Nvidia is expanding the reach of its data-centre ecosystem by allowing AI chip startup d-Matrix to integrate its processors directly into Nvidia-based server infrastructure.
D-Matrix will use Nvidia’s NVLink Fusion technology to connect its forthcoming Raptor inference processors into larger Nvidia data-centre systems.
The compatible server racks are expected to become available in 2027, while d-Matrix plans to complete the final design stage of Raptor before the end of 2026.
Financial terms of the collaboration were not disclosed.
The arrangement is important because it illustrates how Nvidia is attempting to maintain its position at the centre of AI infrastructure even when some of the processors inside a system are developed by other companies.
AI Moves From Training to Everyday Use
Much of the first stage of the generative AI boom was driven by the enormous computing resources required to train models.
The next stage increasingly involves inference — running those trained models for millions or potentially billions of daily requests.
D-Matrix specialises in processors designed specifically for that task.
Its Raptor chips are being developed for applications requiring extremely fast responses, including AI coding assistants, chatbots and voice agents.
The startup is also working with Astera Labs on specialised connectivity technology intended to move information quickly across AI systems.
Nvidia Builds an Ecosystem, Not Just Chips
NVLink Fusion could become strategically important for Nvidia.
The technology allows custom processors to connect with Nvidia infrastructure using specialised memory and high-speed links.
Rather than forcing data-centre customers to choose entirely between Nvidia and rival processors, the approach potentially allows other chips to operate within Nvidia’s wider architecture.
That could help Nvidia preserve its influence over AI data centres even as hyperscalers and startups develop specialised silicon.
Microsoft Backs d-Matrix
D-Matrix also has substantial financial backing.
Microsoft has supported the company since a $110 million financing round in 2023, while the startup reached a valuation of about $2 billion after raising $450 million in 2025.
It shipped its first AI processor in November 2024.
The larger commercial question is whether specialised inference processors can gain meaningful market share as AI usage scales.
If demand for AI assistants, autonomous agents and other applications grows as expected, inference could ultimately represent one of the largest computing markets created by the AI revolution.
That explains why Nvidia is simultaneously defending its own processor leadership while creating technology that allows other chipmakers to participate within its ecosystem.
