Who is d-Matrix?
d-Matrix is a semiconductor and systems company that develops compute hardware and software for large-scale Artificial Intelligence (AI) inference workloads in data centers.
- Compute-in-memory AI acceleration platforms for data center inference (AI infrastructure)
- Digital in-memory computing architecture targeting transformer and Large Language Model (LLM) inference
- Chiplet-based system designs and modules for deployment in existing server environments
- Software stack for model deployment, compilation, and optimization on d-Matrix hardware
- Focus on power efficiency and Total Cost of Ownership (TCO) for enterprise AI inference at scale
Show more
More About d-Matrix
d-Matrix focuses on AI inference infrastructure, with hardware and software designed for transformer and LLM workloads in enterprise and cloud data centers.
The company develops digital in-memory compute architectures that aim to keep data close to compute elements, reducing data movement relative to conventional Central Processing Unit (CPU) and Graphics Processing Unit (GPU) designs. This approach targets lower power consumption and higher efficiency for matrix operations that dominate transformer-based models. d-Matrix positions its solutions for organizations running inference on models such as large language models, recommendation systems, and other deep learning workloads.
d-Matrix offers AI acceleration platforms (AI infrastructure) that integrate its chips into boards or systems compatible with standard data center environments. These platforms are intended to be deployed alongside or as alternatives to general-purpose accelerators, with an emphasis on serving high-throughput, latency-sensitive inference. The hardware is tailored to batch and real-time inference scenarios common in enterprise applications such as search, conversational interfaces, content ranking, and analytics.
On the software side, d-Matrix provides a software stack (AI infrastructure software) that includes tools for model import, graph compilation, and runtime execution on its hardware. The stack is designed to integrate with common Machine Learning (ML) frameworks and to support quantization and other optimizations for running large models within data center power and cost constraints. This software layer is central for mapping transformer architectures efficiently onto the company’s in-memory compute engines.
From a marketplace taxonomy perspective, d-Matrix fits into AI infrastructure, data center acceleration, and specialized silicon for inference. Its offerings are positioned for enterprises, cloud providers, and service operators that deploy large-scale AI services and seek dedicated inference capacity separate from training infrastructure.
Our description of d-Matrix. Updated January 2026.
