Singapore-based Ropedia raised $30 million in pre-Series A funding to scale its data infrastructure...
The AMW Read
Incremental update to a new entrant in the robotics data infrastructure space; significance is sub-segment as the $30M round does not reshape the segment.
Singapore-based Ropedia raised $30 million in pre-Series A funding to scale its data infrastructure for physical AI, comprising an $8 million tranche announced in March and a newly disclosed $22 million close. The startup's Homie wearable captures synchronized multimodal data — video, audio, depth, hand tracking, gaze, body motion — to feed its Xperience-10M dataset, which contains over 10 million interaction episodes and 10,000 hours of real-world recordings. Ropedia sells access through dataset licensing, hardware access, and research collaborations, targeting robotics and embodied AI labs that need task-specific interaction data rather than static video or text corpora.
This raise sits squarely within the ongoing infrastructure buildup for physical AI, where data pipelines are emerging as a distinct layer, analogous to how data centers supported cloud computing and web-scale text corpora supported LLM training. Ropedia’s approach of using a wearable to capture first-person human activity bypasses the expensive teleoperation methods that rely on robot fleets, claiming cost reductions of up to 50x compared to traditional techniques. The company’s model mirrors the acqui-licensing pattern seen in foundation models, but here the “product” is structured real-world interaction data rather than model weights, suggesting a new sub-category in the data infrastructure segment.
The company’s positioning as a data infrastructure provider for embodied AI addresses a structural bottleneck: scaling physical AI requires orders of magnitude more training data than recent language model releases, but that data must be multimodal, synchronized, and task-specific. Ropedia’s geographic spread across Southeast Asia and North America, alongside its academic ties to Nanyang Technological University, gives it a diverse capture environment. The $30 million pre-A round is modest by hyperscaler standards, but it signals early investor conviction that data generation, not just data labeling, will be the differentiator for robotics and physical AI in the next cycle.