This startup thinks robotics is about to have its ChatGPT moment
THE SO WHAT
Training physical AI on millions of hours of game data is a bet that sim-to-real plus foundation models can replace bespoke robotics stacks. If it works, the constraint on deploying smarter robots moves from data collection to integration and safety sign-off.
READ THE SOURCE
MORE FROM THE WIRE
Robotics & AutonomyAutonomous drone delivery startup Manna plots major US expansion
A 1,000-person operations and manufacturing footprint in Tulsa means drone delivery is graduating from pilot to regional infrastructure. Retail, QSR, and pharmacy operators in Manna’s target zones should start modeling last-mile mixes where air is a real channel, not a novelty.
Robotics & AutonomyChinese autonomous driving company Momenta saw a muted trading debut in its Hong Kong IPO, valuing the company at ~$9B, after raising $751M
A $9B valuation with a muted debut for a GM-backed AV player suggests public markets are still discounting long commercialization timelines in autonomy. If you’re building in AV or physical AI, assume capital will demand clearer near-term revenue paths than private rounds over the last five years did.
Robotics & AutonomyENCY Software and Stäubli Robotics partner to simplify robot programming
Lowering the programming barrier for industrial robots shifts the constraint from specialist engineers to process design and change management. If you run factories or warehouses, you should be re-auditing tasks for automation viability assuming cheaper, faster programming cycles.
Robotics & AutonomyMistral AI Releases Robotics Model to Support Physical AI Push
Mistral’s new robotics navigation model plus European industrial deals is another data point that “physical AI” is becoming a distinct product line, not a demo. If you run factories or logistics in Europe, start mapping where navigation and manipulation models could plug into existing automation vendors.