This Microsoft-backed startup wants to rival Nvidia in AI chips: Meet its Indian-American founders

D-Matrix, a California-based startup co-founded by Indian-American entrepreneurs Sid Sheth (CEO) and Sudeep Bhoja (CTO), is challenging Nvidia’s dominance in AI chips with its Corsair processor. The company targets AI inference workloads like chatbots and voice agents, where low-latency performance is critical. Unlike traditional GPUs relying on DRAM, Corsair uses an SRAM-based design, placing memory and compute closer together to reduce energy use and speed up processing. Sheth, a veteran semiconductor executive, and Bhoja, former CTO of Inphi’s datacenter unit, claim Corsair delivers up to 10x faster inference speeds, one-third the cost, and five times better energy efficiency when paired with Nvidia’s Blackwell GPUs, according to research by Gimlet Labs. The chip is manufactured by TSMC using a 6nm process and is already shipping to hyperscalers, neocloud providers, and AI labs in the U.S., Middle East, and Southeast Asia. However, Stanford electrical engineering professor Rick Bahr warns that SRAM-based chips may struggle with today’s largest AI models, which require trillions of parameters. These models exceed the capacity of SRAM designs, limiting D-Matrix’s scalability for high-end AI workloads. The company is addressing this by developing rack-scale systems called SquadRack in partnership with Broadcom, Arista Networks, and Super Micro Computer. D-Matrix has secured backing from investors, including Microsoft, and is expanding its product line. While its Corsair chip shows promise for latency-sensitive applications, its long-term success depends on overcoming architectural limitations for next-generation AI models.