Canadian Startup Mecka AI Acquires Vancouver's Docula to Build the Data Infrastructure That Robots Need to Learn
Toronto and New York-based physical AI startup Mecka AI quietly acquired Vancouver's Docula earlier in 2026, adding the three-person bootstrapped team to help it process petabytes of human motion data used to train robots. The deal, disclosed publicly on June 29, reflects a fundamental constraint in physical AI development that is receiving far less attention than the robots themselves: the data infrastructure problem.
Mecka AI bought Vancouver's Docula early this year. The companies didn't disclose deal terms, but all three members of the bootstrapped Docula team joined Mecka. Mecka aims to collect physical data from human movement and sell that to companies making robots. Docula built an AI data-processing engine for medical billing and auditing that could ingest records, normalize codes, run edits, benchmark fees, and spit out defensible reports instantaneously. Though Docula's product was in healthcare and not directly related to robotics, the core capability - dealing with massive volumes of data at speed - is exactly what Mecka needed.
What Mecka Is Building and Why the Data Problem Is Hard
Mecka's core business model is both simple to describe and technically demanding to execute. The company sends iPhones and custom cameras to hundreds of thousands of "contributors" across 12 countries, collecting physical, three-dimensional data about how people move around and interact with objects with their hands. That motion data - walking, grasping, manipulating, sorting - is then sold to companies training humanoid robots and robotic arms.
The problem Docula solves is not collection. It is processing. Motion data from hundreds of thousands of contributors across 12 countries arrives in varied formats, with varied quality, captured on varied hardware. Before any of it is useful for robot training, it needs to be ingested, normalized, validated, and formatted in ways that AI training pipelines can consume. That is an enormous data engineering challenge that requires exactly the kind of high-throughput processing capability Docula built for medical billing.
Mecka special projects lead Mark Grinev explained the logic directly: "They are dealing with massive volumes of data. They were dealing with petabytes of complex structured data, and we need to do the same thing."
Physical AI's Data Problem Is the Next Bottleneck
The robotics industry is facing the same fundamental challenge that language AI faced five years ago: the training data required to make capable systems does not naturally exist in a form that machine learning can use. For language AI, the solution was scraping the internet. For physical AI, there is no internet of human movement data to scrape. The data has to be collected, which requires building the collection and processing infrastructure from scratch.
The US robotics market was worth $11.4 billion in 2026, a nearly 30% boost year over year. Canada lags behind peer countries on industrial robotics adoption. Mecka opened a New York City office this month as it looks to expand its US customer base, though more than half of its 45-person team is Canadian. Most of the startup's customers are US-based and exceptional growth has been seen in the past eight to 12 months.
Mecka's $60 million raise announced earlier in 2026 provides the capital to scale this data flywheel. The more contributors provide motion data, the more training data Mecka can sell. The more training data available, the more capable the robots trained on it become. And the more capable robots become, the more demand there is for the next round of training data.
What This Means for Businesses Thinking About Robotics
For business leaders in logistics, manufacturing, and warehousing evaluating AI agents and automation, Mecka's Docula acquisition is a useful indicator of where the physical AI bottleneck currently sits. The capability of humanoid robots is advancing rapidly - Tether's $1.4 billion investment in Neura Robotics this month put 5 million robots by 2030 as a funded production target. The constraint is not the robot hardware or the AI models. It is the training data quality and quantity.
Companies like Mecka building the data infrastructure layer are doing for physical AI what AWS did for software - creating the underlying infrastructure that makes the end products possible. In a market projected to exceed $11 billion in the US alone, the data layer is likely to be as valuable as the robots it trains.
Cut Through the Noise
What did Mecka AI acquire and why?
Toronto- and New York-based physical AI startup Mecka AI acquired Vancouver's Docula, a three-person bootstrapped AI data-processing company, earlier in 2026. All three Docula team members joined Mecka. Docula's core capability was processing massive volumes of complex structured data at high throughput - originally built for medical billing - which Mecka needs to process petabytes of human motion data collected from hundreds of thousands of contributors across 12 countries for robot training.
What does Mecka AI do and how does it make money?
Mecka collects three-dimensional physical data about human movement and object manipulation by distributing iPhones and custom cameras to contributors in 12 countries. This motion data - capturing how people walk, grasp, sort, and interact with physical objects - is sold to companies training humanoid robots and robotic arms. The data layer Mecka builds is the missing ingredient that robot companies need to create capable AI models for physical tasks.
Why is physical AI data collection a hard problem?
Unlike language AI, which could train on existing internet text, physical AI requires purpose-built motion data that does not exist in a usable format. Creating it requires building collection hardware infrastructure, contributor networks across diverse environments, and processing pipelines that can handle petabytes of varied-format motion data. Docula's high-throughput data processing capability directly addresses the processing side of that challenge.
How big is the robotics market Mecka is targeting?
The US robotics market reached $11.4 billion in 2026, up nearly 30% year over year. Mecka opened a New York City office to expand its US customer base while maintaining a majority-Canadian 45-person team. The company raised $60 million earlier in 2026 and has seen exceptional growth in the past eight to 12 months as demand for robot training data increases alongside humanoid robot deployments.
