Tesla Optimus: A Full Guide to the Fremont Line, Gen 3, Optimus Academy, AI5, Price and Limits

Tesla says Optimus is finally in production. It still hasn't said how many robots it has built, what they can do on their own, or when you can buy one.

That gap is the whole story of Tesla Optimus at the end of 2026. Tesla's Q3 2026 update says production began in Fremont on high-volume lines, and Elon Musk told investors the company could make 20,000 robots a week by the end of 2027. In the same call he admitted the hands, durability and AI aren't finished.

This full guide covers where Optimus came from, what Tesla has actually disclosed about the Gen 3 design, the Fremont and Texas lines, Optimus Academy, the AI5 chip, price and leasing, the 1 million robot goal in Musk's pay package, how Optimus stacks up against Figure and the rest, and the honest limits. Last updated: November 11, 2026.

Key Takeaways

• Tesla's Q3 2026 update says Optimus production began in Fremont, listed as Early Ramp, with no unit count disclosed.
• The Fremont line is designed for 1 million robots a year; a Texas line targets 10 million a year long term.
• Musk targets 20,000 robots a week by end of 2027 and first outside deliveries around the same time.
• Tesla says it needs 10,000 to 20,000 robots in its internal Optimus Academy to generate training data first.
• Musk's 2025 pay award includes a 1 million Bots delivered milestone, so robot volume is now a contractual target.

Tesla Optimus at a glance

Let's start with what's on the record. Every row below comes from Tesla's SEC-filed shareholder updates, its earnings call, or reporting on that call. Where Tesla hasn't disclosed something, the table says so.

Source: Tesla Q3 2026 update and Q1 2026 update; TechCrunch and The Verge on the Q3 2026 call; TechCrunch on the price range Musk gave in October 2024.

Notice how many rows are targets rather than results. That's the right lens for everything that follows.

From a dancer in a suit to Gen 3: the Optimus timeline

Optimus started as a slide and a person in a robot costume at Tesla's 2021 AI Day. The first working prototype came a year later, and the program has moved in roughly annual steps since.

Source: Electrek on AI Day 2022; Electrek on Gen 2; TechCrunch on We, Robot; TechCrunch on Kovac; Tesla Q4 2025, Q2 2026 and Q3 2026 updates.

The We, Robot row matters more than it looks. TechCrunch, citing Bloomberg, reported that the robots serving drinks and chatting with guests were controlled by humans for much of the night, while walking was handled by the robot's own AI. Tesla didn't disclose that on stage.

Here's why that matters. Every Optimus video since then gets judged against that night, and Musk himself now promises footage of tasks that aren't preprogrammed.

What Tesla has actually said about Optimus Gen 3

The public calls it Optimus V3. Tesla's filings call it Gen 3, and they say very little about it.

In its Q4 2025 update, Tesla said it planned to unveil Gen 3 in the first quarter of 2026, with major upgrades over version 2.5, including its latest hand design. It called Gen 3 its first design meant for mass production and said it planned an eventual capacity of 1 million robots a year.

That unveiling never got a follow-up in the written record. The Q1, Q2 and Q3 2026 updates don't mention it, and Electrek noted in late September that the promised V3 reveal by mid-2026 still hadn't happened.

So what are the specs? There aren't any official ones. Tesla hasn't published Gen 3 height, weight, payload, battery life, degrees of freedom or runtime.

The Fremont line: how Optimus production actually started

Killing the Model S and Model X to make room

Tesla gave up two of its most famous cars for this robot. Its Q1 2026 update said the first-generation Optimus line, designed for 1 million robots a year, would replace the Model S and Model X lines in Fremont.

By July, the Q2 2026 update said those car lines had been decommissioned and Optimus lines were being installed, with production expected soon. The first builds were earmarked for internal use, not customers.

What production-quality means, and what it doesn't

The Q3 2026 update is careful. It says the bots coming off the Fremont line are the first production-quality bots Tesla has manufactured, and that the current process is suitable for a commercial product, while hardware and process iteration continues.

It doesn't give a production number. The capacity table lists California Optimus as Early Ramp and Texas Optimus as Construction, under a standard warning that installed capacity isn't the current production rate.

Reporting fills in some of the gap. Electrek, summarizing The Information, said Tesla went from a few dozen robots a week in Q2 to several hundred a week in August, with managers aiming for more than 1,000 a week by year-end. Tesla hasn't confirmed those figures.

Musk's own caveats

On the Q3 call, Musk called it early production with a ramp that will be slow and somewhat bumpy, according to TechCrunch. He said the robot can't wear out after a month and that there's still a lot of AI work before it's useful for open-ended tasks.

He also said reaching 20,000 a week would take at least a year, maybe a little longer, because parts of the design are still changing. That's a frank admission from a CEO who usually leads with the upside.

Texas and the 10 million robot plan

Fremont is the first act. Tesla's Q1 2026 update said Gigafactory Texas is being prepared for a second-generation Optimus line designed for long-term annual capacity of 10 million robots.

The Q2 2026 update said building construction in Texas was in full swing, and the Q3 table still lists Texas Optimus as Construction. No start date has been given.

To put 10 million in context: that's more than Tesla's entire installed car capacity. The Q3 2026 update lists Model 3 and Model Y capacity of more than 550,000 a year in California and more than 950,000 in Shanghai. A 10 million robot line would be a different scale of factory altogether.

Optimus Academy and the training data problem

Tesla's cars have one advantage Optimus doesn't: millions of vehicles collecting driving data every day. Optimus starts with no fleet.

Tesla's answer is Optimus Academy. The Q2 2026 update said the initial builds would be used there for training data collection and further functionality development.

On the Q3 call, Musk said Tesla needs roughly 10,000 to 20,000 robots practicing tasks at the Academy, learning through self-play before customers get them. He compared the process to a toddler gradually learning harder skills.

Electrek's summary of The Information adds detail: Tesla reportedly has more than 500,000 hours of training data and wants to double it by year-end, using data collectors in camera helmets and motion-capture suits. It also reported that Optimus still takes several days to learn basic tasks.

That means the Academy isn't a side project. It's the bottleneck that decides whether the factory output is useful or just inventory.

Hands, actuators and the supply chain

Hands are where humanoids live or die, and Tesla is shipping before they're done. On the Q3 call Musk said, in The Verge's report, that Tesla is building robots that don't have perfected hands.

Electrek's reporting describes why. The hand and forearm reportedly contain more than 100 screws and small components still assembled by hand, some touch sensors have had reliability problems, and Tesla plans a replaceable sensing glove next year so it doesn't have to swap whole hands.

Suppliers are the other pressure point. The same report says Tesla relies on outside suppliers, many in China, for motors and precision gears, and that some struggle to hold quality at volume. Inside Deep Tech's guide to robot dexterity and manipulation explains why fingers, tactile sensing and force control are the hardest part of the stack.

AI5 and the Optimus brain

Optimus runs on Tesla's own inference hardware and the same vision-first approach as its cars. The next step is a new chip.

Musk said he hopes to put Tesla's AI5 chip into Optimus around April 2027, with roughly 10 times the computing power of AI4. He expects AI5, better software and mechanical fixes to come together before customer deliveries.

He also promised a video within a month or two showing Optimus doing tasks that aren't preprogrammed. If it arrives, look for unedited runs, failure rates and whether anyone is teleoperating off-camera.

Price, leasing and who can get an Optimus

Tesla hasn't announced a price. The numbers in circulation are old guidance: Musk said at AI Day 2022 that Optimus would cost less than $20,000 at volume, and at We, Robot in 2024 the figure was around $20,000 to $30,000.

You can't buy one today. Musk said on the Q3 2026 call that Tesla could start delivering Optimus to outside companies around the end of 2027, starting with industrial work and moving to homes later.

The first commercial model may not be a sale at all. Electrek reported that Tesla intends to lease, not sell, to a short list of companies whose factories and warehouses look like its own, then use those deployments to improve the AI.

If you're budgeting for humanoids in 2027, plan around robots you can actually contract for, and treat Optimus as an option to revisit once Tesla publishes terms.

The 1 million robot milestone in Musk's pay package

Optimus volume is now written into Tesla's governance. Shareholders approved the 2025 CEO Performance Award on November 6, 2025.

The award agreement lists 12 operational milestones, and one of them is 1 Million Bots Delivered. The agreement defines a Bot as an AI-driven mobile robot or physical product made by or for Tesla, including Optimus, and excludes vehicles. Each tranche also needs a market-cap milestone.

On the flip side, that milestone doesn't measure usefulness. A delivered robot counts whether it does one scripted task or a hundred general ones.

How Optimus compares with Figure, Atlas, Digit, Apollo and Unitree

Tesla's edge is manufacturing scale and capital. Its rivals mostly have a head start on real deployments. Inside Deep Tech's humanoid robots full guide covers each one in depth; here's the short version.

Source: Inside Deep Tech's humanoid robots full guide, Brett Adcock and Figure AI guide and Tesla Optimus vs Figure AI comparison.

Humanoid Robots in 2026: A Full Guide to Figure, Optimus, Atlas, Digit, Unitree, and Apollo

Inside Deep Tech's evidence scorecard of every major humanoid robot program: deployments, money, verification and what to watch.

Inside Deep Tech

Chinese competition is also moving. Electrek noted that XPeng started running an automated IRON humanoid production line in Guangzhou in September 2026, targeting commercial sales in 2027.

What Optimus costs Tesla

Tesla doesn't break out Optimus spending, but the company's numbers show the strain of its AI and robotics push. Capital expenditures were $4,732 million in Q3 2026, up from $2,248 million a year earlier, per Tesla's own financial summary.

Net income fell 43% to $787 million in the quarter, and TechCrunch attributed part of the drop to spending on AI projects and Optimus. That capex figure covers the whole company, so don't read it as an Optimus budget.

The point is simpler. Tesla is funding the robot from a car business with flat revenue, which raises the cost of every quarter the Academy needs before Optimus earns money.

Inside Deep Tech's take

Honest limits of what's known about Optimus

• No unit count. Tesla's Q3 2026 update confirms production but gives no output figure. Third-party weekly numbers are unconfirmed reports.
• No Gen 3 spec sheet. Height, weight, payload, battery life and degrees of freedom haven't been published.
• No autonomy metrics. Tesla hasn't disclosed task success rates, intervention rates or how often robots are teleoperated.
• No price or terms. Tesla hasn't announced a price, lease terms, warranty, safety certification or support model.
• Call targets aren't filings. The 20,000 a week, end-2027 delivery and AI5 dates come from the earnings call, not the written update.

FAQ

What to watch next

• Tesla's Q4 2026 update in January 2027: any unit count, any change from Early Ramp, and any Texas start date versus the Q3 2026 table.
• The non-preprogrammed task video Musk promised, and whether it shows failures and full uncut runs.
• AI5 integration around April 2027, and whether the first external leases get named customers.
• Rival milestones from Figure, Agility and Apptronik, tracked in Inside Deep Tech's Top 100 Physical Robotics Companies directory.
• Any update on the Gen 3 hand, the replaceable sensing glove and supplier quality at volume.

Optimus has crossed from prototype to production line. The next 12 months decide whether it crosses from production line to product. Watch the Academy, not the factory: the first robot doing paid work outside Tesla is the milestone that counts.

Cover background: photo of Tesla Optimus robots by Ulkl, public domain via Wikimedia Commons.