryX states that its accelerators can operate alongside x86, Arm, and RISC-V host processors, allowing the accelerator to handle AI inference while the host performs other application processing.
Is MemryX designed for computer vision?
Computer vision is one of MemryX’s primary applications. Its software supports workloads including object detection, classification, segmentation, pose estimation, depth estimation, and OCR, while its pipelined dataflow architecture is designed to handle continuous streaming inputs such as cameras.
Can MemryX technology be used in robotics?
Yes. MemryX identifies robotics and industrial systems among its target applications. Its accelerators can process computer-vision and sensor workloads locally, making them applicable to robotic systems requiring real-time perception and edge inference.
Do You Have an Update for This Company?
Have an update, correction, or additional information about this company? Help us keep the futureTEKnow company database up to date by submitting the latest company information.
Related Companies
Latest Articles
OpenAI Astra for Law Brings GPT-6 Astra Into Legal Research and Law Firm Workflows
OpenAI’s Astra for Law combines GPT-6 Astra with specialized legal search, integrations, and governance controls. The launch shows how frontier AI is moving deeper into professional legal workflows.
Are your AI projects stuck on your own data?
AI projects often stall because business data isn’t consistent or trustworthy. Here’s how to identify data gaps before they become an AI roadblock.
EVAS Intelligence’s RISC-V AI Chips Land 1.5 Billion Yuan Series B to Power Global Expansion
EVAS Intelligence has raised 1.5 billion yuan to mass‑produce its RISC-V Epoch AI chips, deepen its full‑stack platform, and accelerate global expansion, signaling rising investor confidence in alternative AI compute.
Orkes Raises $60M to Orchestrate Enterprise AI Agents in Production
Orkes has raised 60 million dollars to turn its Netflix‑born workflow engine into a control plane for enterprise AI agents. The piece unpacks how its orchestration platform works, why investors are backing it now, and what this means for companies trying to move AI from pilots into production.
Sillage’s AI signal engine raises €1.7M to help enterprise sales teams follow the right buying signals
Paris-based Sillage has raised €1.7 million to launch an AI signal engine that helps enterprise sales teams follow the right buying signals, prioritize high-intent accounts, and improve pipeline efficiency without replacing human-led conversations.
Cloneable’s agentic AI platform targets the infrastructure knowledge gap with $4.6M seed round
Cloneable is launching an agentic AI platform for infrastructure operations that captures institutional knowledge from retiring experts and turns it into scalable agents, targeting energy, utilities, and other asset-heavy sectors. The $4.6M seed-backed startup aims to connect field data capture with back-office execution so founders and operators can modernize infrastructure faster and with fewer bottlenecks.
Ricursive Superintelligence’s $500M bet on self-improving AI raises the stakes in the AGI funding race
Excerpt: Ricursive Superintelligence has raised at least $500 million to build self‑improving AI, with GV and Nvidia backing a four‑month‑old lab at a $4 billion valuation. The deal crystallizes how quickly capital, talent and regulatory scrutiny are converging around autonomous AI research—and what that means for builders trying to compete outside the frontier club.
Bond AI raises US$2M to automate SME accounting in Brazil’s brutal tax maze
Brazilian startup BOND has raised US$2M to automate accounting for SMEs in Brazil’s complex tax system. Combining AI with human accountants, it targets a fragmented market of manual processes, aiming to turn back-office compliance into a faster, more strategic function for founders and operators.
Loop’s AI supply chain platform raises $95M Series C to predict disruptions before they hit
Loop just raised a $95M Series C to expand its AI-native supply chain platform, turning messy logistics data into early disruption warnings and better financial outcomes for global operators and shippers.
London AI startup Lua wants to give every business its own AI workforce with $5.8M seed round
Linkedin X-twitter-square Facebook-square Startups & Business News AI agents are finally moving from demos to the day-to-day stack of real
Factory.ai raises $150M Series C to build fully autonomous software engineering “factories”
Factory has raised a $150M Series C at a $1.5B valuation to scale its autonomous “Droids” platform, betting that enterprises will standardize software engineering around agent‑native missions, governance and cost‑optimized AI infrastructure.
Solidroad’s $25M bet on AI quality assurance for customer support
Solidroad has raised $25 million to bring AI-native quality assurance to every human and AI-powered customer interaction. The new funding will accelerate product development and expansion for support-heavy businesses leaning into automation.
futureTEKnow is focused on identifying and promoting creators, disruptors and innovators, and serving as a vital resource for those interested in the latest advancements in technology.
© 2026 All Rights Reserved.