AI chip startup d-Matrix adopts NVIDIA’s NVLink Fusion technology to connect its Raptor processors with data-centre systems, targeting low-latency inference applications such as chatbots, coding assistants and voice agents.
D-Matrix, an AI chipmaker based in Silicon Valley, is implementing NVIDIA’s NVLink Fusion in order to embed its chips directly into NVIDIA’s data centre ecosystem. This move comes amid growing needs for AI inference, which refers to running trained models in practical applications.
The d-Matrix company is designing its new Raptor processors especially for AI inference processing. Though NVIDIA’s powerful graphics processors still prevail in the field of AI model training, the growing popularity of generative AI services requires a more specialised hardware for fast inference processing.
Also read: UAE and Germany Strengthen Longstanding Ties During State Visit
Through this partnership, Raptor processors will be connected to NVIDIA servers via NVLink Fusion, which offers the unique connection and memory technology that is necessary for the development of custom AI accelerators within NVIDIA’s larger data centre ecosystems. It is anticipated that Raptor processors will undergo final design by the end of 2026, while server hardware is due in 2027.
These integrated systems have been developed in order to cater to application scenarios where response time is an issue. Some of these are artificial intelligence based code assistants, bots, and voice agents that need rapid processing capabilities. Using the combination of d-Matrix's processors optimised for inference along with NVIDIA's data centre technology, the firms want to tackle this challenge.
The alliance also underscores the growing significance of special-purpose AI accelerators in addition to the use of standard GPUs. With increasing adoption of AI technologies, there is an increased need for computing hardware to manage specific workloads rather than generic accelerators.
In order to improve the interconnectivity of its system, d-Matrix is also collaborating with Astera Labs, which is a semiconductor interconnect company. The collaboration is centred on designing custom high-speed data links in order to ensure effective communication between the components in the AI ecosystem.
Also read: How AI Agents are reinventing the Factory Floor
The start-up has witnessed considerable investor attention owing to the rising demand for AI hardware. Microsoft has invested in d-Matrix since 2023, when it raised a funding of $110 million. The company is shipping its first AI chip in November 2024 and is valued at $2 billion after raising $450 million in 2025.
This collaboration paves the way for d-Matrix to strengthen its footprint in the fast-evolving world of AI inference applications while facilitating the ability of NVIDIA’s hardware to support specialised third-party processors. It is also indicative of the larger trend in the field of AI computation towards heterogeneity.