Tesla Shipped 150 Robots Last Year. It Just Converted the Factory That Made the Model S to Build a Million.
On May 10, 2026, the last Model S rolled off the Fremont assembly line after 14 years. The line now builds Optimus Gen 3 humanoid robots. Tesla's target is one million units per year. Last year, the entire United States shipped 450 humanoid robots. China shipped 18,000.
One hundred and fifty. That's it. Every Optimus humanoid robot Tesla shipped in 2025, the full annual output of a company that Wall Street values north of $1 trillion, squeezed into a number you could count on your fingers and toes. The target was 10,000 units, a 98.5% miss.
On May 10, 2026, the last Tesla Model S sedan rolled off the Fremont, California assembly line, ending a 14-year production run that began with the car that made the world believe electric vehicles could be desirable, the car that silenced skeptics who said lithium-ion couldn't power a luxury sedan and range anxiety would kill it in the showroom. The Model X followed it into retirement after 11 years. Three hundred and fifty Signature Edition farewell units. Even those saw their deliveries postponed, sparking customer backlash.
Those lines now build humanoid robots. One million per year. That is Tesla's stated target, and getting from 150 to that number requires a 6,667-fold increase in production volume, a ramp so steep that no manufacturer in history has attempted anything comparable within a two-year window.
The Math Nobody Ran
Tesla's own history provides the benchmark. The company delivered its first Model S in June 2012 and crossed one million cumulative vehicle deliveries in early 2020, roughly eight years later. Annual production didn't reach one million until 2022, a full decade after launch.
Consider the EV ramp: 2,650 vehicles in 2012, then 22,477, then 31,655, then 50,580, then 76,230. Each year roughly doubled or tripled the last, but the base was always growing from thousands toward tens of thousands, a gentle upward slope that rewarded patience and compounding and the slow accumulation of manufacturing know-how across three continents. Going from 150 to 1,000,000 is not an S-curve. It is a vertical line.
| Metric | Tesla EV Ramp | Tesla Robot Target |
|---|---|---|
| Year 1 production | 2,650 (2012) | 150 (2025) |
| Target volume | 1,000,000/year | 1,000,000/year |
| Multiplier needed | 377x | 6,667x |
| Time to reach 1M/year | ~10 years (achieved ~2022) | 1-2 years (stated goal) |
| Manufacturing precedent | Toyota (50+ years of auto manufacturing knowledge via NUMMI) | None (no company has mass-produced humanoid robots) |
There is a more generous reading. Tesla currently produces roughly 1,000 Optimus units per month at Fremont, up from near-zero in early 2025. If that rate holds, annual output would be 12,000, not one million. To hit one million by the end of 2027, Tesla needs to increase monthly production from 1,000 to roughly 83,000 within 18 months. Even if the lines ramped at 3x per quarter, the math produces 81,000 per month by Q4 2027. Tight. Not physically impossible if the supply chain cooperates, but tight.
The Production Gap: 40 to 1
Zoom out. Tesla's bet looks less like hubris and more like desperation. In 2025, the entire United States shipped approximately 450 humanoid robots, all companies combined, in a country that put humans on the moon and builds fighter jets that cost $100 million apiece. Tesla accounted for 150. Figure AI shipped roughly 150 more, and Agility Robotics contributed the balance.
China shipped 18,000. Chinese vendors generated $440 million in humanoid robot revenue in 2025, with Unitree and AgiBot each moving between 5,000 and 10,000 units. Forty robots for every one the United States shipped.
In electric vehicles, the gap was never this wide. Tesla was the global EV sales leader from 2018 through 2022, and even after BYD overtook it, American EV production remained competitive. Not here. In humanoid robots, the race started with the US already deep in a deficit it didn't know it was running, while Chinese manufacturers built domestic actuator supply chains, exploited lower labor costs for assembly, and collected government subsidies specifically targeting humanoid robot production, including a projected 94% production growth target for 2025 alone.
The Revenue Substitution
Tesla investor relations has not published this calculation. The Model S and Model X lines at Fremont produced an estimated 15,000 to 20,000 vehicles per year in their final years, down from a peak of roughly 50,000 combined. At an average transaction price of approximately $85,000 for the Model S and $95,000 for the Model X, those lines generated somewhere between $1.3 billion and $1.8 billion in annual revenue.
Now replace every sedan with a bipedal robot. If Tesla hits its one-million-unit Optimus target at the company's stated price range of $20,000 to $30,000 per unit, the same factory floor would generate $20 billion to $30 billion in annual revenue. Same building, same square footage, fourteen to seventeen times the money.
| Product Line | Annual Units (Est.) | Avg. Price | Annual Revenue |
|---|---|---|---|
| Model S/X (final years) | 15,000-20,000 | ~$90,000 | $1.3-1.8B |
| Optimus (target) | 1,000,000 | $20,000-$30,000 | $20-30B |
| Revenue multiplier | 14-17x |
Piper Sandler analyst Alexander Potter values the Optimus business at approximately $100 per Tesla share, arguing that at the current stock price of around $420, investors are effectively getting the entire robot division for free. Free. His base case assigns $400 per share to Tesla's 17 existing product lines, leaving the robot as an unpriced option with, in his view, $15 trillion in total addressable market upside.
Margin is the missing variable. Cars have gross margins Tesla reports quarterly, hovering around 17-18% in recent periods, but robot margins are a total unknown because no company on Earth has mass-produced humanoid robots at anything approaching consumer volumes. The bill of materials for Optimus Gen 3 includes 50 actuators per hand, a custom AI5 chip, the same 4680 battery cells used in Tesla vehicles, and the full FSD neural network stack. Whether all of that can be assembled profitably at $20,000 per unit is a question that will only be answered after the lines are running and the scrap rates and yield losses and warranty claims start rolling in.
The Competition Is Not Waiting
Figure AI, the Sunnyvale-based humanoid startup, has built over 350 Figure 03 robots and 9,000 actuators at its BotQ factory. In January 2026, it produced one robot per day; by May, one per hour, twenty-four times faster in under 120 days. The Helix-02 system completed an autonomous eight-hour warehouse shift without human intervention, sorting, packing, and storing inventory from clock-in to clock-out without a single person touching a box.
Germany's Schaeffler, a $14 billion automotive supplier, signed a deal with Humanoid Inc. to deploy 1,000 to 2,000 humanoid robots across its manufacturing sites by 2032, with pilots beginning in 2026-2027. The contract includes a five-year actuator supply agreement, making Schaeffler both customer and component supplier. That's a first: no major industrial manufacturer has ever signed a long-term robot-as-a-service contract with a humanoid robot maker.
Norwegian startup 1X offers its NEO robot at $20,000 or $499 per month on a lease, establishing a consumer price point that undercuts Musk's stated $20,000-$30,000 range for Optimus before Tesla has shipped a single consumer unit. IDTechEx projects average humanoid robot costs will fall from $114,700 in 2024 to $37,000 by 2030, driven by actuator commoditization and shared AI platforms.
Strongest Counterargument
The strongest case against Tesla's robot bet is Tesla's own history with promises. Elon Musk said 10,000 Optimus units in 2025 and shipped 150. He promised Full Self-Driving would be feature-complete by the end of 2020, then 2021, then 2022. Still waiting. The Semi was unveiled in 2017 with deliveries promised for 2019; the first units shipped in late 2022, three years late. The Cybertruck was announced in 2019 for late 2021 delivery; production began in late 2023.
A pattern holds: Musk announces a target, misses it by years, and eventually delivers a version of the product at a fraction of the promised volume. If that pattern holds here, one million Optimus units per year could arrive not in 2027 but closer to 2030 or 2031, a timeline that would still represent the fastest manufacturing ramp in history and would still make the Fremont conversion a rational bet, but that would also hand Chinese manufacturers four to five additional years of uncontested production experience, potentially widening the 40:1 gap before Tesla can even begin to close it.
Converting a car line to a robot line is also not a simple retooling. Cars and robots share some supply chain overlap: batteries, wiring harnesses, control boards. But they diverge sharply on the components that matter most. Actuators. Multi-axis joints. Vision and tactile sensor arrays with no automotive analog. Tesla is building the actuator supply chain from scratch while Figure AI and Chinese competitors source from an existing ecosystem of industrial servo and motor manufacturers who have been refining these components for decades.
Limitations
Tesla does not break out Model S and Model X production volumes separately from the Cybertruck in its "Other Models" reporting category, so the 15,000-20,000 annual unit estimate for the S/X lines in their final years is derived from industry trackers and registration data, not Tesla's own filings. The $20,000-$30,000 Optimus price target comes from Musk's public statements and analyst models, not a published price list. China's 2025 humanoid shipment figures vary by source between 13,000 and 18,000 units depending on whether wheeled humanoids are included. Figure AI's production rate is self-reported and unaudited. Margin projections for Optimus at scale remain speculative; no humanoid robot has been mass-produced in volumes sufficient to establish real unit economics.
What You Can Do
If you work in manufacturing, logistics, or warehousing, the question is not whether humanoid robots arrive but when and at what hourly cost. Schaeffler's robot-as-a-service contract is the first template. Ask your operations team whether they have evaluated pilot programs from Figure, Agility, or Tesla's internal deployment pipeline. Prior LITF analysis calculated the break-even cost at $1.48 per hour for current-generation humanoids against warehouse labor. Two shifts, and the economics are already close to parity.
If you are an investor, track three numbers: Tesla's quarterly Optimus production disclosures due in Q2 2026 earnings in July, Figure AI's BotQ output rate currently at one robot per hour and targeting four per hour by year-end, and China's monthly humanoid export data from customs. The spread between Tesla's stated targets and actual shipments is the single clearest signal of whether the factory conversion is working.
If you are a policymaker in a manufacturing-dependent district, the Fremont conversion is a preview of something recursive and unprecedented. Tesla's Fremont factory employs over 22,000 people, and while Model 3 and Model Y lines continue operating, so the S/X conversion affects a subset, the direction is unmistakable: the same factory that needed thousands of assembly workers to build cars will eventually need fewer workers to build the robots that replace assembly workers at other factories, warehouses, and fulfillment centers across the country. That loop has no precedent, and workforce transition planning for it has not started in any state.
The Bottom Line
On June 22, 2012, the first Model S rolled off the Fremont line and changed what the world thought an electric car could be. Fourteen years later, the last one rolled off and was replaced by a bipedal robot with 50 actuators per hand. Tesla is betting that the second event will prove more consequential than the first. Getting there requires a ramp 17.7 times steeper than the one Tesla pulled off with cars, attempted in one-fifth the time, using a supply chain that does not yet exist, against competitors who already ship 40 units for every one the US produces. Every number in that sentence should make you uncomfortable. The question is whether it makes you uncomfortable enough to pay attention.