en.Wedoany.com Reported - US AI chip company d-Matrix announced that its Corsair AI inference acceleration platform has entered full production and will begin volume shipments to priority customers. Headquartered in Santa Clara, California, the company's initial shipments target hyperscale cloud providers, Neocloud providers, and frontier AI labs, primarily for data center AI inference scenarios.

The Corsair is not a training chip; its focus is on large model inference.

In generative AI applications, after model training is complete, every user query, voice interaction, code generation, and multimodal content generation requires backend inference computation. d-Matrix states that the Corsair is designed for low-latency inference tasks and can form heterogeneous, decoupled computing systems with GPUs, allowing different computing units to handle tasks best suited to them. The company views this full production milestone as a commercial delivery point, marking the Corsair's transition from early customer validation to volume supply.

Supply chain is a prerequisite for scaling deliveries. d-Matrix stated that it has secured multi-year supply and manufacturing service guarantees and will proceed with product shipments based on this foundation. Alchip Technologies participated in the design and mass production support of the Corsair. Its management noted that the two companies have collaborated since the early design phase of the Corsair and will continue to support the platform's scaling.

d-Matrix's product portfolio is not limited to a single accelerator card. According to the company, the Corsair can be paired with the JetStream I/O accelerator and the Aviator software stack to form an inference platform for data center rack deployment. Reference solutions on the company's website include multiple servers, multiple accelerator cards, and various memory configurations, aiming to reduce data movement overhead in large model inference while improving response speed and performance per watt.

This shipment milestone arrives during a shift in the focus of AI infrastructure construction. Over the past two years, market investment has primarily focused on training clusters and GPU computing power expansion. With chatbots, agent applications, real-time voice agents, and enterprise AI tools entering high-frequency usage, inference-side latency, concurrency capabilities, and operating costs have become key metrics in customer procurement. With the Corsair entering full production, d-Matrix must now prove that its platform can operate stably under real data center workloads and achieve compatibility with servers, networks, models, and the software ecosystem.

Subsequent milestones will focus on customer deployment scale, rack-level performance validation, and long-term supply cadence. If volume usage by priority customers progresses smoothly, d-Matrix will secure a more defined delivery position in the commercialization competition for US AI inference chips.

This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com