Three humanoid doctrines update — Tesla hits 1,000+ units, Figure ships 17-hour endurance, 1X holds the consumer flag
Tesla Optimus Gen 3 crosses 1,000 deployed units. Figure 03 streams a 17-hour, 22,000-package warehouse run. 1X continues consumer deliveries. The three humanoid doctrines we mapped earlier today now have new data points — and the gap between them is widening.
The three doctrines, updated
This morning we mapped three doctrines: Apptronik picks factories, Figure picks the controlled-to-home gradient, 1X picks consumer-first. The PM news cycle adds Tesla as the fourth axis and updates Figure with concrete endurance data.
Tesla's 1,000-unit milestone
Tesla now has 1,000+ Optimus Gen 3 units across its global manufacturing facilities. V3 reveal is targeted for late July, with Fremont production lines being installed. A second factory at Giga Texas is under construction. Musk has named a 10M unit/year target — aggressive enough that public-market analysts treat it as marketing, but the capital commitment behind it is real.
The 1,000-unit data point is what changes the competitive picture. None of the other humanoid programs have units in the hundreds, let alone thousands. Tesla's in-factory data-flywheel advantage compounds with every unit deployed.
Figure's 17-hour endurance run
Figure 03 livestreamed a 17-hour continuous warehouse-style run handling 22,000+ packages on the Helix-2 stack. The endurance shape — sustained throughput rather than peak capability — is the metric Figure has been chasing for two release cycles. It's the first publicly-disclosed multi-hour autonomous-operation milestone outside controlled-factory tier.
The procurement implication is direct. Figure can now credibly bid into 24/7 warehouse and logistics deployments — the actual jobs the humanoid program needs to support unit economics at scale.
1X holds the consumer flag
1X's NEO consumer humanoid continues delivering at $20K outright or $499/month subscription. The sustained-delivery phase makes NEO the only humanoid with home-environment longitudinal data — every novel object, every unscheduled interaction adds to a training corpus none of the other programs can replicate without putting their own units in homes.
The four-way comparison
| Program | Doctrine | Best disclosed metric |
|---|---|---|
| Tesla Optimus | Factory-and-consumer simultaneously | 1,000+ units deployed, 10M/year target |
| Figure | Factory-to-home gradient | 17 hours continuous, 22,000+ packages |
| Apptronik | Factory-first, no consumer ambition | $935M raised, Mercedes + GXO deployments |
| 1X | Consumer-first, learn from field | Only home-environment longitudinal dataset |
What the next 18 months pick
The doctrines are diverging fast enough that the next 18 months will pick a winner — or two — by deployment density. Tesla wins if 10M/year scales. Figure wins if endurance-plus-home-gradient translates to logistics-scale unit economics. Apptronik wins if factory-first produces the cleanest balance sheet. 1X wins if home-environment data turns into a capability lead that the others can't catch up to.
Most likely outcome: two of the four win, two recede. The interesting question is whether Tesla's deployment density alone determines the winner regardless of doctrine, or whether one of the smaller programs catches up through methodology advantage. The interleaved vision-language paper is the kind of methodology input that could tilt the smaller-program side; whichever doctrine adopts it fastest gets the largest near-term improvement on out-of-distribution generalization.
Robot Report — Tesla Optimus 10M → · Humanoid Press → · Qviro — humanoid robots 2026 →