DexForce: training humanoids on virtual data
-
2026-10-15 (first record): Chinese startup DexForce (Shenzhen) trains its humanoid robots on virtual data — simulation-based training in place of real-world data collection — which it says shrinks development costs and accelerates practical use (Reuters via MSN syndication, ~4:45 AM ET; Traders Union syndication of the same report, ~4:45 AM ET). New company in the record; the sim-training lane sits beside the data-layer ticks on file (Config Intelligence's 150,000-hour CFG-1, October 14) without overlapping them. Captured coverage carries no model name, no cost figure against a baseline, and no named customer or hardware target. Watch: named model, quantified savings vs real-data pipelines, whose hardware it trains, funding layers. https://www.msn.com/en-ca/news/other/chinese-startup-dexforce-trains-humanoid-robots-using-virtual-data-to-slash-costs/ar-AA2egQns?ocid=BingNewsVerp
-
2026-10-15 (~4:47 PM ET, note.com): the thread's first backstory layer — a Shenzhen headquarters-visit profile recounts DexForce starting with 3D cameras for factories before the yellow humanoid, from a first contact at WRC. Blog-grade color, not a financial or deployment fact; the ~8:51 AM ET $1.4 billion valuation remains the financial layer. Watch unchanged: investors behind the valuation, named model or product.
-
2026-10-15 (~8:51 AM ET, domain-b): the tick's first development hours later — the company is reported to have reached a $1.4 billion valuation as it develops synthetic data, simulation, and robot training (DexForce reaches $1.4 billion valuation with synthetic robot training). Hedged by the outlet ("reportedly"), single-source syndication-grade, no investor names or round type. The thread's financial layer opens beside the cost-savings claim from the debut Reuters capture (same day, ~4:45 AM ET). Watch: company confirmation, investors, named model. https://www.domain-b.com/technology/artificial-intelligence/dexforce-synthetic-data-humanoid-robot-training-2026