Kinetic Blocks, an Oslo startup and NVIDIA Inception member, opened its long-awaited marketplace for robotics training data on September 1, 2026, bringing standardised verification and quality grading to a corner of the physical-AI supply chain where bilateral deals have historically taken months and buyers routinely discovered they had paid for mislabelled or dead-time footage.
Trust built into every listing
Every dataset on the platform carries two independent results before anyone pays. A verification pass checks each factual claim on the seller's dataset card against the actual files and stamps every claim as Verified, Doesn't match or Can't confirm, always with the evidence attached. A separate KB Quality grade then scores footage from A+ to F across four pillars: technical quality (30 percent), temporal integrity (25 percent), content usefulness (30 percent) and completeness (15 percent). Weights are versioned, so a grade means the same thing across every seller and every listing.
Household to construction, on real robots
Launch inventory spans household manipulation, cleaning POV footage, industrial, hospitality, construction and personal-care environments, sourced through the company's supplier network. A verified 20-hour egocentric household manipulation dataset from South America is listed at $800; a 10-hour cleaning POV dataset from Oceania sits at $300. Sellers keep full ownership, pick the price, licence and exclusivity, and their files stay read-only inside the platform.
Feeding the physical-AI stack
The launch lands as embodied-AI and humanoid teams from Neura Robotics to XPeng race to feed foundation models the diverse teleoperation and egocentric footage they need to generalise across tasks and environments. Kinetic Blocks is betting an open market with visible quality signals will outrun bilateral deals as demand scales.
Reporting based on coverage from Kinetic Blocks, Humanoid Guide and industry analyses.