D-Matrix Starts Shipping Corsair Chips, Claiming 10x Faster AI Inference Than Nvidia GPUs
Updated
Updated · CNBC · Jun 9
D-Matrix Starts Shipping Corsair Chips, Claiming 10x Faster AI Inference Than Nvidia GPUs
3 articles · Updated · CNBC · Jun 9
Summary
D-Matrix said it will begin shipping its Corsair AI inference chips this month to unnamed high-profile customers, with commitments from hyperscalers, neoclouds and frontier AI labs.
Corsair targets small, latency-sensitive inference workloads and, when paired with an Nvidia Blackwell GPU, can deliver 10 times faster performance, three times lower cost and up to five times better energy efficiency than a standalone GPU, the company said.
The chip uses SRAM integrated with compute on a single die, avoiding the DRAM and high-bandwidth-memory bottlenecks that constrain many GPU systems; about 90% of customers are in the U.S., with others in the Middle East and Southeast Asia.
That design also limits Corsair's reach: analysts and academics said SRAM-based chips excel at fast inference but cannot handle the trillion-parameter reasoning models used by leaders such as OpenAI and Anthropic.
Founded in 2019, D-Matrix has raised about $500 million at roughly a $2 billion valuation, joining a wave of AI chip challengers seeking niches around Nvidia in what its CEO called a potential $1 trillion market.
Valued at $2 billion, is D-Matrix the next AI giant or a temporary fix until the global memory shortage ends?
Is D-Matrix’s bet on SRAM a brilliant bypass of the AI memory crisis, or a fatal flaw for future, larger models?
d-Matrix Corsair Hits Full Production: DIMC™ Architecture Delivers Ultra-Low Latency and High Throughput for Enterprise AI Inference
Overview
d-Matrix has launched its Corsair AI inference platform into full production to meet the urgent demand for specialized AI solutions. As AI applications push the limits of traditional GPU-only infrastructure, hyperscalers and AI labs need faster and more interactive experiences at scale. Startups like d-Matrix are now challenging established leaders such as Nvidia in the intensifying AI chip market. Corsair is designed to deliver high throughput and extremely low latency, directly addressing the needs of real-time, complex AI workloads. This marks a pivotal shift towards purpose-built hardware for next-generation AI inference.