Bharat Semi raises $24M seed for programmable AI sensors
A New Delhi chip startup has closed a $24 million all-equity seed round to build programmable sensors for physical AI. Info Edge Group, South Park Commons and Indus Bridge Ventures led the round, according to TechCrunch. Valour Capital, Micromax, Gaingels and individual investors also took part, and Entrackr named two angels, Mukul Agarwal and Karthik Sundar Iyer. No valuation has been disclosed. The company was founded in late 2021 by Vrinda Kapoor and Vinayak Dalmia, who serves as managing director, and its chief technology officer previously held the same role at image-sensor maker OmniVision.
The chips combine image capture, processing and AI on one chip, so devices can run inference at the point of capture rather than relying on separate processors. Kapoor says that integrated design could cut system costs by at least 70% compared with stacks built from a separate image sensor, image signal processor and system-on-chip, as TechCrunch and TokenPost reported. Entrackr reports the company claims up to 99% less data movement between components and up to ten times lower system costs than multi-chip systems. Bharat Semi plans to stay fabless, outsourcing manufacturing while it focuses on design and software.
The money will fund a bigger engineering team, product completion and market launch. Kapoor told Mint, as reported by CASRAI, that the round covers about two and a half years of work, with a goal of becoming revenue-positive and deploying roughly six products with major customers. The company is pre-revenue. First-generation sensors are in prototyping, second-generation prototypes are expected by mid-2027, and customer delivery is targeted for the end of 2027. Kapoor has said the company holds design wins with General Atomics and Kaynes Semicon, and it employs about 75 people, with 10 to 12 in the US.
Why it matters
For teams building robots, drones, cameras or vehicles, the pitch is that handling perception and inference on one chip can cut data movement, power draw and component cost. If that holds, sensor-level compute changes which parts a hardware team must source and integrate, and it shortens the path from capture to decision. Engineers building edge devices should track this layer early, because it can shape architecture choices before the software stack is written.