Ropedia Raises $30 Million to Build the Data Layer for Physical AI

Ropedia Raises \$30 Million to Build the Data Layer for Physical AI

Singapore-based Ropedia has raised \$30 million in pre-Series A funding to scale a bet that the next great leap in robotics will be decided not by better motors or smarter models, but by data — specifically, the messy, multimodal record of how humans actually move through the world.

The company disclosed the milestone on July 23, 2026, saying the total came across two rounds: a \$22 million raise announced this week and an earlier \$8 million round in March. The capital will fund an expansion of Ropedia's real-world data-collection operations across Southeast Asia and North America, ramp production of its wearable capture device, and grow its research into data and world models. Investors were not individually named, though Ropedia said the latest round was led by venture backers with experience in AI, deep technology and infrastructure, alongside long-term financial investors and strategic partners in robotics and mobility. An earlier raise, the company added, included super angels connected to Google, Andreessen Horowitz, NVIDIA and Amazon.

The bottleneck behind embodied AI

The pitch rests on an analogy the industry has repeated to the point of cliche, but which remains stubbornly unsolved: internet-scale text taught large language models to reason and write, so what corpus will teach robots to act? Pictures of objects and unstructured online video do not translate cleanly into robotic trajectories. A machine reaching for a cup needs synchronized, annotated information about movement, geometry, timing and the physical consequences of its actions — data that barely exists at scale and is expensive to produce through robot teleoperation.

That gap is Ropedia's market. "There's not yet going to be a massive deployment of robots," founder and Chief Executive Zhaoxi Chen told SiliconANGLE. "We need to unlock the ChatGPT moment for robotics first."

Ropedia's answer is a head-mounted wearable called HOMIE, short for Human-centric Omni Interaction and Experience. Shaped roughly like a crown, weighing under half a pound and fitted with four cameras pointing in the cardinal directions, it records first-person video, audio, depth, gaze, hand movement, body motion and camera pose simultaneously — every stream timestamped so it can be reconstructed into clean, model-ready datasets. Because it sits out of the wearer's line of sight, people can perform ordinary work while it captures. "This kind of 'anywhere, by anybody' is really valuable for enriching the diversity of the data," Chen said.

A dataset measured in millions of episodes

The output feeds what Ropedia calls Xperience-10M, which it describes as one of the largest human-experience datasets in the industry: more than 10 million interaction episodes and over 10,000 hours of multimodal recordings, spanning billions of synchronized video, depth, motion-capture and inertial-sensor frames. Chen frames the company less as a hardware maker than as the invisible plumbing of physical AI, comparable to cloud infrastructure for internet services. "Infrastructure means that you can push the boundary of data production from, let's say, 1,000 hours to 1 million hours of data production," he said.

The model appears to be finding customers. Ropedia says it already works with more than 20 robotics and foundation-model companies across North America, China and Singapore, and that its pipeline cuts data-collection costs by up to 50 times versus traditional methods. As a proof point, Chen said Ropedia supplied more than half of the data used to train MolmoMotion, a 3D motion-forecasting model released by the Allen Institute for AI. The company plans to unveil a lighter second-generation HOMIE next month, with an eventual production target of up to 10,000 devices, and to build out a larger U.S. presence to sit closer to North American robotics labs.

Why it matters

The timing is not accidental. Humanoid and general-purpose robots have absorbed enormous investment on the promise that they will soon move from demo floors to factories and homes, but the field's central constraint has quietly shifted from hardware to training data. Whoever can supply diverse, high-quality real-world interaction data at scale occupies a position analogous to the labeled-image and web-text providers that underwrote the last decade of AI. Ropedia is one of a growing cohort — alongside efforts to harvest egocentric video and scale teleoperation — racing to own that layer. Its wager is that curation, not just collection, is the hard part. "Data is also a form of science," Chen said. "It's not only about collecting or accumulating the data. It's more like how you organize and understand the flow between different domains."

What to watch: whether the second-generation wearable and the push toward 10,000 devices translate into a genuinely defensible dataset, whether Ropedia can convert its 20-plus early customers into recurring licensing revenue, and whether the "anywhere, by anybody" collection model runs into the privacy and consent questions that always trail always-on cameras. If real human experience proves to be for physical AI what the internet was for language models, the companies that own that data early stand to matter far beyond their current size. A \$30 million pre-A is a modest sum in today's AI market — but Ropedia is betting it is buying a foundational position.