Nvidia plans to invest in d-Matrix, a Santa Clara, California, startup that designs chips for running AI models, The Information reported Thursday, citing three people with knowledge of the deal, according to Investing.com. The financial terms were not clear, and neither company has publicly confirmed the investment.

What Happened

According to the report, the investment is part of a broader effort by Nvidia to make rival chip designs compatible with its technology and bring them into its hardware ecosystem, rather than competing with them only head-on. No stake size, valuation or closing date has been disclosed.

d-Matrix, founded in 2019, builds accelerators for inference, the stage at which trained AI models generate answers for users. Its current Corsair platform uses a memory-centric architecture aimed at generative AI workloads. The company raised a $275 million Series C in November 2025 at a $2 billion valuation, bringing its total funding to about $450 million. In July, The Information reported that d-Matrix was seeking new funding at a valuation of at least $5 billion.

RelatedFirmus Pulls A$44 Billion ASX Listing After Weak Investor Demand

Why It Matters

The two companies are already technical partners. In September, d-Matrix announced a multi-year arrangement under which its next-generation Raptor chips will plug into Nvidia's MGX rack designs through NVLink Fusion, alongside Nvidia Vera CPUs, NVLink switches, BlueField-4 data processing units and ConnectX-9 and Spectrum-X networking, according to MLQ.ai.

That design pairs the two companies' chips through a technique called speculative decoding. A smaller draft model on d-Matrix silicon proposes sequences of tokens, and Nvidia GPUs running the larger model check them, which can speed up responses. The approach is part of a wider push toward disaggregated inference, where different stages of serving a model run on hardware suited to each.

An equity stake would give Nvidia a financial interest in a company that pitches cheaper, faster inference than GPUs alone. As inference takes a growing share of AI computing spending, startups such as Groq, Cerebras and d-Matrix have marketed themselves as alternatives to Nvidia. Licensing NVLink Fusion and investing in those companies lets Nvidia keep its racks, interconnects and software at the center of systems that use rival accelerators.

Context

The reported move follows Nvidia's $20 billion licensing and hiring deal with inference chipmaker Groq in December 2025, previously its largest arrangement with an AI chip startup. d-Matrix's existing backers include Microsoft's venture fund M12, Temasek, Qatar Investment Authority and EDBI, and its Series C was led by Bullhound Capital, Triatomic Capital and Temasek.

What to Watch

Key open questions are whether either company confirms the deal and its terms, and whether Nvidia's money comes as part of the round d-Matrix was reported to be raising at a valuation of $5 billion or more. Raptor-equipped Nvidia racks are the other milestone; their arrival will show how much of the partnership translates into deployed systems.