The humanoid race is now a production race — and production has two speeds

A thousand Optimus units work a gigafactory while Fremont's line starts 'extremely slow.' Unitree targets twenty thousand shipments at a tenth of the price. The question stopped being whether humanoids work. It's who can make them fast enough to matter.

Over a thousand Optimus robots now sort parts and run inspections inside Giga Texas as Tesla's dedicated Fremont line spins up. The demo era ended without a press release. Robots doing real shifts in real factories is the new baseline across the industry — Figure bills BMW by the robot-operating-hour, Agility runs warehouse deployments, Atlas is in the field. Capability arguments are over; throughput arguments have begun.

Two production philosophies

The race has split into speeds. Unitree — already the world's volume leader at roughly 5,500 units in 2025 — targets 10,000 to 20,000 shipments this year at a tenth of Western prices. That is the consumer-electronics playbook: ship cheap, iterate on the installed base, ride the cost curve. The Western tier answers with capability-per-unit and verified productive work. Drones and EVs both ran this exact bracket, and the volume tier won both times.

The flywheel argument

Tesla's wager is that neither price nor polish decides it — data does. A thousand robots working your own factory generate the manipulation data that trains the next generation, and a robot-built production line closes the loop entirely. If embodied capability scales with fleet-hours the way language capability scaled with tokens, the largest working fleet wins regardless of who ships the most units to customers.

Which is why the only number worth tracking this year is not reveals or specs but robots-at-work, by maker, by month. The humanoid industry has become a manufacturing industry. Manufacturing industries are won on curves, not keynotes.

Robot News Today — Tesla Optimus and the automation revolution → · EVST — Top 8 humanoid robot companies to watch in 2026 →