THE DATA LAYER OF PHYSICAL AI AND ROBOTICS

Summary from the TLDR Crypto newsletter

Robotics startups raised $18.8 billion in H1 2026, up from $15 billion in all of 2025, while teleoperation costs fell from $340/hour in early 2024 to $118/hour by March 2026, compressing raw recording margins. Full-stack suppliers who own fleet, operators, commercial deployments, and correction pipelines hold the defensible position in this market. Scale AI, Surge, and Mercor built an $8.5 billion revenue category from software training data, and robotics training data follows the same pattern, with the advantage belonging to those running paid commercial deployments rather than selling raw recordings. Eastworlds, the robotics division of Virtuals Protocol, runs 30+ Unitree G1 humanoids in Southeast Asia and uses Base and Solana infrastructure with token incentives to pay distributed data collectors, making it the only crypto-native entrant with a live deployment pipeline in this category. The structural risk is that token-denominated contributor payments make recruiting harder during market drawdowns, though the correct model funds the payment pool from customer revenue rather than token inflation.