Embodied AI Raises ~$100M Seed Round: Don't Applaud Just Yet

Just saw the news: Liu Nianqiu, technical partner at DeepRoute.ai, left three months ago, and his new company, Femo One Tech, has completed a seed round of nearly $100 million, led by IDG Capital, with Didi and Hesai following. Less than half a year since establishment, the first product hasn't been made public yet, but the money has arrived.

My first reaction was still that phrase: this kind of demo is far from mass production. This isn't directed at Liu Nianqiu personally; he has done autonomous driving and hardware-to-mass-production delivery, and the team's resume is indeed solid. But a seed round buys direction; results depend on what comes after. The core metrics weren't given in the news.

Looking at embodied intelligence over the past two months, what I fear most are pretty videos. Grasping, walking, delivering—single-point success is easy, continuous stability is hard. Key factors are task success rate, mean time between failures, cost per unit, maintenance responsibility, and whether customers are willing to pay monthly. The reshuffling logic in autonomous driving seems the same here; those who survive are those who can deliver stably.

Didi and Hesai coming in indicates that scenarios and sensor supply chains are betting on last-mile operations. But betting doesn't equal orders. If startups want to copy this path, I'd suggest finding a narrow scenario first, nailing yield and cost, and then talking about generalization.

So this funding is worth watching, but I wouldn't recommend treating it as an industry turning point right now. Wait until the first batch of machines runs continuously for three months in real workstations, then we'll talk.