Robot-brain startup Ruoyu raises $14.9M Pre-A, led by Sunway
What's the deal? Shenzhen Ruoyu Technology, a robot-brain company, has raised a Pre-A round of US$13.8M, led by Sunway Communication with participation from Jinmu CapitalDealroom has a profile for this one. Try Dealroom →, Qindao Capital, Huanghai Capital, and Zhongchuang CapitalDealroom has a profile for this one. Try Dealroom →. Founded in 2023, the startup has now raised several hundred million yuan in total.
What's the endgame? Ruoyu builds "Jiutian," a self-developed robot brain structured around a three-part system of brain, body, and scene data. It treats the physical robot as a tool for collecting real-world data rather than the end product, aiming to become an intelligence layer that moves across hardware platforms, robot bodies, and scenes.
Why now? The round lands as the company completes the first half of its strategy: entering a high-value scenario and proving out both its technical loop and real-world validation. The money will fund full-stack brain development, real-scene data loops, and commercialisation.
Why fuelling robots? Ruoyu chose smart refuelling — fuelling robots — as its first deployment, a setting that combines explosion risk, continuous outdoor work, and extreme temperatures. The logic, chief executive officer Sun Teng said, is that harder scenes surface more real problems: "We can use a higher density of real problems to force the system to expose bottlenecks and accelerate evolution."
Explosion-proof certification, Sun said, reshapes the brain's architecture itself, pushing it from centralised high-compute designs toward a layered "lightweight body safety controller plus edge computing" structure that trims onboard compute, sensors, and bandwidth.
What could go wrong? The approach is slow. High-value scenes take longer to accumulate data, and copying them carries high marginal cost, so Ruoyu cannot claim the "one model fits many scenes" narrative of rivals chasing general-purpose capability. Its bet is that once the loop works, followers struggle to catch up, because real-scene data must be "soaked" out of field time rather than stacked from compute.
What's proven so far? Ruoyu began trial operations at a Foshan site in April and later disclosed partial metrics: for fixed vehicle models, refuelling tasks reach a high success rate in open environments, with human takeover triggered only on anomalies. The skills built there — complex manipulation, perception, task planning, and decision-making — can transfer to charging, ports, and inspection work, the company says.
The signal: The round reflects embodied AI shifting from technical storytelling to value delivery, and a widening split over data strategy between simulation and real-world collection. As the sector moves from "does it exist" to "is it usable," teams willing to send robots into the most dangerous sites — building moats from field data — may prove the scarcest force as the industry matures.
Read more: 36kr.com
Image credit: yumikrum
