Tesla Optimus
Last reviewed
Jun 9, 2026
Sources
44 citations
Review status
Source-backed
Revision
v7 ยท 7,384 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Jun 9, 2026
Sources
44 citations
Review status
Source-backed
Revision
v7 ยท 7,384 words
Add missing citations, update stale details, or suggest a clearer explanation.
Tesla Optimus (originally announced as the Tesla Bot) is a general-purpose humanoid robot being developed by Tesla. First revealed at Tesla AI Day in August 2021, the program aims to build an autonomous bipedal robot that can perform dangerous, repetitive, or boring tasks in factories, warehouses, and eventually homes. The project leverages Tesla's expertise in neural networks, computer vision, and battery technology, sharing significant architectural overlap with the company's Full Self-Driving (FSD) software. As of mid-2026, Tesla has not yet begun mass production of Optimus; the company's Q1 2026 shareholder update confirmed that pilot production at Fremont is expected to start in late July or August 2026, after the Model S and Model X production lines are dismantled and converted [1][2][33].
Elon Musk first announced the Tesla Bot concept at Tesla AI Day on August 19, 2021. During the presentation, a person in a spandex robot suit danced on stage to represent the planned humanoid form factor, a moment that drew both laughter and skepticism from the robotics community. Musk described a 5-foot-8-inch, 125-pound robot that would use the same artificial intelligence systems powering Tesla's vehicles and could handle tasks that are "unsafe, repetitive, or boring" [3].
The original concept specifications announced at AI Day 2021 outlined a robot with a carrying capacity of 45 pounds (20 kg), a top speed of 5 miles per hour, and a screen on its face for displaying information. Musk emphasized that the robot would be designed so that a human could easily overpower or outrun it, addressing safety concerns about autonomous humanoid robots. He also stated that Tesla was already "the world's biggest robotics company" due to its experience with autonomous vehicle AI and factory automation [3].
The announcement was met with widespread doubt. Robotics experts pointed out that building a functional humanoid robot capable of general-purpose tasks is one of the hardest unsolved problems in engineering, and that Tesla had no track record in robotics. Bloomberg News described the announcement as "mission creep" that fell outside Tesla's clean-energy initiatives, while The Washington Post noted Tesla's history of "exaggerating timelines and overpromising at its product unveilings and investor presentations." Despite the skepticism, Musk set a bold target: a working prototype would be ready by 2022 [3].
At Tesla's second AI Day on September 30, 2022, the company presented a working prototype nicknamed Bumble-C. Built from off-the-shelf actuators and components over approximately six months, Bumble-C walked slowly and untethered across the stage for the first time, waved to the audience, and performed simple dance moves. Musk emphasized that this was the first time the robot had walked without a tether. Tesla also showed pre-recorded footage of the Bumble-C prototype handling packages in an office setting, watering plants, and doing basic manual work on a factory floor [4].
After Bumble-C walked off stage, Tesla wheeled in a second, more refined prototype that used custom-designed Tesla actuators, battery systems, and control electronics. Though this version could not yet walk under its own power, it demonstrated the design direction for the production robot, including Tesla's proprietary hardware. The Bumble-C prototype carried a 2.3 kWh battery pack, which Tesla engineers said was sufficient for roughly a full day of work. While the demonstration was modest compared to the acrobatic capabilities of robots like Boston Dynamics' Atlas, it showed meaningful progress from concept to hardware in roughly one year [4].
The AI Day 2022 event also provided the first detailed look at Tesla's actuator strategy. The engineering team presented the robot's joint architecture, showing six distinct actuator types designed for different parts of the body, ranging from high-torque actuators for leg joints to lighter, faster units for arm and hand movements. This hardware-first approach signaled that Tesla was serious about building a physical robot, not just an AI demonstration [4][22].
Throughout 2023, Tesla released several video updates showing improved capabilities. The Gen 1 Optimus could sort objects, pick up and place items, and perform basic assembly tasks. Tesla demonstrated the robot folding laundry (though later reporting suggested this task was partially teleoperated rather than fully autonomous). Gen 1 used Tesla-designed actuators throughout its body and ran on a single Tesla System-on-Chip (SoC) for onboard computation [5].
Notable Gen 1 milestones during 2023 included walking on varied terrain, autonomous calibration of its own limbs, and the ability to handle objects of different shapes and weights. Tesla published video showing the robot performing yoga poses and balancing on one leg, though these demonstrations occurred in controlled laboratory settings rather than real-world environments [5].
Tesla unveiled Optimus Gen 2 in December 2023, showcasing significant improvements over the first generation.
| Specification | Gen 1 | Gen 2 |
|---|---|---|
| Height | 5'8" (173 cm) | 5'8" (173 cm) |
| Weight | ~73 kg | ~57 kg |
| Walking speed | ~5.6 km/h (3.5 mph) | ~8 km/h (5 mph) |
| Hand degrees of freedom | 6 DoF | 11 DoF per hand |
| Carry capacity | ~20 kg (45 lbs) | ~20 kg (45 lbs) |
| Deadlift capacity | N/A reported | ~68 kg (150 lbs) |
| Battery | 2.3 kWh | 2.3 kWh |
| Actuators | Tesla-designed | Improved Tesla-designed |
| Balance | Basic walking | Improved balance, squats |
| Neck degrees of freedom | 1 DoF | 2 DoF |
Gen 2 achieved a 10 kg weight reduction (from approximately 73 kg to 57 kg), 30% faster walking speed (up to 8 km/h or 5 mph), and substantially more dexterous hands with 11 degrees of freedom per hand (up from 6 in Gen 1). The robot demonstrated smoother, more natural walking, the ability to perform squats, and tactile sensing in its fingertips that allowed it to handle delicate objects like eggs [6].
Key design changes in Gen 2 included a new neck with 2 degrees of freedom (compared to 1 in Gen 1), boot-integrated foot force and torque sensors for improved balance, and a redesigned lower body that contributed to the overall weight reduction. The hands featured metal tendons connected to the fingertips, providing tactile feedback for fine manipulation [6].
On October 10, 2024, Tesla hosted the "We, Robot" event at Warner Bros. Discovery Studios in Burbank, California, where Optimus appeared alongside the unveiling of the Cybercab robotaxi. Optimus units walked among guests, served drinks at a bar, played rock-paper-scissors with attendees, and held conversations with the crowd. The event was widely covered as an inflection point for Tesla's robotics ambitions [34][35].
In the days following the event, multiple outlets reported that many of the interactive capabilities had relied on hidden human teleoperators rather than autonomous AI. Bloomberg reported on October 14, 2024 that "sources familiar with the matter" said employees stationed remotely oversaw many of the interactions between the Optimus units and attendees, although the robots' walking was performed autonomously. TechCrunch and The Verge published similar reporting. Morgan Stanley analyst Adam Jonas wrote in a client note that the bots "relied on tele-ops (human intervention)." At least one Optimus unit at the bar was captured on video acknowledging to a guest that it was being "assisted by a human" [34][35][36].
Tesla and Musk did not publicly confirm or deny the level of teleoperation during the event itself. The disclosure controversy reinforced longstanding skepticism about the gap between Tesla's demonstrations and autonomous capabilities, and became a recurring reference point in subsequent reporting on Optimus demonstrations [34][35].
At the 2025 Tesla Annual Shareholder Meeting in November 2025, Tesla showed an interim version of the robot referred to as Optimus V2.5, which walked autonomously through the venue, waved, interacted with attendees, and danced. Musk used the event to outline a forward production roadmap rather than to claim large existing production volumes [37].
On June 6, 2025, Bloomberg and TechCrunch reported that Milan Kovac, Tesla's Vice President of Engineering for Optimus, was leaving the company effective immediately. Kovac cited personal reasons in an X post, writing that the decision was "the most difficult" of his life and that he intended to spend more time with his family abroad. Ashok Elluswamy, the head of Tesla's Autopilot/AI team, took over responsibility for the Optimus program [38][39].
Kovac had joined Tesla in 2016 as a machine learning engineer on Autopilot, was promoted to lead Autopilot software engineering in 2019, and transferred to lead the Optimus engineering team in 2022. His departure during a critical pre-production phase drew attention from analysts as one of several senior departures at Tesla in 2025 [38].
On February 17, 2026, Musk posted "This bot got hands" on X alongside a demonstration video showing the Gen 3 hand system. The hands feature 22 degrees of freedom plus 3 in the wrist/forearm, driven by 25 actuators per hand-forearm assembly (50 total for both hands), a roughly 4.5x increase in actuator count over Gen 2. Most of the actuators are housed in the forearm and connected to the fingers via tendons in a biomimetic design, allowing for a slimmer overall hand profile [7][16].
The Gen 3 hand reveal was the most detailed public look at any Gen 3 subsystem to that date. Tesla had originally indicated that the full Gen 3 robot would be unveiled in Q1 2026, but the full reveal slipped (see below) [7][16].
On the Q4 2025 earnings call (January 28, 2026), Musk said: "We will probably unveil Optimus 3 in a few months. I think it's going to be quite surprising to people." Tesla's prepared Q4 2025 commentary indicated that a production-intent prototype of Optimus V3 would be ready in early 2026, after which Tesla intended to build a one-million-unit annual production line at Fremont with a targeted production start "toward the end of next year" (i.e., late 2026) [2][40].
In the same call, Musk was unusually candid about the program's current state, telling analysts: "It's still in the R&D phase. We have had Optimus do some basic tasks in the factory. But as we iterate on new versions of Optimus, we deprecate the old versions. It's not in usage in our factories in a material way." This walked back earlier suggestions that hundreds or thousands of robots were doing useful productive work in Tesla factories [2][40].
On the Q1 2026 earnings call (April 22, 2026), Musk further delayed the Optimus V3 unveiling. He said he would prefer to unveil the robot closer to the start of production, which he described as "somewhere around the late July, August time frame." He cited concerns about competitor reverse-engineering: previous Optimus generations had given competitors detailed footage from which to copy joint geometry and actuator placement. He stated that compressing the gap between public reveal and shipping would make imitation less practical [33][41].
Musk also described the conversion of the Fremont Model S/X production lines into an Optimus pilot line as "an insanely fast speed" if completed within four months of the last Model S/X rolling off the line in early May 2026. Regarding the part complexity, he said: "It will move as fast as the least lucky, slowest, dumbest part in the entire 10,000. It is impossible to predict these things." He declined to provide specific 2026 unit production targets [33][41].
The Gen 3 hands represent the most significant hardware upgrade revealed to date, featuring 22 degrees of freedom plus 3 in the wrist and forearm. Each hand-forearm assembly is driven by 25 actuators (50 total for both hands), a roughly 4.5x increase from Gen 2. This gives the hands enough dexterity to manipulate small objects, turn screws, and perform tasks requiring fine motor control. Tesla has stated that the Gen 3 hand system enables a much wider range of tasks than Gen 2, including delicate battery cell manipulation and handling fragile objects [7][16].
| Specification | Gen 2 | Gen 3 |
|---|---|---|
| Hand degrees of freedom | 11 DoF per hand | 22 DoF + 3 (wrist/forearm) per hand |
| Hand actuators | ~11 per hand | 25 per hand (50 total) |
| Actuator location | Primarily in hand | Primarily in forearm (tendon-driven) |
| Tactile sensing | Basic fingertip | Advanced multi-point |
| Fine motor control | Moderate | High (screws, small objects) |
Tesla designed the Optimus hardware platform from the ground up, drawing on its experience with electric vehicle powertrains, battery systems, and autonomous driving electronics. The hardware architecture is organized around three main subsystems: the actuator system, the compute and sensing platform, and the power system.
Tesla developed a portfolio of six custom actuator types for Optimus, divided into three rotary reducers and three linear actuators. Each type is optimized for the specific force, speed, and range-of-motion requirements of different body joints.
| Actuator Type | Category | Force/Torque Rating | Location |
|---|---|---|---|
| Small rotary | Rotary reducer | 20 Nm | Wrist, neck |
| Medium rotary | Rotary reducer | 110 Nm | Elbow, ankle |
| Large rotary | Rotary reducer | 180 Nm | Hip, shoulder |
| Small linear | Linear actuator | 500 N | Hand, wrist |
| Medium linear | Linear actuator | 3,900 N | Elbow, spine |
| Large linear | Linear actuator | 8,000 N | Knee, hip |
The linear actuators use inverted planetary roller screws, a relatively uncommon mechanism in consumer and most industrial applications. In this design, the screw remains stationary while the nut rotates, driven directly by the servo motor's rotor. Planetary roller screws offer higher shock load resistance than standard ball screws, making them well suited for the dynamic impacts that occur during bipedal walking. The configuration of each linear actuator includes a frameless torque motor, a planetary roller screw, a force sensor, an encoder, a driver board, a ball bearing, and a four-point contact ball bearing [22][23].
The robot contains 14 linear joints distributed across the body: 2 in the elbows, 4 in the wrists, and 8 in the legs. The rotary joints use harmonic drive reducers paired with brushless DC motors, providing smooth motion with minimal backlash. Together, the actuator system gives Optimus a total of 28 degrees of freedom in the body (excluding the hands), enabling human-like range of motion for walking, bending, reaching, and turning [22][23].
Optimus runs on a single Tesla-designed System-on-Chip (SoC), the same chip family used in Tesla vehicles for autonomous driving. The FSD chip (Hardware 3 generation) features twelve ARM Cortex-A72 CPUs operating at 2.6 GHz, two neural network accelerator systolic arrays operating at 2 GHz, and a Mali GPU operating at 1 GHz. Tesla claims the FSD chip can process images at 2,300 frames per second, a 21x improvement over earlier hardware generations [24].
The newer Hardware 4 (HW4 / AI4) chip, which Tesla has begun integrating into later Optimus builds, features 20 ARM Cortex-A72 CPU cores at up to 2.35 GHz, along with a dual-redundancy architecture where two separate computing modules run in parallel and cross-check each other's results. Samsung manufactures the HW4 processor on a 7 nm process. The HW4 board includes 16 GB of RAM and 256 GB of storage, double and quadruple the amounts in HW3, respectively [24].
For perception, Optimus uses a vision-only approach with no LiDAR. The robot is equipped with multiple autopilot-grade cameras positioned around its head and body, providing stereo depth perception and 360-degree awareness. Additional sensors include:
This sensor architecture mirrors the philosophy behind Tesla's FSD system: rely on cameras and neural networks to interpret the environment, rather than expensive specialized sensors like LiDAR or radar [9][24].
Optimus is powered by a 2.3 kWh lithium-ion battery pack, designed and manufactured by Tesla using its expertise in EV battery technology. Tesla engineers have stated the battery provides approximately a full day of light to moderate work on a single charge. The battery integrates directly into the robot's torso, doubling as a structural element that contributes to the overall rigidity of the chassis. The power distribution system uses Tesla's in-house power electronics to deliver precisely regulated voltage and current to each of the robot's dozens of actuators [4][6].
The Optimus robot's intelligence is built on the same foundational technology as Tesla's Full Self-Driving system, an approach that distinguishes it from most other humanoid robot programs.
Ashok Elluswamy, who took over the Optimus program in June 2025 following Milan Kovac's departure, has stated that the same end-to-end neural network architecture used for FSD will transfer to the humanoid robot. FSD version 12, which replaced approximately 300,000 lines of hand-coded C++ with learned neural networks, demonstrated that Tesla's AI team could build systems that learn complex behavior from data rather than explicit programming [8][38].
Gen 3 robots are expected to run on the FSD-v15 neural architecture, the latest iteration of Tesla's autonomous driving AI stack adapted for humanoid robotics. This architecture processes visual, tactile, and proprioceptive sensor data through a unified neural network that outputs motor commands for the robot's actuators [8][16].
At the 2025 Annual Shareholder Meeting, Musk also teased integration with Tesla and xAI's Grok large language model to allow natural-language command of the robot [37].
Tesla's approach to training Optimus has evolved substantially since the program began. The current pipeline combines several complementary techniques:
Teleoperation demonstrations: Human operators control the robot remotely using VR controllers or specialized rigs, demonstrating desired behaviors and generating labeled training data. Each teleoperated session produces paired sequences of sensor inputs and motor outputs that serve as supervised training examples [8][9]. This is the same technology that is widely understood to have been used at the 2024 "We, Robot" event for the interactive demonstrations (see above) [34].
Video-based imitation learning: In a notable shift from the initial training strategy, Tesla moved away from motion capture suits and toward a vision-only approach. Workers now wear camera rigs consisting of helmets and backpacks equipped with five in-house cameras that record mundane tasks like folding a shirt or picking up an object. These videos are then used to train Optimus to mimic the actions. According to insiders, this change was designed to help Tesla scale data collection faster, since recording video requires far less specialized equipment than traditional motion capture [25].
Sim-to-real transfer: Tesla's neural world simulator generates millions of synthetic scenarios for the robot to practice in. The robot acquires skills through a "Sim-to-Real" training pipeline, simulating actions millions of times in a virtual world before transferring these skills to real hardware. This approach dramatically reduces the amount of real-world data needed for each new task [8][9].
Real-world data collection: Robots deployed in Tesla factories gather data continuously, creating a feedback loop that improves models over time. Tesla has been recording video of human workers performing factory tasks at its Fremont facility for over a year to build its training dataset [8][9].
The combination of these techniques creates a data flywheel similar to the one Tesla uses for FSD: more robots in the field generate more data, which trains better models, which make the robots more capable, which justifies deploying more robots. This self-reinforcing cycle is central to Tesla's scaling strategy [9].
The video-based training approach is a core differentiator for Tesla's robotics program. Rather than requiring explicit programming for each new task, Optimus learns by observation:
This pipeline enables Optimus to learn new tasks without requiring manual programming for each one, a key requirement for the general-purpose flexibility Tesla envisions [8][9][25].
Since mid-2024, Tesla has piloted Optimus units at its Fremont, California and Austin, Texas (Giga Texas) factories. The deployment began with simple tasks and has gradually expanded in scope, though the primary purpose has remained data collection and algorithm refinement rather than productive labor.
One of the first operational tasks demonstrated for Optimus was sorting 4680 battery cells at Tesla's factories. This task showcased the robot's ability to combine visual perception with precise manipulation. Tesla shared video of this task in May 2024, with the company stating that Optimus's end-to-end neural network ran in real time on the bot's FSD computer, using only 2D cameras along with hand tactile and force sensors to identify, grasp, and sort battery cells [26].
Musk stated at the beginning of 2025 that Tesla expected to build 5,000 to 10,000 Optimus units that year. The target was reportedly cut to roughly 2,000 internally as the year progressed. Independent estimates and reporting put actual mid-2025 production at no more than a few hundred units, well below either target. Causes reported by industry press included a Gen 3 hand redesign that took longer than expected and supply-chain disruptions tied to China's April 2025 restrictions on neodymium-iron-boron rare-earth magnet exports, which forced Tesla to apply for individual export licenses [27][42].
On the Q4 2025 earnings call on January 28, 2026, Musk acknowledged that Optimus was "still in the R&D phase" and "not in usage in our factories in a material way." He described existing Optimus units as primarily being used for learning rather than productive work: "We have had Optimus do some basic tasks in the factory. But as we iterate on new versions of Optimus, we deprecate the old versions." Tesla CFO Vaibhav Taneja told investors to expect approximately $20 billion in capital expenditure in 2026, with investments in Optimus, new factories, and AI computing resources [2][40].
On December 7, 2025, Tesla held an "Autonomy Visualized" event in Miami where an Optimus unit handed out water bottles to attendees. In a video that subsequently went viral, the robot knocked over its tray, raised its hands in a gesture observers described as resembling a person removing a VR headset, and then fell backward. Multiple commentators noted that the gesture closely matched the motion of a teleoperator pulling off a headset, suggesting that the unit had been remotely operated. Tesla did not publicly comment on the cause of the fall [19][43].
Given Tesla's documented use of teleoperators at "We, Robot" the previous year, the Miami incident reinforced criticism that Tesla's public demonstrations frequently rely on hidden teleoperation rather than autonomous AI. Critics argued that demonstrations of even simple tasks like distributing water bottles were still being assisted by humans more than two years after the program's first major public reveal [19][43].
| Task Category | Status (Mid-2026) | Complexity Level | Description |
|---|---|---|---|
| Battery cell sorting | Demonstrated (R&D / data collection) | Low | Sorting and organizing 4680 cells |
| Parts transport | Limited pilot (R&D / data collection) | Low | Moving parts between workstations |
| Pick-and-place operations | Limited pilot (R&D / data collection) | Low-Medium | Retrieving and positioning components |
| Component assembly | Testing | Medium | Fitting parts together on assembly lines |
| Quality inspection | Testing | Medium | Visual inspection of manufactured parts |
| Complex manipulation | Development | High | Multi-step handling of irregular objects |
| Multi-step manufacturing | Development | High | Sequential production workflows |
On the Q4 2025 and Q1 2026 earnings calls, Tesla confirmed that it would convert the Fremont production lines previously used for the Model S sedan and Model X SUV (ending production in early May 2026) into an Optimus pilot line, with a target capacity of 1 million units per year. Musk emphasized on the Q1 2026 call that this would require dismantling the existing line and installing the new one in approximately four months, a timeline he described as "an insanely fast speed" for a complex manufacturing line [2][33].
Tesla has also confirmed plans to build a second, dedicated Optimus production facility at Giga Texas, with a longer-term capacity of up to 10 million units annually. Musk indicated on the Q1 2026 call that the Texas facility was expected to begin production around summer 2027 and would eventually produce a higher-volume Gen 4 variant [33].
| Facility | Location | Purpose | Target Capacity | Timeline |
|---|---|---|---|---|
| Fremont (R&D pilot) | Fremont, CA | Prototyping and testing | Hundreds per year | Active since mid-2024 |
| Fremont (converted lines) | Fremont, CA | Gen 3 pilot production | 1 million units/year (long-term) | SOP late July/August 2026 |
| Giga Texas (dedicated) | Austin, TX | High-volume production | 10 million units/year (long-term) | Production ~summer 2027 |
| Milestone | Target Date | Status |
|---|---|---|
| Bumble-C prototype | September 2022 | Achieved |
| Gen 1 demonstrations | 2023 | Achieved |
| Gen 2 unveiling | December 2023 | Achieved |
| Internal factory pilot deployment | Mid-2024 | Achieved (small scale) |
| "We, Robot" event | October 10, 2024 | Achieved (largely teleoperated) |
| 5,000-10,000 units in 2025 (Musk goal) | 2025 | Missed (a few hundred per independent estimates) |
| Optimus V2.5 at shareholder meeting | November 2025 | Achieved |
| Q4 2025 earnings disclosure: still R&D | January 28, 2026 | Reported |
| Gen 3 hand reveal on X | February 17, 2026 | Achieved |
| Last Model S/X off Fremont line | Early May 2026 | Planned |
| Optimus V3 unveiling | Mid-to-late 2026 (delayed from Q1) | Pending |
| Fremont V3 pilot production SOP | Late July/August 2026 | Planned |
| Meaningful production volumes | End of 2026 / 2027 | Planned |
| Optimus useful "outside Tesla" | "Sometime next year" (i.e., 2027) | Stated by Musk Q1 2026 |
| Consumer sales | End of 2027 (per Musk WEF Jan 2026) | Planned |
| Giga Texas dedicated facility SOP | Summer 2027 | Under construction |
| 1M units/year run-rate | 2027-2028 | Long-term target |
| 10M units/year (Giga Texas) | 2029+ | Ultimate capacity |
Independent reporting and analyst commentary have repeatedly noted that each successive Optimus timeline has slipped, including Musk's 2022 forecast of production-ready robots by 2023, his early-2025 forecast of 5,000-10,000 units in 2025, and the Q1 2026 forecast that V3 would be unveiled in Q1 2026 [12][18][27][33].
Musk has consistently targeted a long-run consumer price of $20,000 to $30,000 for Optimus once mass production is achieved. At the World Economic Forum in January 2026, Musk said Tesla planned to sell Optimus to the public by the end of 2027 at this price range, stating, "I think everyone on earth is going to have one and want one." Tesla has not officially disclosed current unit cost. Industry analysts estimate current manufacturing cost is well above the long-term target, with the Gen 3 hand system alone estimated at $30,000 to $80,000 at current low volumes. Initial commercial units sold to enterprise customers are widely expected to be priced in the $100,000 to $150,000 range, though Tesla has not announced a confirmed price list [1][11][44].
| Component | Estimated Current Cost | Projected Mass Production Cost |
|---|---|---|
| Gen 3 hand system (pair) | $30,000 - $80,000 | $3,000 - $5,000 |
| Tesla SoC and compute | $5,000 - $10,000 | $1,000 - $2,000 |
| Actuators and motors | $10,000 - $20,000 | $3,000 - $5,000 |
| Battery and power system | $3,000 - $5,000 | $1,000 - $2,000 |
| Frame and structural | $2,000 - $5,000 | $1,000 - $2,000 |
| Sensors and cameras | $3,000 - $5,000 | $1,000 - $2,000 |
| Assembly and testing | $5,000 - $10,000 | $2,000 - $3,000 |
| Total estimated | $58,000 - $135,000 | $12,000 - $21,000 |
These cost estimates are industry analyst projections rather than official Tesla disclosures and are subject to significant uncertainty. The dramatic cost reduction at scale depends on Tesla achieving automotive-style mass production volumes, which has not yet been demonstrated for humanoid robots [1][11].
The humanoid robot space has grown significantly since Tesla announced Optimus, with several well-funded competitors pursuing similar goals. Each company brings different strengths, philosophies, and target markets.
| Robot | Company | Height | Weight | Key Strength | Hand DoF | Estimated Price | AI Approach | Status (Mid-2026) |
|---|---|---|---|---|---|---|---|---|
| Optimus Gen 3 | Tesla | 173 cm | ~57 kg | Mass production path, FSD AI | 22 DoF + 3 wrist | $20K-$30K target | FSD neural networks, end-to-end learned | Pre-production; V3 reveal pending |
| Atlas (Electric) | Boston Dynamics | 150 cm | ~89 kg | Agility, 50 kg lift, rugged | Multi-DoF grippers | ~$140,000+ | Model-based + learned | Production version; 2026 deployments committed |
| Figure 03 | Figure AI | ~170 cm | ~60 kg | BMW factory proven track record | 16 DoF (Figure 02) | TBD | OpenAI-partnered LLM + vision | In development; Figure 02 retiring |
| Digit v4 | Agility Robotics | 175 cm | ~65 kg | Logistics specialization, 8hr battery | Grippers | ~$250,000 (RaaS) | Task-specific learned | Active commercial pilots |
| NEO | 1X Technologies | 167 cm | ~30 kg | Consumer-ready, soft-body, light | 22 DoF per hand | $20,000 / $499 per month | Embodied AI, NVIDIA Jetson Thor | Pre-orders open, shipping 2026 |
| Phoenix Gen 8 | Sanctuary AI | 170 cm | ~70 kg | Dexterous manipulation, 24hr task learning | 20 DoF (Carbon hand) | ~$65,000 | "Carbon" cognitive architecture | Pilot with Magna International |
| G1 | Unitree Robotics | 127 cm | ~35 kg | Low cost, mass produced | 3-finger grippers | $16,000 - $53,400 | Reinforcement learning | ~5,000 units shipped (H1 2025) |
Boston Dynamics Atlas: The electric Atlas, unveiled in April 2024 and shown in production form at CES 2026 (January 5, 2026), features 56 degrees of freedom, fully rotational joints at key articulation points, a 2.3-meter reach, and can lift up to 50 kg (110 lbs). It operates in temperatures from -20 to 40 degrees Celsius and has a runtime of about four hours. All 2026 deployments are already committed, with fleets scheduled to ship to Hyundai's Robotics Metaplant Application Center and Google DeepMind. Hyundai, which owns Boston Dynamics, is investing $26 billion in U.S. operations, including a new robotics factory capable of producing 30,000 robots per year. The company plans to use Atlas in car plants by 2028 for parts sequencing, expanding to component assembly by 2030 [13][28].
Figure AI: Figure gained attention through its partnership with OpenAI for conversational AI integration. The Figure 02 completed an 11-month pilot at BMW's Spartanburg plant, assisting in the production of over 30,000 BMW X3 vehicles. During that deployment, the robot moved over 90,000 components, clocked 1,250 operating hours and approximately 1.2 million steps, retrieving and positioning sheet metal parts for welding with millimeter accuracy. Figure 03 is now in development, incorporating lessons from the BMW deployment [29].
Agility Robotics Digit: Purpose-built for logistics, Digit v4 can load and unload totes from AMRs, flow racks, shelves, and conveyor belts. At GXO's Flowery Branch facility, Digit moved over 100,000 totes, and a small fleet is deployed at a Spanx facility. Agility closed a $400M Series C round in 2025 with backing from Amazon [30].
1X Technologies NEO: The NEO launched for pre-order in late October 2025 at $20,000 (or $499/month subscription) with U.S. shipments beginning in 2026. At just 30 kg, it uses a patented Tendon Drive actuation system with high-torque-density motors. It runs on an NVIDIA Jetson Thor-based compute module with up to 2,070 FP4 TFLOPS, and uses vision-only navigation with dual 8.85 MP stereo fisheye cameras at 90 Hz. OpenAI is an investor in 1X Technologies [31].
Unitree G1: The Chinese-made G1 is notable for its low price point ($16,000 to $53,400 depending on configuration) and early mass production. Approximately 5,000 G1 units shipped in the first half of 2025, making it one of the highest-volume humanoid robots on the market. At 127 cm tall and 35 kg, it is smaller than most competitors but features 23 to 43 degrees of freedom and a 2 m/s walking speed [32].
Tesla's primary differentiator is its stated path to mass production at consumer-friendly price points, leveraging the same manufacturing infrastructure used for vehicles. No other humanoid robot maker has announced production capacity at the volumes Tesla is targeting, though Tesla has not yet demonstrated the ability to reach those volumes either [1][13].
The broader humanoid robotics market is attracting significant investment and growing analyst attention. Major financial institutions have released detailed forecasts that frame the industry as a multi-trillion-dollar opportunity over the coming decades.
| Source | Forecast Period | Market Size Projection |
|---|---|---|
| Goldman Sachs | By 2030 | 250,000+ humanoid shipments (base case) |
| Goldman Sachs | By 2035 | $38 billion (revised upward 6x from $6 billion) |
| Morgan Stanley | By 2030 | 40,000 humanoid robots in operation |
| Morgan Stanley | By 2050 | $5 trillion (including supply chains, repair, maintenance) |
| Morgan Stanley | By 2050 | 1 billion+ humanoid robots (90% industrial/commercial) |
| MarketsandMarkets | 2025 to 2030 | $2.92 billion to $15.26 billion (39.2% CAGR) |
Goldman Sachs notably revised their 2035 forecast upward by 6x, from $6 billion to $38 billion, reflecting how quickly the supply chain and AI capabilities have matured. Morgan Stanley Research estimates the humanoid market could reach $5 trillion by 2050 when including related supply chains, repair, maintenance, and support services. However, Morgan Stanley also projects that adoption will remain relatively slow until the mid-2030s, accelerating in the late 2030s and 2040s [20][21].
The Optimus program has drawn significant skepticism from robotics researchers, industry analysts, and Wall Street firms on several fronts.
Timeline credibility: Musk has a well-documented history of setting aggressive timelines that are not met. He initially suggested Optimus could be in production by 2023, a target that proved far too optimistic. Each subsequent deadline has been pushed back. Musk initially set internal goals for Tesla to produce at least 5,000 Optimus units in 2025, a target that was reportedly slashed to 2,000 mid-year and ultimately resulted in only a few hundred units being built. The Gen 3 reveal itself, originally promised in Q1 2026, slipped to "a few months" on the January 28, 2026 earnings call and again to "late July, August" on the April 22, 2026 earnings call [12][18][27][33].
Teleoperation vs. autonomy: Several of Tesla's most impressive Optimus demonstrations, including folding laundry, the 2024 "We, Robot" event interactions, and the December 2025 Miami water-bottle demonstration, were reported by named outlets (Bloomberg, TechCrunch, The Verge, Electrek) to involve human teleoperation. While teleoperation is a standard technique for collecting training data, critics argue that Tesla has sometimes presented these demos in ways that overstate the robot's autonomous capabilities [12][19][34][43].
General-purpose feasibility: Building a robot that can perform a wide range of tasks in unstructured environments (homes, varied workplaces) is an extraordinarily difficult problem. Many robotics experts believe that purpose-built robots designed for specific tasks will deliver more practical value than humanoid generalists for the foreseeable future. A Gartner analyst covering emerging technologies and robotics stated that humanoid robots "face too many limitations to be practical" in the near term [12][18].
Economic case: At $20,000 to $30,000, Optimus would need to deliver enough value to justify the investment for consumers or small businesses. The economic case is stronger in industrial settings where robots can operate continuously, but the consumer market remains uncertain [11].
Public demonstration failures: The December 2025 Miami incident, in which an Optimus unit fell backward in a motion that observers read as a teleoperator removing a VR headset, became a touchpoint in public skepticism about Tesla's autonomy claims. The clip went viral on X and Reddit and was covered by Electrek, BGR, and other outlets [19][43].
Wall Street opinion on Optimus is divided:
| Firm | Stance | Key Assessment |
|---|---|---|
| Goldman Sachs | Cautiously optimistic | Notes Tesla is "happy with hardware progress"; highlights capability, reliability, and manufacturability as key factors for scaling |
| Morgan Stanley | Skeptical on valuation | Downgraded Tesla in early 2026; warns stock already reflects full AI and robotics valuation |
| Gartner | Skeptical on timeline | Humanoid robots face too many limitations to be practical in near term |
Goldman Sachs analyst Mark Delaney noted that Tesla "highlighted the importance of the capability, reliability, and manufacturability of its design for scaling," while acknowledging that the stock could maintain a higher multiple reflecting long-term AI and robotics opportunities. However, Morgan Stanley downgraded Tesla in early 2026, citing concerns that the company's stock price already fully reflects its robotics and AI ambitions [20][21].
Despite these concerns, Tesla's advantages in vertical integration, AI talent, manufacturing scale, and access to billions of miles of real-world driving data for computer vision training give the company credible capabilities that few competitors can match.
If Tesla or any competitor succeeds in building a mass-produced humanoid robot at consumer price points, the implications extend far beyond the robotics industry. Musk has described Optimus as potentially "the most important product Tesla ever makes" and suggested it could eventually represent the majority of Tesla's long-term value.
The economic implications are significant. A functional, affordable humanoid robot could transform industries including manufacturing, logistics, agriculture, construction, and elder care. Musk has spoken about Optimus enabling a future of "universal high income," where robots handle most physical labor and human workers shift to supervisory, creative, and interpersonal roles. Critics counter that this vision raises serious questions about employment displacement, wealth concentration, and the social structures needed to manage a transition of this scale [15][19].
From a technical perspective, the development of Optimus is accelerating progress in several adjacent fields: actuator design, tactile sensing, sim-to-real transfer, imitation learning, robot teleoperation, and whole-body control. Even if Optimus itself falls short of Musk's most ambitious claims, the research and engineering investment is producing advances that benefit the broader robotics ecosystem.
As of mid-May 2026, Tesla Optimus has not yet entered mass production. The last Model S and Model X vehicles are expected to roll off the Fremont production line in early May 2026, after which Tesla will begin a roughly four-month conversion of those lines into an Optimus pilot line, with start of production targeted for late July or August 2026. The full Optimus V3 robot has not yet been unveiled in production form; Tesla has shown only the Gen 3 hand system publicly (in February 2026), and Musk has stated that he intends to delay the full V3 reveal until just before production begins, citing concerns about competitor imitation.
In factory deployments, Musk acknowledged on the Q4 2025 earnings call (January 28, 2026) that Optimus is "still in the R&D phase" and is "not in usage in our factories in a material way." On the Q1 2026 call (April 22, 2026), he said the robot would "probably" be useful "outside of Tesla sometime next year" (2027). Consumer sales remain targeted for end of 2027 at the $20,000-$30,000 price point, with a dedicated second factory at Giga Texas targeted to begin production around summer 2027 with eventual capacity of 10 million units per year. The program remains one of the most ambitious and closely watched efforts in the robotics industry, though the recurring gap between Musk's stated vision and current capabilities continues to fuel both excitement and skepticism.