US startup d-Matrix to use Nvidia technology for AI servers
US startup d-Matrix, based in Santa Clara, California, plans to use Nvidia’s chip-linking technology in server systems for artificial intelligence. As Channel NewsAsia reports, the company will use NVLink Fusion to directly integrate its own processors into Nvidia data center systems.
Raptor chips for inference
d-Matrix’s new chips, called Raptor, will be connected to Nvidia server racks through NVLink Fusion. According to the startup, this technology includes connectors and specialized memory that make it possible to connect specialized AI chips to larger Nvidia data center systems.
d-Matrix specializes in inference — the use of already trained artificial intelligence models in everyday tasks. The company said the joint systems are designed for fast, low-latency AI services, including coding assistants, chatbots, and voice agents.
More current news is available on the UA.News Telegram channel Telegram.
Timeline and partners
Nvidia-compatible server racks are expected to be available in 2027. d-Matrix plans to complete the final design stage of the Raptor chips by the end of the current year. The companies did not disclose the financial terms of the cooperation.
The startup is also working with connectivity technology company Astera Labs to create customized solutions intended to ensure fast data flow within the system. Microsoft has supported d-Matrix since its $110 million funding round in 2023. The company shipped its first AI chip in November 2024, and it was valued at $2 billion when it raised $450 million last year.
