Nvidia pairs GPUs with startup d-Matrix's inference chips instead of competing with it

Nvidia is combining its GPUs with inference chips from startup d-Matrix for a joint system to run AI models — the first customer of the new solution will be AI cloud company Parasail, reports The Information.

Why Nvidia Chose Partnership Over Competition

The AI chip market today is split into two major segments. The first is model training: here Nvidia is an undisputed leader, with its GPUs becoming the de facto standard for major AI labs and cloud providers. The second is inference — that is, directly servicing user requests to already trained models. The requirements are fundamentally different: in inference, low latency, energy efficiency, and cost per token matter more than peak computational power.

It is precisely in the inference chip niche that d-Matrix specializes — a startup developing specialized hardware for fast and cheap servicing of language models in production. According to The Information, instead of competitive pressure, Nvidia chose integration: the companies will create a joint system that combines Nvidia GPUs with d-Matrix inference chips into a single product. This architecture covers both key scenarios — heavy computations on GPUs and economical high-speed inference on specialized chips.

Who Will Receive the Joint System First?

The first customer of the new equipment will be Parasail — an AI cloud company that provides computational infrastructure for running large language models. Parasail specializes in corporate clients who need scalability, low latency, and predictable operational expenses when running LLMs. Specific delivery timelines and commercial terms of the deal are not disclosed.

Key facts:

• Source: The Information
• The system will combine Nvidia GPUs and d-Matrix startup inference chips
• First buyer — AI cloud company Parasail
• The product is aimed at AI model inference, not their training

Nvidia Changes Strategy: Integration Instead of Displacement

The partnership with d-Matrix reflects an important shift in AI infrastructure. Nvidia traditionally built a closed vertical ecosystem: GPUs, proprietary drivers, CUDA platform, developer tools. However, the inference market is fragmented — actively competing in it are Groq, Cerebras, Etched, SambaNova, Tenstorrent and other startups with their incompatible architectures.

Integrating d-Matrix instead of acquiring or ignoring it is a possible response to this fragmentation. For Nvidia, it is advantageous when niche players become part of its ecosystem without disrupting the core GPU business. For startups, this means a fundamentally new path to scaling: entering the wider market through partnership with a leader, rather than in direct collision with it.

What This Means

The inference market in 2025–2026 is growing faster than the training segment and is becoming one of the key competition points in AI infrastructure. If the joint system from Nvidia and d-Matrix proves successful, the strategy of technological integration instead of displacement of niche competitors could become a template for other partnerships in AI hardware.