← All companies

Chips / Company profile

d-Matrix

What d-Matrix does

d-Matrix is a private AI semiconductor company building digital in-memory compute (DIMC) hardware and an end-to-end inference software and networking stack aimed at reducing the latency, energy use, and cost of generative AI inference in data centers. Its flagship product family is centered on the Corsair™ inference accelerator platform, which combines on-chip “performance memory” for fast token generation with off-chip “capacity memory” for offline/batched workloads, plus Aviator™ software for compiling and executing inference graphs over the company’s architecture. Corsair is designed to scale from card-level deployments up to rack-level systems using standard fabrics, and to operate in heterogeneous inference clusters where it can pair with conventional GPU compute for different parts of the inference pipeline. d-Matrix also offers JetStream™ I/O accelerators (transparent NIC/ethernet-based networking) and SquadRack™ reference rack-scale solutions that integrate Corsair accelerators, JetStream networking, and Aviator software into an off-the-shelf-style deployment blueprint.

Strategically, d-Matrix has been moving from silicon milestones (e.g., chiplet interconnect and DIMC engine improvements) toward production and deployment readiness. In mid-2026 it announced Corsair is in full production with volume shipping to priority customers, and it has also been acquiring capabilities to reduce “time to deploy” in heterogeneous inference setups—most recently acquiring Wallaroo.ai (August 2026) and previously acquiring GigaIO’s data center business (April 2026). The company’s investor and partner ecosystem includes hyperscale- and platform-adjacent participants and networking/system partners, positioning it as an infrastructure provider for low-latency inference workloads rather than a standalone training-only chip vendor.

News

Company record

Aug 22, 2026 · Official · d-Matrixd-Matrix Reaches New Milestone in Efficient AI Inference with Jayhawk II

d-Matrix announced Jayhawk II, describing a DIMC engine and chiplet interconnect and providing comparative claims about memory bandwidth, compute scalability, and throughput/latency/TCO relative to incumbent GPU solutions; the company positioned Jayhawk II for demo/evaluation.

Aug 03, 2026 · Official · d-Matrixd-Matrix Acquires Wallaroo.ai to Speed up Deployment of Heterogeneous AI Inference Workloads

d-Matrix announced the acquisition of Wallaroo.ai, described as an AI inference deployment and orchestration software company, to add deployment/orchestration technology and engineering talent to d-Matrix’s silicon-to-software inference stack.

Jul 07, 2026 · Official · d-MatrixParasail to Combine NVIDIA AI Infrastructure with d-Matrix Accelerators to Achieve 10x Faster Token Generation

Parasail and d-Matrix announced a production deployment of d-Matrix Corsair accelerators alongside NVIDIA Hopper and NVIDIA Blackwell architectures, positioned to deliver up to 10x faster token generation and more cost-efficient inference services.

Jun 25, 2026 · Official · d-Matrixd-Matrix Corsair Inference Accelerator Wins 2026 AI Breakthrough Award

d-Matrix announced Corsair received the 2026 AI Breakthrough Award (AI Processor Innovation Award), with the company describing Corsair as an emerging infrastructure choice for heterogeneous AI compute and token generation when working alongside GPUs.

Jun 09, 2026 · Official · d-Matrixd-Matrix Corsair AI Inference Platform Enters Full Production to Meet Customer Demand

d-Matrix stated Corsair is in full production and that products would begin shipping in volume to priority customers; the announcement also describes integration into rack-scale systems (SquadRack) and manufacturing partnerships (including TSMC and Alchip Technologies).

Apr 02, 2026 · Official · d-Matrixd-Matrix Boosts Rack-scale AI Capabilities With Acquisition of GigaIO Data Center Business

d-Matrix announced acquisition of GigaIO’s data center business, described as expanding d-Matrix’s capabilities for rack-scale deployments and adding systems engineering talent and data-center technologies.

Mar 12, 2026 · Official · d-Matrixd-Matrix and Gimlet Labs to Deliver 10x Speed Ups, Massive Power Efficiency for Frontier AI Workloads

d-Matrix and Gimlet Labs announced that Gimlet Cloud would incorporate d-Matrix Corsair accelerators alongside GPUs to deliver claimed 10x speedups for agentic AI inference workloads.

Nov 12, 2025 · Official · d-Matrixd-Matrix Raises $275 Million to Power the Age of AI Inference (Series C)

d-Matrix announced it closed a $275 million Series C at a $2 billion valuation and stated total raised to date reached $450 million; the release describes an inference performance/cost/energy positioning and lists investors and co-leads.

Show 2 earlier updates
Source map · 2 recurring channels · 13 references

Still resolving: Newsroom · Official X

Funding

Latest disclosed valuation$2B
Tracked capital$429M5 sourced rounds

What it builds

Corsair™ inference accelerator platform

Corsair is d-Matrix’s generative AI inference compute platform that uses a digital in-memory compute architecture with integrated performance memory for low-latency token generation and off-chip capacity memory for offline/batched inference; it is packaged and scaled from card to rack configurations and is paired with JetStream networking for multi-node deployments.

source ↗
JetStream™ I/O accelerator (Transparent NIC)

JetStream is d-Matrix’s I/O accelerator / transparent NIC designed to deliver low-latency accelerator-to-accelerator communication using standard Ethernet in multi-node deployments.

source ↗
Aviator™ software stack

Aviator is d-Matrix’s software stack for deploying and executing models on the Corsair architecture, described as integrating with open frameworks such as PyTorch and Triton and using an MLIR-based compiler toolchain plus runtime components.

source ↗
SquadRack™ rack-scale solutions

SquadRack is d-Matrix’s rack-level solution/reference architecture that integrates Corsair accelerators, JetStream networking, and Aviator software into an industry-standard rack-scale deployment blueprint.

source ↗

Key metrics

Milestones & partnerships