Kinara built two AI chips for under $50M, 90% out of Hyderabad, and sold to NXP for $307M all cash — one of India's largest, least-talked-about deep tech exits. Vikram Vaidyanathan sits down with Ravi Annavajhala to unpack how it happened: the disruption playbook behind it, training vs. inference explained plainly, why edge beats cloud (cost, latency, privacy, reliability), how to shrink a frontier model to fit on a chip, why India's real AI edge is industrial data (not talent or capital), and the difference between frugal and cheap.








.webp)
.avif)
%20(1).avif)


