Nomura Securities' robotics research report points out that the continued decline in the price of physical AI first-person perspective video (Ego) data is squeezing the profit margins of lower-level data collectors. The value of the industry chain is gradually concentrating on Tier 1 suppliers directly connected to model labs, data verification, and automated annotation. Nomura's industry research shows that some Tier 2 data collection operators have been nearing losses since mid-2026.
According to Nomura's estimates, in the Southeast Asian low-cost raw Ego data supply chain, Tier 1 suppliers can retain approximately $5 per hour, while actual collection workers receive approximately $1-2, resulting in a profit margin of approximately 10%-20% for Tier 1 suppliers.
However, this profit margin also depends on whether the data can be resold, and most datasets can only be sold once.
Automated annotation is further changing the industry's cost structure. Nomura's research found that one Tier 1 supplier, after switching to machine annotation in August 2026, reduced its annotation team from over 1,000 people to approximately 50. Nomura believes that in the future, the competitive advantage of physical AI data suppliers will come more from leading lab customer channels, data processing efficiency, and quality verification capabilities. Business models relying solely on manual data collection for scale expansion will continue to face pressure.
