Microsoft-backed tech startup d-Matrix launches first AI chip to revolutionize chatbots and video generation

Silicon Valley-based AI startup d-Matrix announced Tuesday that it has started shipping its first AI chip, designed to support applications like chatbots and video generators.

This milestone comes just over a year after the company secured $110 million in funding to fuel the development of chips for generative AI platforms such as ChatGPT. So far, d-Matrix has raised over $160 million in funding. d-Matrix backers include Microsoft’s venture capital arm, Temasek, a Singapore-based investment firm, and Playground Global.

Early customers are already testing the chip samples, with full-scale shipments expected to ramp up next year. While the company hasn’t disclosed specific customer names, it did confirm that Super Micro Computer plans to offer servers equipped with d-Matrix chips.

Founded in 2019 by Sid Sheth and Chaloem Khompitoon, d-Matrix focuses on developing AI chips tailored to generative AI applications. The company is building its first compute engine modeled after the human brain, targeting a range of inferencing tasks for cloud and edge infrastructure markets, which together represent a multi-billion-dollar opportunity.

The chips leverage advanced digital “in-memory compute” technology to make AI processing more efficient. This approach reduces energy consumption while handling the massive data requirements of generative AI, ensuring the technology is optimized for these specific workloads.

What differentiates d-Matrix from a major player like Nvidia is its focus on inference processing. While Nvidia dominates in training large AI models, d-Matrix has carved out its niche in managing the tasks required after a system is trained—handling the real-time interactions and requests from end users.

“We have solved the computer architecture,” said Sasha Ostojic, partner at Playground. “We have solved the low power requirements and the needs of a data center – (we) built a software stack to deliver the lowest latency in the industry by orders of magnitude.”

The startup’s chips are built to manage multiple user requests simultaneously on a single chip. This capability is particularly valuable for applications such as video generation, where users often seek real-time customization and updates.

“We are getting a lot of interest in video use cases,” d-Matrix’s CEO Sid Sheth told Reuters. “We have customers saying, ‘We want to generate videos, and we want a collection of users, all interacting with their own respective video.’”

By complementing Nvidia’s dominance in AI training, d-Matrix is positioning itself as a critical player in inference processing—a step that’s key to scaling generative AI for everyday use.