India Just Bet $24M on Chips That See and Think at the Same Time
Bharat Semi raised $24M to build image sensors with AI baked into the silicon — cutting camera system costs by up to 70% and turning the sensor into the brain.
- Published
- 16 Nov 2026
- Written by
- Md Tayobur Rahman
- Topic
- Tech News
- Reading time
- 4 min
Article
A chip startup out of New Delhi just raised $24 million before shipping a single product — and the pitch is unusually specific. Bharat Semi wants to put the AI inside the image sensor itself, so the camera stops being a dumb eye that ships raw pixels off to a separate processor.
The seed round, reported by TechCrunch, is all equity and was led by Info Edge Group, South Park Commons and Indus Bridge Ventures, with Valour Capital, Micromax, Gaingels and a group of individual investors chipping in. The company is nearly five years old, still pre-revenue, and its first commercial production run is targeted for 2027.
One chip instead of three
Today's camera pipeline is a chain: a sensor captures light, an image signal processor cleans it up, and a separate system-on-chip runs whatever AI model makes sense of the scene. Three parts, three power budgets, three cost lines.
Bharat Semi's bet is to collapse that into programmable sensors that capture, process and run inference on a single die. According to co-founder Vrinda Kapoor, the approach can cut system cost by at least 70% versus conventional setups — because you no longer need the extra silicon sitting behind the lens.
That kind of saving matters most where cameras are deployed in volume and compute is expensive: smart homes, industrial automation, surveillance and aerospace. Kapoor's examples are refreshingly concrete. An air conditioner that notices a room is empty and dials itself back. A drone that interprets what it sees without carrying a costly companion computer.
Design wins before the product exists
The most interesting part of the story is who has already signed on. Bharat Semi says it has design wins with General Atomics, a U.S. defense technology company, and India's Kaynes Semicon — even though the sensors are still under development.
First-generation parts are in prototyping. A second generation with heavier on-chip AI is expected to hit prototype by mid-2027, with customer hands-on by the end of that year. The fresh capital goes toward engineering headcount, finishing product development and getting to market.
Why this fits India's chip ambitions
India has committed billions to building a domestic semiconductor industry, hoping to become an alternative manufacturing hub as companies look to diversify beyond China. It has not yet established itself as a major fab player, but the AI boom gives Indian startups a narrower opening: design specialised chips for global buyers rather than compete on leading-edge manufacturing.
Bharat Semi is deliberately fabless. It outsources fabrication and concentrates on design and software. The company has around 75 employees across India and the U.S., including 10 to 12 in Santa Clara doing engineering and business development, and plans to grow that U.S. presence to roughly 20 by mid-2027 while also hiring in Europe. A separately announced 2024 fabrication project with a U.S. Space Force technology partnership is still in development and is not being funded by this round.
The takeaway
"Our aim here is to move from sensors to sense-making solutions," Kapoor told TechCrunch. That line is the whole thesis. For years the value in computer vision sat with whoever owned the model; the sensor was a commodity. If inference moves onto the sensor, the economics flip — cheaper systems, lower bandwidth, faster decisions at the edge, and a hardware company that suddenly owns part of the intelligence layer.
The risk is real: pre-revenue, pre-product, with a 2027 timeline and defence-grade customers who will not tolerate slippage. But the direction is right. The next wave of AI will not all live in data centres. A lot of it will live in the thing pointed at the world, and whoever makes that thing smart at a fraction of today's cost gets to define the price floor for everyone else.
Share
Contact
Available for Laravel platforms, native Android apps, and the systems that run behind them.