coScene, a Shanghai-based data infrastructure company formally known as Shanghai KeXing ShiKong Technology Co., raised nearly 100 million yuan (about $14 million) in a Pre-Series A round led by Yonggui Fund, with Tianji Capital, Changzhou Zhirong and several state-backed and industrial investors participating, the company said on November 9, 2026. Founder Huang Yangming says coScene is the first company in China to simultaneously cross 100 million yuan in real robot data orders, secure formal data-asset rights, and deliver that data under compliance review.
What does coScene actually sell?
coScene builds what it calls a collection-training-simulation-inference data pipeline for robotics and autonomous-driving companies, covering pose tracking and multi-dimensional annotation of real-world robot deployments. Its two core products are StarBase, a data-production management tool, and the coScene platform itself, a spatiotemporal multimodal data manager, according to PEdaily. The company's data has reportedly been used to train cleaning robots, robotic lawn mowers, low- and mid-speed autonomous vehicles, and humanoid embodied-intelligence robots, including real-world operating data captured from Unitree hardware.
Who is running it, and what's the traction so far?
Founder Huang Yangming built his career in data infrastructure before starting coScene in May 2022, including work on Apple Watch health-data architecture and a full data closed-loop system for autonomous-vehicle company Chiedu. His founding team draws from Apple, Alibaba, Google and Chiedu. coScene says it has delivered more than 300,000 hours of multimodal robot data on a single overseas contract worth over 200 million yuan, with delivery-quality rates near 100% — figures the company cites as evidence that selling robot data, not just robots, is now a standalone business.
How does this round fit coScene's funding history?
The Pre-Series A follows a smaller angel round from Linear Capital in August 2022, meaning coScene spent roughly four years building its data pipeline before raising a meaningful institutional round. The jump from an angel check to a nearly 100-million-yuan Pre-A, led by an institutional fund rather than an individual backer, suggests investors now treat labeled, rights-cleared robot data as its own infrastructure layer rather than a function robotics companies build in-house.
Why does a data company matter in China's robotics race?
China's humanoid and embodied-AI makers have been pouring capital into training models rather than hardware, a shift that raises the value of clean, rights-cleared real-world data to feed those models. coScene's pitch — that it has already resolved the thorniest parts of that problem, namely data-asset ownership and compliant delivery — positions it as a vendor for robotics companies that don't want to build their own data operations from scratch. Competing approaches include robot makers like Unitree collecting and using their own fleet data internally, and large-model labs building synthetic-simulation pipelines instead of paying for real-world capture. With a fresh round and an overseas contract already running past 200 million yuan, the question for coScene is whether it can keep winning orders from multiple robot makers rather than becoming an extension of any single one.