Tesla Optimus: Inside the Scale Play

Tesla is building factories to produce millions of humanoid robots per year. No other company comes close. The question isn't ambition — it's whether execution matches the vision.

---

No company matches Tesla's production ambition in humanoid robotics. Not Figure AI. Not Boston Dynamics. Not AgiBot. Tesla is building factories to produce robots at a scale the industry hasn't seen — millions of units per year.

The question isn't whether Tesla has the ambition. It's whether the ambition matches the execution.

Program Overview

| | | |---|---| | Robot | Optimus (Gen 2 current, Gen 3 hands, V3 patent filed Apr 2026) | | HQ | Palo Alto, CA / Austin, TX | | Leadership | Ashok Elluswamy (Head of Optimus, since Jun 2025) — replaced Milan Kovac | | Company type | Public (NASDAQ: TSLA) | | Funding | Internally funded from Tesla balance sheet |

Technical Specifications

Confirmed Specs

| Spec | Value | Source | |---|---|---| | Height | 173 cm (5'8") | Tesla AI Day 2022 | | Weight | 57 kg (Gen 2) | Tesla AI Day 2023 | | Payload | 20 kg carry | Tesla AI Day 2023 | | Hand DOF | 22 DOF per hand (Gen 3) | Tesla Nov 2024 | | Actuator type | Electric (planetary roller screws) | Tesla AI Day 2022 | | AI system | Tesla FSD adapted for robotics | Tesla AI Day 2022 | | Computing | Tesla AI5 chip | Q1 2026 Update Letter | | Target price | $20,000-$30,000 | Musk, WEF Jan 2026 | | Design type | Bipedal | — |

Unconfirmed / Undisclosed Specs

| Spec | Status | Note | |---|---|---| | Walking speed | Undisclosed | 2021 AI Day target was ~2.24 m/s (5 mph). No confirmed measurement published. | | Battery capacity | Undisclosed | Custom Tesla battery pack. Not publicly disclosed. | | Total body DOF | Undisclosed | Industry estimate 60-80+ for V3 (speculative). | | Battery life | Undisclosed | — | | Max speed | Undisclosed | — | | Computing specs | Partial (AI5) | AI5 chip tape-out completed Apr 2026. TFLOPS not disclosed. |

Spec confidence: Medium. Height, weight, payload, actuator type, and hand DOF are confirmed. Walking speed, battery capacity, and total body DOF remain undisclosed by Tesla.

The V3 Hand Patent (April 2026)

On April 16, 2026, USPTO published Tesla's V3 hand/arm patents — the most significant technical detail released about Optimus in months.

Key details from the patent:

• Tendon-driven architecture: Actuators in the forearm pull tendons that control finger movement — a biomimetic approach similar to 1X Technologies' NEO
• 4 DOF per finger: Each finger has 4 degrees of freedom
• 2 DOF wrist: Wrist adds 2 additional degrees of freedom
• Total: 22 DOF per hand (4×4 fingers + 2 thumb + 2 wrist = 22)

This is a meaningful design choice. Tendon-driven hands are more backdrivable (safer around humans) and can achieve higher DOF in a compact package, but they add mechanical complexity and potential failure points. 1X Technologies uses a similar approach for NEO.

The AI5 Chip

The AI5 chip tape-out completed in April 2026, per Tesla's Q1 2026 Update Letter. The chip is designed specifically to power Optimus units in the field.

Context: Tesla's AI hardware progression:

• HW3 (2019) — Original FSD computer
• HW4 (2023) — Current FSD computer in production vehicles
• AI5 (2026) — Next-generation, designed for both vehicles and Optimus

The training infrastructure is also significant: 230,000+ H100-equivalent GPUs across the Cortex 1 & 2 training clusters at Giga Texas.

The thesis: Tesla's FSD AI, trained on billions of miles of driving data, transfers to robotics because both domains require real-time perception, prediction, and planning in physical environments. The AI5 chip brings this capability to the robot itself (edge inference), not just the training cluster.

Production Roadmap

| Milestone | Status | Date | |---|---|---| | 1,000+ Optimus units deployed internally | Confirmed | 2025 | | Fremont pilot production line operational | Confirmed | Apr 2026 | | Fremont target: 1M units/year | Announced | — | | Giga Texas Optimus factory under construction | Confirmed | — | | Giga Texas target: 10M units/year | Announced | — | | External sales planned | Announced | End of 2027 |

The scale is staggering. AgiBot, the current production volume leader, has produced 10,000 units. Tesla's Fremont line targets 1M/year — 100x AgiBot's total. Giga Texas targets 10M/year — 1,000x. If even partially achieved, this would dwarf every other humanoid manufacturer combined.

The reality check: Tesla has a track record of announcing ambitious timelines and delivering late. The Fremont line is operational but output rate is undisclosed. The Giga Texas factory is under construction. The V3 reveal, tipped for Q1 2026, has not yet happened. External sales remain 18+ months away.

Leadership Transition

Milan Kovac resigned as head of Optimus in June 2025. Ashok Elluswamy, previously head of Tesla's Autopilot team, took over.

This was a significant change. Kovac was a visible and respected leader of the Optimus program. His departure raised questions about program direction. Elluswamy brings deep Autopilot/FSD expertise — reinforcing the thesis that Tesla's robotics strategy is fundamentally an AI strategy, not a hardware strategy.

Deployment and Milestones

| Date | Milestone | |---|---| | Aug 2021 | AI Day announcement | | Sep 2022 | Gen 1 prototype (Bumble-C) | | Dec 2023 | Gen 2 revealed — 30% faster, 10kg lighter, finger sensors | | Oct 2024 | We Robot event demo (criticized for teleoperation) | | Nov 2024 | Gen 3 hands with 22 DOF | | Jun 2025 | Milan Kovac resigns; Ashok Elluswamy takes over | | Dec 2025 | Optimus at Miami showroom event | | Mar 2026 | Appeared at AWE Shanghai | | Apr 2026 | Fremont pilot production line operational (1M units/yr target) | | Apr 16, 2026 | V3 hand/arm patents published | | Apr 2026 | AI5 chip tape-out completed | | Apr 20, 2026 | Optimus at Tesla Boston showroom for Boston Marathon |

The Teleoperation Question

The October 2024 "We Robot" event drew criticism when observers noted that Optimus demos appeared to involve teleoperation — remote human control — rather than full autonomy. Tesla did not explicitly confirm or deny the extent of teleoperation.

This matters because the gap between a teleoperated demo and autonomous operation is the gap between a remote-controlled puppet and a useful worker. Tesla's internal deployment of 1,000+ units suggests the robots are doing useful work — but the autonomy level of that work remains unclear.

Market Impact Assessment

What makes Tesla different:

1. Manufacturing infrastructure — No other humanoid company has Tesla's factory-building capability. They build gigafactories for cars; they can build them for robots.
2. AI pipeline — FSD is the most advanced real-world AI system ever deployed. Transferring it to robotics is a massive advantage if it works.
3. Balance sheet — Tesla can fund Optimus development from operations. No need for venture rounds or IPOs.
4. Vertical integration — Tesla designs its own chips (AI5), batteries, actuators, and software. This reduces supply chain dependency and could improve unit economics.

What constrains Tesla:

1. Opacity — Fewer disclosed specs than any other major humanoid company. Battery, speed, DOF, and autonomy levels are all unclear.
2. Timeline risk — External sales at end of 2027 means 18+ months of internal-only deployment. Competitors (Figure AI, Agility) are selling to external customers now.
3. Execution risk — The gap between Musk's announcements and actual delivery is well-documented.
4. Credibility — The teleoperation criticism and the broader "Musk discount" on timelines create skepticism among enterprise buyers.

Competitive Positioning

| Dimension | Tesla Optimus | Figure AI Figure 03 | Agility Digit | AgiBot A2 | |---|---|---|---|---| | Production scale | 1K+ (internal) | Pilot (BotQ 12K/yr) | ~100 shipped | 10K produced | | External sales | End 2027 | In progress | Active (RaaS + purchase) | In progress | | AI system | FSD adaptation | Helix 02 (proprietary VLA) | Arc (cloud platform) | GO-1 + Lingqu OS | | Spec transparency | Low | Low | Medium | Low | | Price target | $20-30K | Undisclosed | Undisclosed (est. ~$250K) | Undisclosed | | Key advantage | Scale + AI pipeline | Funding + deployment data | First to market | Volume + cost |

The take: Tesla is playing a different game than everyone else. Where Figure AI and Agility are proving the market with early deployments, Tesla is building the infrastructure for mass production. If the Optimus timeline holds, Tesla becomes the volume leader by 2028. If it slips, the window closes.

---

Data verified as of April 2026. Sources: Tesla Q1 2026 Update Letter, USPTO patent filings, Tesla AI Day presentations, Reuters, Bloomberg. Want to compare Optimus with 30+ other humanoid robots? — specs, funding, timelines, and side-by-side comparisons on Humanoid Inc.