Can Optimus reach a unit cost that makes humanoid labor cheaper than people?

Thesis

Inside Tesla’s accounts sit a mature vehicle, storage and services business and a physical AI venture. The mature business produced $94.8 billion of revenue in 2025 and earned $4.36 billion of operating income on it. The venture, covering driving software, a robotaxi network, a custom chip program and a humanoid robot, consumes most of the incremental cash. Annual revenue was $96.8 billion in 2023 and $94.8 billion in 2025, and Q3 2026 revenue of $28.05 billion was roughly flat against a year earlier. Operating income fell to $718 million because research and development rose 69% to $2.76 billion.

On labor cost alone, the arithmetic for Optimus is forgiving. The Bureau of Labor Statistics put total employer compensation for manufacturing workers at $48.62 per hour in June 2026. Over 10,000 hours of work, which is five years of a single 2,000-hour shift (our assumption), that is about $486,000 of compensation per position. On Tesla’s Q4 2024 earnings call in January 2025, Elon Musk said Optimus could sell for under $20,000 once production reaches one million units a year, as Electrek reported. That is a stated selling price, not a disclosed cost. Tesla has not published an actual Optimus build cost. At $20,000, hardware would be about 4% of the labor it replaces (our calculation). Even at five times that figure it would be about a fifth.

Purchase price is unlikely to be the binding constraint. The harder variables are what fraction of a human’s output a robot delivers, how many hours a year it runs, how often it needs service, and who supervises it. None has been disclosed. Tesla has disclosed that Optimus production began in Fremont on lines it calls high volume, that the first robots are early ramp, and that Musk, on the October 21, 2026 earnings call, tied outside deliveries to remaining issues being solved around the end of 2027.

The financial setup is demanding. The market capitalization of about $1.58 trillion on October 30, 2026 is roughly 456 times trailing twelve month operating income (our calculation from the filings cited in the Valuation section), so the price rests on expected outcomes in autonomy and robotics. The company has also changed its balance sheet posture, signing $30 billion of new credit commitments in September 2026 against $45.4 billion of cash and investments.

Company History

Tesla was incorporated in Delaware on July 1, 2003 and converted to a Texas corporation on June 13, 2024. It began as an electric vehicle maker and added stationary storage and solar over time. By the third quarter of 2026 it had delivered its 10 millionth vehicle, with 10.2 million cumulative deliveries on the company’s count.

The recent history is a change in self-description. The 2025 annual report opens by saying Tesla is focused on bringing AI into the real world through FSD (Supervised), Robotaxi and AI robots including Optimus. Tesla launched its Robotaxi service in June 2025. In September 2026 it began Robotaxi rides in Cybercabs in Austin, the purpose-built autonomous vehicle.

Capital allocation followed the narrative. Tesla decommissioned the Model S and Model X lines at Fremont and began installing the first generation Optimus lines there, as stated in the Q2 2026 update. In March 2026 it invested $2.00 billion in SpaceX common stock, a position that began as a preferred share investment in xAI and represents under 1% ownership. Tesla accounts for it under the equity method because it presumes significant influence.

Governance changed as well. The board adopted a 2025 CEO Performance Award whose operational milestones include 20 million vehicles delivered, 10 million active FSD subscriptions, 1 million bots delivered and 1 million Robotaxis in commercial operation, plus Adjusted EBITDA tiers up to $400 billion. The award matters here because it defines, in Tesla’s own filings, what success in physical AI is meant to look like. Today’s counts, 10.2 million vehicles and 1.74 million FSD subscriptions, show how far each target is from the current base.

Product

Vehicles

Tesla delivered 486,532 vehicles in Q3 2026, down 2% from 497,099 a year earlier, against production of 464,391. Model 3 and Model Y accounted for 478,237 of those. Other models, a group that includes Cybertruck, Semi and the remaining Model S and X, fell to 8,295 from 15,933.

Inventory is lean at 11 days of supply. Tesla says battery availability was the limiting factor on auto volume and that it ended the quarter with its largest passenger vehicle backlog in four years. For full year 2025, deliveries were 1,636,129, a decline of about 9%.

The Tesla Semi entered volume production in Nevada. Tesla lists installed Semi capacity of 25,000 units a year and Cybercab capacity at Texas of more than 125,000 a year, with the caveat in the same table that installed capacity is not a production rate.

FSD and Robotaxi

FSD (Supervised) is software sold as a purchase or a subscription, and it requires an attentive driver. Tesla reports 1.74 million active FSD subscriptions, up from 1.04 million a year earlier, a rise of about 67%. The count includes upfront purchases, per the CEO award definition. On the earnings call, CFO Vaibhav Taneja said over 75% of North American Q3 deliveries included a paid FSD sign-up at delivery. FSD is now approved in nine European countries, used by over 100,000 customers who have logged more than 160 million kilometers there.

The Robotaxi service reached nearly two million cumulative unsupervised miles by the end of Q3, and Tesla says the figure grew eightfold quarter over quarter. Austin service hours now run from 6am to midnight. On the call, Musk said the Austin Cybercab fleet had grown by a factor of 10 since launch about a month earlier. Tesla gave no fleet count.

These are disclosed facts. The following are targets stated by management on the call: Ashok Elluswamy, Tesla’s VP of AI software, said the first customer vehicles could join the Robotaxi network around Q2 2027; Musk said Robotaxi becomes noticeable around Q2 2027 and a big deal by Q3 or Q4 2027; and Taneja said Tesla is not yet at its $0.30 per mile operating cost goal but sees a path to it. Tesla does not report Robotaxi revenue, fleet profitability, or the share of rides that involve remote assistance.

Optimus

Optimus is a general purpose humanoid robot. Tesla’s current disclosures are short. The Q3 update says Optimus production began in Fremont on high-volume lines and describes the output as the “first production-quality bots” Tesla has made. Tesla also says the current manufacturing process is suitable for a commercial product. The capacity table lists California Optimus as Early Ramp and the Texas site as under construction. Tesla reports no Optimus unit count, revenue or cost.

The program has slipped against its own earlier dates. The Q4 2025 update said Tesla planned to unveil the Gen 3 version in Q1 2026 and start production before the end of 2026. The production start was met in Fremont. We found no report of the Gen 3 unveil in the Q3 update, and on the call management said a video of Optimus doing tasks that are not pre-programmed may be ready in a month or two.

Targets, as opposed to facts, are these. The Fremont line is designed for 1 million robots a year, and a second generation line at Giga Texas is being designed for 10 million a year of long-term capacity. Management said reaching an annualized rate of about 20,000 robots a week will take at least a year, maybe longer, and that integration of the AI5 chip into Optimus is targeted for approximately April 2027. The Q2 update said initial builds would be used in an Optimus Academy for data collection and development, which suggests the first units serve as training assets.

AI compute and chips

Tesla trains its driving and robot models on its own clusters in Texas. The Q3 update lists Cortex 1 at more than 105 MW and Cortex 2 at more than 120 MW, both in production, and states a plan to reach 400 MW of training compute by the end of 2026. That is a target. The two clusters listed total about 225 MW, so roughly 175 MW more would have to come online in two months (our arithmetic; Tesla’s chart also counts 40 MW of other compute, so the gap may be smaller).

On silicon, Tesla is building a semiconductor fab in Austin and plans to start commissioning equipment in 2027. On the call, Musk said machinery commissioning would begin in the first half of 2027 and estimated Terafab equipment spending at tens of billions of dollars through early 2028, and more than $100 billion over the long term.

Energy storage

Tesla sells Megapack for utilities and large commercial sites and Powerwall for homes. It deployed 13.7 GWh in Q3 2026, up from 12.5 GWh a year earlier, and 36.0 GWh in the first nine months (our sum of the quarterly figures of 8.8, 13.5 and 13.7 GWh). Full year 2025 deployments were 46.7 GWh. Megafactory Texas began production with planned capacity of 50 GWh, and the first Megapacks using Tesla-made LFP cells from Nevada have shipped. Tesla links the business to AI demand, noting that energy storage is critical to data center uptime.

Services

Services and other covers charging, insurance, parts, collision work and used vehicles. Revenue was $4.61 billion in Q3, up from $3.48 billion, and gross profit was a record $734 million. The Supercharger network had 84,465 connectors at quarter end.

Market

Customer

Tesla has five distinct customer groups. Vehicle buyers are consumers and fleets, and they are the revenue base: automotive revenue was $20.56 billion in Q3, about 73% of the $28.05 billion total. Software buyers pay for FSD, and on the call an analyst estimated FSD annual recurring revenue at roughly $1 billion, and Taneja said the numbers were correct. Robotaxi riders are a new group that pays per ride and is served by Tesla’s own fleet.

Energy customers are utilities, developers and, increasingly, compute builders. One is a related party: Tesla recognized $221 million of Q3 revenue and $626 million over nine months from SpaceX for Megapacks, with $729 million of receivables from that counterparty. That is a concentration worth watching in a segment with $2.88 billion of quarterly revenue.

Optimus has no disclosed customers. Musk said on the call that early uses would probably be industrial applications, with home robots as production scales, and Tesla’s own factories are the first user.

Market Size

Tesla’s addressable markets differ widely in maturity. Autos and storage are measured markets in which Tesla already has revenue. Autonomous ride hailing is nascent: Waymo, the leader, said in 2026 that it was handling about 500,000 paid rides a week and targets 1 million by year end. Humanoid robots are earlier still. Unitree, which its prospectus says ranks first globally, reported shipping 5,500 humanoids in 2025. We cite no 2035 market forecast because the ones we found could not be traced to a primary publication.

The labor pool that humanoids target is large and expensive. Private industry compensation averaged $46.89 per hour in June 2026, and manufacturing averaged $48.62, made up of $32.50 in wages and $16.12 in benefits. A robot that substitutes for even a fraction of those hours at a low enough cost has an obvious economic case on paper. The unresolved question is capability.

Competition

Competitive Landscape

Competition differs by business line. In vehicles, Tesla faces price pressure and share loss from Chinese makers. In autonomy, Waymo has a head start in driverless paid service, while Tesla has a far larger base of cars collecting data and a software revenue stream. In humanoids, Chinese hardware makers are selling real units at low prices, and U.S. startups have raised large private rounds on thin disclosure. Tesla’s differentiator is claimed vertical integration of chips, training compute, factories and a data source in its vehicle fleet, which no humanoid startup has.

The practical test is cost per unit of useful work. Tesla has disclosed neither a robot cost nor a robotaxi cost per mile, so comparisons below use competitors’ published numbers where they exist.

Competitors

Waymo

Waymo operates driverless service in 11 U.S. cities over more than 1,400 square miles as of May 2026, with a fleet reported at roughly 3,500 vehicles and more than 20 million trips (the article is inconsistent on fleet size, giving both 3,500 and 3,000). Its stated goal is 1 million weekly rides by the end of 2026. Tesla’s disclosed figure is miles, not rides or paid trips, so the two cannot be compared directly. Waymo’s lead is in paid driverless operations at scale; Tesla’s potential edge is a cheaper vehicle and an installed base of 1.74 million FSD subscribers.

Unitree

Unitree filed for a listing on Shanghai’s STAR Market in March 2026. Its prospectus showed 2025 revenue of CNY 1.7 billion (about $246 million), up 335%, and net profit of CNY 600 million. Unitree’s G1 humanoid starts at CNY 85,000, and its R1 Air at CNY 29,900. The company reported a humanoid unit cost of CNY 62,200 in the first three quarters of 2025. Unitree says comparable overseas products cost $30,000 to over $130,000. This is the most relevant external cost benchmark: a real, profitable, volume producer of smaller humanoids. At the exchange rate implied by Yicai’s own conversion of its revenue, CNY 62,200 is about $9,000 (our calculation), below Tesla’s stated cost target, though the robots differ in size and capability.

Figure AI

Figure raised more than $1 billion at a $39 billion post-money valuation in September 2025 and set an initial manufacturing target of 12,000 robots a year at its BotQ factory. The report named no customers. Figure has no public financial statements, so its output and revenue cannot be checked against Tesla’s.

BYD

In vehicles, BYD sold 2,254,714 battery electric vehicles in 2025, up 27.9%, against Tesla’s 1,636,129 deliveries, and 4,550,036 vehicles including plug-in hybrids. Tesla cites lower average selling prices as a drag on revenue, and this core business funds the physical AI spending.

Business Model

Tesla earns money four ways: selling vehicles, selling or leasing energy storage and solar, selling services and software, and selling regulatory credits to other automakers. It also receives interest income on cash. In Q3 2026 automotive sales were $20.09 billion, regulatory credits $113 million, leasing $353 million, energy $2.88 billion and services $4.61 billion.

Profit is more concentrated than revenue is. Total gross profit was $5.49 billion, of which energy contributed roughly $954 million (revenue of $2.88 billion less cost of revenue of $1.93 billion), or about a third of that segment’s revenue, and services contributed a record $734 million. Those two segments together supplied about 31% of gross profit on 27% of revenue (our calculation). Automotive gross margin was 18.5%, or 18.1% excluding regulatory credits.

Regulatory credits are a shrinking support. They were $417 million a year ago, and in the 10-Q Tesla reports it reversed $475 million of credit revenue for contractual return of previously delivered credits after regulatory changes.

The model Tesla wants investors to underwrite is different. The Q3 update says that over time it expects hardware profits to be accompanied by an acceleration of SI, software and fleet-based profits (SI is Tesla’s term for super intelligence, its AI platform). In plain terms, vehicles become fleet assets that earn per mile, robots earn per hour, and the same trained models serve both. That model has no reported revenue yet beyond FSD subscriptions.

Financials

Annual results. Revenue was $94.83 billion in 2025, down from $97.69 billion in 2024 and $96.77 billion in 2023. Gross margin was 18.0%. Operating income fell from $8.89 billion in 2023 to $7.08 billion in 2024 and $4.36 billion in 2025. Net income attributable to common stockholders was $3.79 billion, or $1.08 per diluted share. Operating cash flow was $14.75 billion. Energy revenue grew 27% to $12.77 billion at a 29.8% gross margin.

Q3 2026, the latest reported quarter (ended September 30, 2026). Revenue was $28.05 billion, against $28.1 billion in Q3 2025. Gross margin was 19.6%, up from 18.0% a year earlier. Operating expenses rose to $4.78 billion from $3.43 billion, led by R&D of $2.76 billion against $1.63 billion. Operating income was $718 million, a 2.6% margin, versus $1.62 billion a year earlier. GAAP net income attributable to common stockholders was $787 million, or $0.22 per diluted share; non-GAAP net income was $1.70 billion, or $0.48 per share. Stock-based compensation was $1.17 billion, up from $663 million, a rise that Electrek reports includes the 2025 CEO award.

Nine-month figures show the same pattern. Revenue was $78.67 billion against $69.93 billion, gross profit rose to $14.97 billion from $12.09 billion, and R&D rose to $7.08 billion from $4.63 billion. Gross profit grew faster than revenue, then R&D absorbed the gain.

Cash and capital spending. Operating cash flow was $6.26 billion and capital expenditure $4.73 billion, leaving free cash flow of $1.53 billion. Q2 free cash flow was negative $1.09 billion on $5.79 billion of capex. Trailing four quarter free cash flow was about $3.3 billion, and trailing operating income about $3.5 billion (both our sums of reported quarters). The 10-Q expects 2026 capital expenditure in excess of $20 billion, while Taneja described the figure on the call as approximately $20 billion, down from the more than $25 billion that Electrek reports as earlier guidance. We use the 10-Q wording. Nine-month capex was about $13.0 billion by our sum of the three quarters, which implies roughly $7 billion or more in Q4.

Balance sheet. Cash, equivalents and short-term investments were $45.45 billion. Total debt, excluding finance leases, was $9.28 billion, nearly all non-recourse, and in September 2026 Tesla signed three credit agreements with $30 billion of aggregate commitments. On the call, Taneja said Tesla does not plan to draw on the new facilities in this calendar year. Net property, plant and equipment rose to $50.6 billion from $40.6 billion at year end 2025. Stockholders’ equity was $89.3 billion. Non-operating items add noise to GAAP profit: a $142 million Q3 loss on the SpaceX stake and digital assets (bitcoin) carried at $962 million.

Valuation

All figures are as of October 30, 2026. TSLA closed at $399.81 and the 10-Q cover reports 3,952,757,946 shares outstanding as of October 15, 2026. Multiplying gives a market capitalization of about $1.58 trillion (our calculation). The stock closed between $372.22 and $407.46 over the eight sessions from October 21 to October 30.

Trailing twelve month revenue, the sum of the last four reported quarters of $24.90 billion, $22.39 billion, $28.24 billion and $28.05 billion, is about $103.6 billion. Tesla notes this is its second consecutive quarter of trailing revenue above $100 billion. The market cap is therefore about 15 times trailing revenue and about 17 times 2025 revenue. Trailing GAAP diluted EPS is about $0.91 (sum of quarterly figures of $0.24, $0.13, $0.32 and $0.22), so the stock trades at roughly 440 times trailing GAAP earnings, or about 230 times the trailing $1.72 of non-GAAP earnings. On trailing operating income of about $3.5 billion ($3.47 billion by our sum), the multiple is roughly 456 times.

These multiples have no peer inside the automotive group, so they carry little information on their own. The more useful reading is what they imply: the price assigns most of its value to businesses that currently report little or no revenue. For private comparison, Figure AI’s September 2025 round valued a pre-revenue humanoid developer at $39 billion. Unitree, a humanoid seller with volume sales, reported 2025 net profit of CNY 600 million. We do not publish a fair value. The factors that would change our reading are the disclosed ones: a reported Optimus unit cost, Robotaxi revenue, and free cash flow after the capex step-up.

Key Opportunities

A disclosed Optimus cost curve, first used inside Tesla

If Tesla starts building robots at volume and publishes a build cost, the cost question becomes answerable. The company has a unique internal customer in its own plants. It has also decommissioned an existing vehicle line to host Optimus, a design choice that signals commitment. The Fremont line is designed for 1 million robots a year, and management said Taneja said Optimus costs are trending closer to expected Cybercab costs, though it is a little early to say they have arrived. The labor arithmetic above shows the upside: against compensation of $48.62 an hour, even a robot that performs half the work of a person would have headroom at a price several times the target.

FSD and Robotaxi as recurring software revenue

FSD subscriptions grew from 1.04 million to 1.74 million in a year, and Taneja said an analyst estimate of recurring revenue near $1 billion was correct and promised more FSD metrics. Software revenue has a very different margin structure from vehicles. Musk said he expects EU-wide approval within a few months and would be surprised if it took beyond Q1 2027. Approvals in nine countries are already in hand. Customer vehicles joining the Robotaxi network, which Tesla targets around Q2 2027, would be the larger prize.

Storage demand tied to AI power needs

Energy gross margin reached 33.1% in Q3, up from 20.4%, helped by tariff refunds and warranty true-downs per Taneja, and Tesla reports its highest ever order rate and backlog. Capacity is being added: 50 GWh at Texas, with Shanghai and California capacity of 20 GWh and 40 GWh. If data center builders keep buying Megapack, this segment funds the rest of the plan with a business that already earns a gross margin of about a third.

Key Risks

Optimus may be late, costlier or less capable than targeted

Tesla has not disclosed an Optimus unit count, build cost or customer. The Gen 3 unveil was planned for Q1 2026 and we found no report of it in the Q3 update. Management said the ramp to 20,000 robots a week will take at least a year and that remaining AI, electromechanical and reliability issues should be resolved around the end of 2027. The CEO award milestone of 1 million bots delivered sits far from today’s early ramp status. The 10-K itself warns that Bots operate in a nascent industry that has yet to develop commercially.

Thin operating profit funds a steep capex step-up

Operating margin was 2.6% in Q3, with trailing operating income of about $3.5 billion against 2026 capex of more than $20 billion. The gap is covered by cash, but the cash is finite and the $30 billion of credit commitments are described in the 10-Q as added flexibility for planned investments. Musk estimated Terafab spending at tens of billions through early 2028. Stock-based compensation of $1.17 billion in a quarter dilutes existing holders.

Autonomy faces regulatory and legal exposure

The 10-Q discloses a securities class action alleging misstatements about Autopilot, FSD and Robotaxi for stock purchases made between April 19, 2023 and June 22, 2025. It also describes the Benavides verdict, in which a jury awarded $129 million in compensatory damages and $200 million in punitive damages over a 2019 Autopilot accident, with Tesla assigned 33% of fault. Federal regulators have also reviewed Robotaxi behavior; NHTSA opened defect petition file DP26010 on October 9, 2026, alleging school zone speeding, and has until roughly January 21, 2027 to decide whether to grant it and open a formal investigation. Customer cars joining the Robotaxi network, targeted for Q2 2027, and EU-wide approval both depend on regulators.

Core automotive demand and credit support are weakening

Deliveries fell 2% year over year in Q3 and 9% for 2025. BYD sold more battery electric vehicles than Tesla in 2025, by about 618,600. Regulatory credit revenue has fallen from $417 million to $113 million in a year, and the company attributes the supply limit to batteries, so demand strength does not automatically convert to profit. The business that must fund physical AI is not growing.

Related party and key person concentration

Tesla sold $626 million of Megapacks to SpaceX in nine months, owns a stake in that company, and shares its CEO with it, which is why the 10-Q presumes significant influence. Strategy, compensation and capital allocation are tied to one executive, whose award is denominated in physical AI deliverables. Related party transactions and a single decision maker raise governance risk that a diversified board and independent pricing would otherwise mitigate.

What Would Prove Us Wrong

Our working view is that Optimus hardware cost is probably not the binding constraint, that Tesla’s physical AI claims remain unproven in its financials, and that the burden of proof sits with disclosures. These signposts would show the view is wrong in one direction or the other.

1. Optimus ships to outside customers on time. If Tesla reports paid Optimus deliveries to external customers of at least 1,000 units by the end of 2027, Musk’s end-2027 expectation was credible and our delay framing is wrong. If external deliveries are below 1,000 units at the Q4 2027 report, the schedule has failed by its own measure.
2. A disclosed unit cost lands at or below the target range. If Tesla’s filings or updates through the Q4 2027 report state an Optimus build cost at or below $25,000 at any volume, the unit cost question is answered in the bull direction. If the disclosed cost remains above $50,000 at the Q4 2027 report, or Tesla still declines to disclose it, the labor substitution argument stays unproven. These thresholds are ours, set against Musk’s stated January 2025 target of a price under $20,000 at one million units a year.
3. Robotaxi scale shows in the numbers. Cumulative unsupervised Robotaxi miles are nearly 2 million at Q3 2026. If the figure is below 20 million in the Q1 2027 update, the growth rate has stalled. Tesla’s VP of AI software also said the first customer vehicles could join the Robotaxi network around Q2 2027; if that has not happened by the Q3 2027 update, the autonomy timeline has slipped again.
4. The core business funds the plan. If trailing twelve month GAAP operating income in the Q2 2027 update is below the $4.36 billion earned in 2025, while 2026 capex finishes above $20 billion, then the investment is outrunning the profit base. A reading above that level with positive free cash flow would argue the opposite.

Summary

Tesla’s reported business in October 2026 is a $28 billion a quarter revenue company with flat sales, a 2.6% operating margin and rising investment. Automotive supplies about 73% of revenue. Energy storage and services provide a disproportionate share of gross profit and are the healthiest segments. Physical AI shows up in filings as expense (R&D of $2.76 billion in the quarter), capacity (Cortex, Fremont, Austin) and a fast-growing but still modest software line. It does not yet show up as segment revenue.

If Tesla reaches Musk’s stated cost target, humanoid labor would be far cheaper than a manufacturing worker on hardware cost alone, and the BLS wage of $48.62 an hour makes the bar low. Open items are whether Tesla can build at that cost, whether the robot does enough useful work per hour, and whether it runs reliably enough to deploy. The only independent cost datapoint, Unitree’s CNY 62,200 humanoid unit cost, belongs to a smaller, different robot.

The market capitalization of about $1.58 trillion at the October 30, 2026 close is a bet on disclosures that have not arrived. The signposts above, dated through 2027, would move this assessment: external Optimus deliveries, a stated unit cost, Robotaxi mileage, and operating income against capex.