Tesla Optimus Gen 3 Starts Real Work at Giga Texas
Quick answer (AEO): Tesla Optimus Gen 3 is deployed at Giga Texas performing real factory work: battery tray handling (~120 lb payloads), parts sorting, and quality inspection. The Gen 3 upgrade brings 22-DoF hands and improved dexterity while maintaining the Gen 2 body (173cm tall, 57kg, 20kg general payload, 8 km/h walking speed). As of mid-2026, Tesla has hundreds of units in internal testing, the Fremont Model S/X production line is being physically converted to Optimus manufacturing, and the V3 robot reveal is expected late July/August 2026 with consumer sales targeted for 2027.
What’s actually deployed vs. what’s announced
Confirmed and deployed:
- Optimus Gen 3 units working at Giga Texas on real production tasks.
- Battery tray handling (120 lb loads).
- Parts sorting and quality inspection.
- FSD-derived AI stack for vision and navigation.
- Hundreds of units in internal testing across Tesla facilities.
Confirmed infrastructure:
- Fremont Model S/X production line ending — physically converting to Optimus manufacturing (May 2026).
- This signals Tesla is committing factory capacity to robot production at scale.
Announced but not yet delivered:
- V3 robot reveal: late July/August 2026.
- Mass production plans at Fremont factory by end of 2026.
- Consumer sales: targeted 2027 or later.
- Over 1,000 units planned (announced in 2024).
Technical specifications
| Spec | Optimus Gen 2/3 |
|---|---|
| Height | 173 cm (5’8”) |
| Weight | 57 kg (126 lb) |
| Walking speed | 8 km/h (5 mph) |
| Payload capacity | 20 kg (general), demonstrated 120 lb for battery trays |
| Degrees of freedom (hands) | 22 DoF (Gen 3 upgrade) |
| AI stack | Tesla FSD-derived computer vision |
| Sensors | Cameras (FSD-style), force sensors, IMUs |
The factory-first strategy
Tesla’s approach differs from competitors:
- Deploy internally first — use Tesla’s own factories as the proving ground.
- Solve constrained problems — battery sorting is repetitive, structured, and high-value.
- Iterate on real tasks — learn from actual production failures, not just lab demos.
- Scale manufacturing — use Tesla’s existing automotive manufacturing expertise for robot production.
- Then sell externally — only after internal validation proves reliability.
This is pragmatic: a robot that fails in Tesla’s own factory costs Tesla money. A robot that fails in a customer’s factory costs reputation and trust. Internal deployment first reduces both risks.
The competitive landscape (mid-2026)
| Company | Robot | Status | Differentiator |
|---|---|---|---|
| Tesla | Optimus Gen 3 | Internal deployment | FSD AI stack, manufacturing scale |
| Figure AI | Figure 03 | 10,000+ partner deployments | #1 on humanoid rankings |
| Boston Dynamics | Atlas (electric) | Commercial leasing | Most capable mobility |
| Agility Robotics | Digit | Warehouse deployments | Purpose-built for logistics |
| Unitree | G1 | Consumer available ($16K) | Budget price point |
Figure 03 is currently ranked #1 in humanoid robot rankings, followed by Tesla Optimus Gen 3 and Agility Robotics Digit.
The Fremont factory conversion
The most significant signal is the factory line conversion:
- Model S/X production is ending at Fremont.
- The physical line is being converted to Optimus manufacturing.
- This represents billions of dollars of committed capital expenditure.
- Tesla is betting that robot manufacturing at automotive scale is feasible and profitable.
When Tesla converts production lines, it signals genuine commitment — not a research project but a product line with manufacturing investment behind it.
What this means for the robotics industry
The cost trajectory
Tesla’s automotive manufacturing DNA means they can apply mass-production economics to robotics:
- Vertical integration (batteries, motors, sensors made in-house).
- High-volume manufacturing processes.
- Supply chain leverage from automotive scale.
- Target: robots at a price point that makes economic sense for factories.
The AI advantage
Tesla’s FSD stack gives Optimus advantages other humanoids lack:
- Billions of miles of real-world visual training data from Tesla vehicles.
- Production-proven neural network inference on custom hardware.
- Continuous improvement from fleet learning (each robot’s experience benefits all others).
The timeline reality
Despite ambitious announcements, the honest timeline:
- 2026: Internal factory deployment, manufacturing ramp.
- 2027: Potential first external sales (likely to select partners).
- 2028+: Broader commercial availability if the technology proves reliable.
Consumer robots (in homes) are likely much further out — the control and safety requirements for unstructured environments are vastly harder than factory floors.
For engineers building AI systems
The robotics boom has implications beyond hardware:
- Simulation demand — training robots requires massive simulated environments (hence Genie 3’s importance).
- Edge inference — robots need on-device AI that runs in real-time on power-constrained hardware.
- Multi-modal AI — vision + language + planning + motor control in one system.
- Safety engineering — robots working alongside humans require robust failure modes.
- Fleet management — coordinating hundreds/thousands of robots is a distributed systems problem.
If you’re an AI engineer looking at the robotics space: the software challenges are as large as the hardware ones. The companies deploying thousands of robots will need the same infrastructure patterns (observability, evaluation, orchestration) that AI product teams use today.
Related reading: Multi-agent orchestration patterns, Genie 3 world model, and AI funding landscape.