Using deep tech is not creating it

Bhandari's argument is that India has confused using deep tech with creating it. Most Indian companies (80 to 90 per cent, on the split named here) are applying somebody else's model, wrapping a ChatGPT call in something that makes a problem easy, while the underlying layer stays foreign; he blames datasets, hardware and large-scale training infrastructure rather than talent, and expects India to leapfrog once those ease. Seafund's answer is a mandate with a hole deliberately cut in it: no direct-to-consumer, a conscious choice given a small fund and partners who describe their consumer instincts as close to non-existent. What is left is B2B, and the test he keeps returning to is the colour of money. Ten thousand dollars from two project customers is a services business whose multiplier is people and hours; five hundred dollars a month from twenty customers is a product that scales without hiring, and that pattern is itself the moat, though he concedes every high wall eventually meets a taller ladder, so the real question is whether you can build, monetise and move on first. From there the cheques go where India is thin: five rocket teams met in six to nine months, an indigenised camera shooting 7 km from a rooftop, a mid-mile logistics drone priced against what a man on a bike would charge for the same fifty kilometres, a vehicle control unit he calls the CPU of an EV, and chip design as the capability India should spotlight instead of fabs. Fund I reported returns of 45 to 48 per cent over four years and its state-government LP came back for Fund II. That is his evidence that patient capital for hard technology now exists here. The stakes: if the next wave of models and silicon is only consumed in India, the value accrues somewhere else.