YY Group has opened a robotics training laboratory in Singapore and begun deploying humanoid robots in pilot programs across the city-state, marking the company's most concrete push yet into the robotics sector after years of positioning itself primarily as an AI training data vendor. The facility will focus on capturing and annotating data from robot operations in real-world environments, which the company plans to package and sell to robotics developers while simultaneously operating its own fleet of humanoids for commercial clients. That dual mandate—selling picks and shovels while also mining for gold—represents an uncommon strategic bet in an industry where most players choose one lane or the other.

The training lab's physical infrastructure includes motion capture systems, sensor arrays, and controlled environments designed to simulate retail, logistics, and hospitality settings. YY Group plans to use the facility to record thousands of hours of human demonstrations performing tasks that humanoid robots might eventually handle autonomously, from stocking shelves to handling customer interactions. The data will feed into foundation models being developed by robotics companies that lack the resources to generate massive proprietary datasets on their own. Industry sources estimate that high-quality robotics training data currently trades at rates between $50 and $200 per labeled hour depending on task complexity and annotation depth, though YY Group has not disclosed its pricing structure. The Singapore location offers regulatory advantages—the city-state has established streamlined approval processes for robotics pilots in commercial spaces—and proximity to manufacturing hubs in Southeast Asia where several Chinese robotics startups have begun establishing regional offices.

The humanoid pilots currently underway involve robots from unnamed manufacturers operating in what the company describes as retail and food service environments. YY Group has not specified which humanoid platforms it is deploying, how many units are in operation, or which commercial partners are hosting the pilots. That opacity is notable given the company's public listing on NASDAQ under ticker YYGH, which typically comes with disclosure expectations around material partnerships and capital expenditures. The workforce-as-a-service model the company is pursuing faces steep unit economics hurdles—humanoid robots currently cost between $150,000 and $250,000 per unit for research-grade platforms, and even optimistic projections put commercially viable humanoids at $50,000 to $75,000 within three years. For the model to pencil out, each robot would need to generate revenue comparable to two or three human workers while maintaining uptime above 90 percent, a target no commercial humanoid deployment has yet demonstrated at scale.

The announcement arrives as the training data market for robotics enters a period of rapid consolidation and standardization. Scale AI, which has raised over $600 million, acquired Sensor Data Management Systems in late 2023 to bolster its robotics data capabilities. Figure AI has built its own data collection infrastructure rather than relying on third-party vendors, a sign that leading robotics companies view proprietary datasets as a competitive moat. Meanwhile, open-source initiatives like the Open X-Embodiment dataset—backed by Google DeepMind and involving contributions from 22 research institutions—are creating publicly available alternatives that could commoditize certain categories of training data. YY Group's vertical integration strategy suggests a hedge against that commoditization risk, but it also introduces potential conflicts: commercial clients deploying YY's humanoids may hesitate to share operational data knowing the company also sells data to their competitors. How YY navigates those information barriers will largely determine whether the dual strategy proves visionary or simply spreads capital too thin across incompatible business models.

What to Watch: YY Group's next quarterly earnings call should reveal capital expenditure figures related to the Singapore facility and humanoid fleet, offering the first concrete data on how heavily the company is committing to the robotics vertical versus its legacy data annotation business. Watch for announcements identifying which humanoid platforms YY has selected—partnerships with manufacturers like Unitree, Fourier Intelligence, or UBTECH would signal focus on cost-optimized hardware, while choosing Figure or 1X would indicate a bet on premium capabilities. Finally, track whether any major U.S. or European robotics developers publicly announce data partnerships with YY Group, as such deals would validate the market for third-party robotics training data despite the trend toward vertical integration.