Elon's Tsunami of AI Robots: Why Optimus Scale Leaves Humanoid Startups Behind

Jason Calacanis walked into Tesla's Optimus lab on a Sunday morning with Elon Musk. People were already there at 10 a.m. He saw Optimus 3. Then he went public with a take that sounds unhinged until you sit with the numbers: in ten years, nobody will remember that Tesla ever made a car. They will only remember Optimus — and that Elon plans to make a billion of them. When that post hit X, Elon answered with two words: probably true.

I have been a Tesla investor since 2012 and worked there from 2017 to 2021. That reply is not casual hype. It is Elon agreeing that the company that put millions of cars on the road with FSD, that builds giant battery systems, that remade the auto industry, could become known for humanoids first. This channel exists to take claims that sound insane and show when the data says they might be real. This is one of those claims. The gap between Tesla and everyone else in humanoid robotics is not closing. Every quarter it widens. And the market underneath it is nearly $40 trillion a year — the value of physical human labor worldwide, larger than U.S. GDP.

The production targets nobody is pricing

Tesla has said it plans to manufacture 50,000 Optimus units in 2026. It is building factory capacity for one million units per year by late 2026. That does not guarantee a million robots shipped that year. It means the manufacturing system is being stood up to hit that cadence soon after. Separately, Tesla has announced a dedicated humanoid line at Giga Texas aimed at at least ten million Optimus robots a year by 2027 — more robots annually than people in New York City.

Now put the well-funded competitors next to that. Figure AI has raised about $1.9 billion at a $39 billion valuation. Its BOTQ facility targets roughly 12,000 units per year, with a goal of 100,000 total over four years. Tesla's one-year target of 50,000 is almost half of Figure's entire four-year stack. By the time Figure hits 100,000 cumulative, Tesla — if it hits plan — could make that many every few weeks.

Boston Dynamics has been building robots for decades. Atlas parkour. Spot dancing. The demos are world-class. Their near-term Atlas production is committed to Hyundai, their parent. They project roughly 30,000 units per year by the end of 2028. When Boston Dynamics gets to 30,000 a year, Tesla's stated path puts that quantity in a few days of output. Boutique watchmaker versus Casio at millions of units — same product category, different business.

Why competitors cannot just "raise and catch up"

If Figure has nearly two billion dollars, why not build bigger factories? The short answer is vertical integration. Since Tesla started building cars in 2008, it has built one of the most vertically integrated manufacturing operations in modern industry. Battery cells, motors, gearboxes, the specialized chips that run the AI — for the critical pieces, Tesla makes them in-house. When Optimus needs a component, Tesla does not wait for a multi-year supplier contract or convince a third party that humanoids are a real market. It builds the capacity.

A startup sourcing motors, gearboxes, batteries, sensors, and Nvidia compute can assemble prototypes and even a few hundred units. At 50,000 or a million units, every supplier that cannot scale stops the line. Tesla already lived through that with cars. It also sits on more than $40 billion in cash and a business that throws off cash — and can raise in public markets. That financing position is not available to a venture-backed robotics company learning manufacturing hell for the first time.

Then there is silicon. Tesla's AI5 chip is being developed specifically for Optimus — purpose-built for the compute load of humanoid robotics by the same people who built the FSD chip. The chip team talks to the AI team talks to the motor team. They co-optimize the whole stack. Most startups run on Nvidia hardware. Nvidia makes excellent chips. It also sets product timelines and prices for data-center priority first. If Nvidia prioritizes servers over robots, the startup waits. If prices rise, costs rise. Tesla iterates chip architecture against its own software and mechanics. Matching that means excelling at mechanical engineering, AI, power electronics, batteries, precision manufacturing, and supply chain at tens of thousands of units — with billions in capital and years of runway. Even with perfect execution, that is most of a decade. Perfect execution never happens.

Twenty thousand dollars and recursive manufacturing

Elon has said Optimus could cost about $20,000 to $25,000 per unit once production reaches million-unit scale. Less than a new car. Less than a year of minimum-wage pay in most developed countries. Tesla's factories already produce millions of vehicles a year with motors, batteries, sensors, compute, and precision assemblies. The same class of cells that power a car can power a robot. The same class of cameras that see the road can see a factory floor or a home. In important ways, Optimus is a simpler manufacturing problem than a Model Y. A car has to survive crashes, extreme heat and cold, and hundreds of thousands of miles. A factory robot in a controlled environment has hard problems of its own — but different ones, not impossible ones.

Boston Dynamics' edge is decades of robotics research. Their edge is not scaled manufacturing. Going from dozens or hundreds of robots a year to tens of thousands is a different company. Custom cars in a garage are not two million cars a year. Tesla is one of the few organizations that has already crossed that bridge with complex electromechanical products — Model 3, Model Y, Tesla Energy. Boston Dynamics' near-term Atlas capacity is spoken for by Hyundai. Those units go into Hyundai factories, not the open market. Venture-backed peers are still learning lessons Tesla paid for during production hell on the Roadster and Model 3. In software you push a patch. In hardware a design flaw can force a retool, a recall, or scrap. You cannot iterate past a supply-chain bottleneck with a blog post.

Here is the advantage most people underweight: recursive manufacturing. Optimus units will eventually help build more Optimus units. Deploy a few hundred inside Tesla factories and two things happen. Labor cost falls on those tasks. More important, every hour of real operation — a box heavier than expected, a forklift in the aisle, a spill on the floor — becomes training data. That data improves the models for the whole fleet. Every robot makes every other robot smarter. Competitors can run pilots. They cannot deploy thousands of units into real plants early because they cannot build thousands of units early. Chicken and egg. Tesla already has the plants.

The data moat goes further. Tesla has billions of miles of real-world driving video. That is physical-world navigation data at a scale no robotics startup has. Architectures that learn to see pedestrians and predict motion transfer to objects on a shelf and warehouse aisles. When Tesla eventually sells Optimus outside — warehouses, manufacturers, whoever — those units can arrive with hundreds of thousands or millions of hours of operational experience across task categories. Competitors' first commercial units will still be discovering maintenance intervals and failure modes Tesla already burned through.

What catch-up actually costs

To catch Tesla in humanoids, a rival needs something like $10 to $20 billion — not two, not five — for custom chips, supply chains, and factories. Multiple high-volume precision plants. Expertise across mechanical, electrical, AI, power electronics, batteries, and supply chain. Components that do not exist in volume because nobody has built humanoids at scale yet. Thousands of engineers in a talent pool that is already thin, with many of the best people already at Tesla. And they have to do it while Tesla extends the lead every quarter. Even Apple or Google starting tomorrow would be near zero on human-scale robots with dozens of degrees of freedom. Tesla has been sitting at the intersection of AI, manufacturing, and precision engineering for more than fifteen years.

None of this guarantees Optimus is profitable next year. Robotics is hard. Timelines will slip. There will be a production hell 2.0. Market share will not be one hundred percent. Failed companies litter the path from demo to scaled deployment. But if you ask who has the highest probability of deploying general-purpose humanoids at real volume, the answer is not close. The demos — coffee pouring, backflips, laundry folding — are not where the fight is. The fight is in supply chains, factory layout, procurement, and component design. In that contest, Tesla looks years ahead and still pulling away.

One percent of a $40 trillion global labor market is $400 billion a year. Five percent is $2 trillion. Manufacturing, warehousing, agriculture, construction, healthcare — hundreds of millions of people doing pick-and-place, move, assemble, clean, sort. Optimus is not a side project next to cars. It is the long game. Jason Calacanis might be right. In a decade, people may remember Tesla as the company that automated physical work and made services cheap — not as a car brand. Watch the factories, not the viral clips.

Check the video here.