How Tesla could become the world's number one company

Nathan Chadwick · August 25, 2026 · 19 min read

TLDRRead the short version

Tesla is still valued like a car company. The public master plans were never about cars as the end state. They were about energy, then autonomy, then physical AI.
The 2006 plan was executed almost to the letter: Roadster, then Model S and X, then Model 3 and Y, with solar and storage running in parallel.
Part Deux (2016) already named Robotaxi. The software is now giving unsupervised rides in the US. Europe has only approved supervised Level 2, which is why the real growth is still bottled up.
Optimus is the largest TAM Tesla has ever pointed at. Musk talks about more than $10 trillion of revenue and roughly 80% of Tesla's value. Independent research is smaller and later, and production is behind, but even a slice of global labour is civilisation-scale.
SpaceX has done reusable launch, Starlink, an xAI combination, and the largest IPO on record. A Tesla merger is not announced. The industrial overlap is already real, and the combination is how you get a plausible number one.
Not financial advice. I have a Tesla referral on this site.

Not a car thesis#

People still argue about Tesla as if the question is "will they sell more cars than Toyota." That was never the bet, and it has not been the bet since Elon Musk wrote it down on 2 August 2006.

The question that actually matters in 2026 is whether one company can own the stack that turns energy, mobility, and labour into software. Cars are the current cash engine and the rolling sensor fleet. They are not the ceiling.

I think Tesla can still become the most valuable company in the world. I do not think that is guaranteed, and I do not think it happens because the Model Y has a good quarter. It happens if the master plans keep landing in the same direction they have for twenty years, and if autonomy and humanoid robots compound on top of a real energy business.

This is not financial advice. I like the company, I have driven the product, and there is a Tesla referral on the home page. Read this as an argument, not a buy ticket.

The 2006 plan actually happened#

The original post is short, and it is still the cleanest document Tesla has ever published. The point was not "build a sports car." The point was to use a sports car to pay for the thing that mattered.

Musk's four steps, in his own words:

Build sports car
Use that money to build an affordable car
Use that money to build an even more affordable car
While doing above, also provide zero emission electric power generation options

That is the Roadster, then the Model S (and later the Model X), then the Model 3 and Model Y, with SolarCity and then Tesla Energy running alongside. He even said the quiet part in 2006: when someone bought the Roadster, they were helping pay for the low cost family car.

At the time this sounded like a blog post from a man who had not yet shipped a production vehicle. The company was burning cash, the Roadster was late, and the idea that an EV startup would walk down-market into volume looked like a way to go bankrupt with extra steps. Lotus chassis, Panasonic cells, a factory that was not a factory yet. The plan was treated as marketing.

Then it shipped. Not on the original calendar. On the original sequence.

That distinction matters for everything that follows. Tesla's critics are usually right about timing and usually wrong about direction. The 2006 plan is the template: write the steps in public, miss the dates, keep the order, end up with the thing you said you would build.

Part Deux: energy, every segment, and the car that earns while you sleep#

Master Plan, Part Deux landed on 20 July 2016. Four more steps, again in public:

Create stunning solar roofs with seamlessly integrated battery storage
Expand the electric vehicle product line to address all major segments
Develop a self-driving capability that is 10X safer than manual via massive fleet learning
Enable your car to make money for you when you aren't using it

Read that last line again. Robotaxi is not a 2024 pivot. It is a 2016 product promise, written while the Model 3 was still a prototype and Autopilot was a highway lane-keep system with a camera and a prayer.

The vehicle-line expansion is the part that is easiest to score. Compact car, crossover, pickup, heavy truck, urban transport. We got the Model 3, the Model Y (now the volume machine), the Cybertruck, the Semi (still ramping), and the Cybercab as the dedicated robotaxi vehicle. Not every name shipped on the year it was teased. The map of segments did.

Energy is the part the car press still treats as a footnote, and it is the part that has been executing. Tesla did acquire SolarCity. Powerwall became a household product. Megapack became a utility product. Full-year 2025 energy storage deployments were 46.7 GWh, with 14.2 GWh in the fourth quarter alone. In Q2 2026 Tesla deployed another 13.5 GWh and put Megapack capacity on the scoreboard at 40 GWh in California, 20 GWh in Shanghai, and a Texas factory in commissioning. That is not a side hustle. That is a company learning to sell electrons and the boxes that hold them.

Autonomy is the part that ran late in a way that damaged trust. Musk talked about unsupervised self-driving on timelines that did not survive contact with reality. Part Deux was more honest than the tweets: software validation would take much longer than putting cameras in the car, and worldwide regulatory approval would need something on the order of 6 billion miles. Tesla already knew, in 2016, that the bottleneck was not only neural nets. It was proof, and it was governments.

The sharing step is Robotaxi. A car that sits in a driveway for 23 hours is a terrible asset. A car that can take itself to work for other people is a different business, with software margins, on a fleet Tesla already sold. That is the whole idea. Europe is why that idea is still smaller than the 2016 blog post.

Part 3: the energy TAM is not a slogan#

Master Plan Part 3, published 5 April 2023, is the least memed of the four and the most important for anyone who still thinks Tesla is "just" an automaker. It is a paper, not a product launch. The claim is that a sustainable global energy economy is technically feasible through electrification plus generation plus storage, and that it takes less material extraction than continuing to burn fossils.

The numbers in that paper are civilisation-scale: on the order of 30 TW of renewable generation and 240 TWh of storage, with manufacturing investment around $10 trillion. Tesla is not going to build all of that. The point of publishing it was to show that the market Tesla is walking into, batteries and inverters and pack factories, is not a car-adjacent hobby. It is the grid.

Watch the actual deployments against that backdrop. Tesla storage went from 14.7 GWh in 2023 to 31.4 GWh in 2024 to 46.7 GWh in 2025. The slope is the story. Vehicle deliveries, by contrast, peaked at 1,808,581 in 2023, slipped to 1,789,226 in 2024, then to 1,636,129 in 2025. Q1 2026 was 358,023 deliveries. Q2 recovered to 480,126. If you only look at cars, Tesla looks like a maturing auto stock under price pressure from China. If you look at storage, it looks like a company whose second act is already printing volume.

A number-one company does not need to win every car in the world. It needs a cash engine plus a larger TAM. Energy is the first of those larger TAMs that you can already see in the accounts.

Part IV: physical AI is the actual ceiling#

Master Plan Part IV was published on 1 September 2025. The language shifts from sustainable energy to sustainable abundance. The products named are the ones that put AI into the physical world: autonomous vehicles, Optimus, the factories that make both.

This is the document that made a lot of people roll their eyes, because it sounds like a TED talk. Read it against the previous three and it is less of a vibe shift than it looks. 2006 was "use a sports car to fund a family car, and sell solar." 2016 was "make the car drive and earn." 2023 was "the grid is the market." 2025 is "once you can move electrons and atoms with software, labour is next."

Tesla is spending like it believes that. Capex for 2026 was described in the mid-year update as around $25 billion, a multiple of the old run-rate, aimed at AI compute, Cybercab, Semi, Megapack, and the first Optimus lines. That spend is either the down payment on being the most valuable company on earth, or a very expensive way to miss a car cycle. The master plans only help you if you think the spend is the former.

Robotaxi is real. Europe is why it is still small.#

Unsupervised Robotaxi is no longer a slide. Tesla has been running a driverless service in Austin since June 2025, expanded the geofence, and by the Q2 2026 call had unsupervised rides in multiple US metros, including Florida cities, with Tesla's AI lead putting the tally at about 380,000 unsupervised miles and no notable incidents. Cybercab production has started in Texas, with installed capacity listed above 125,000 a year, and an Austin launch event set for 3 September 2026.

That is the bull case in miniature: a vision-only stack, trained on a consumer fleet, now taking people from A to B without a safety driver in the seat. Waymo proved the destination is possible with a different architecture. Tesla is trying to prove it is possible at the cost structure of a Model Y.

Here is the honest part. The live unsupervised fleet is still tiny. Electrek's tracking has had it in the dozens of cars, not the tens of thousands. Volume production language for Cybercab slipped in the Q2 2026 letter. A launch event is not a network. The technology is in the "it works in some cities" phase, which is exactly where every serious autonomy programme has to live before regulators let it eat a country's taxi market.

Europe is the clearest picture of that brake.

On 10 April 2026 the Dutch RDW issued the first European type approval for Tesla FSD Supervised. Eighteen months of testing. Track and public roads. The RDW's own statement is the one that matters: a vehicle with FSD Supervised is not self-driving. It is a driver-controlled assistance system. The driver stays legally responsible and must be able to take over. That is UN Regulation 171, Level 2. Hands can come off the wheel. Eyes cannot come off the job.

Reporting around the approval put the evidence base at roughly 1.6 million kilometres of European road data and more than 400 compliance requirements. Other countries then started recognising the Dutch decision. That is real progress. It is also not Robotaxi.

Full driverless operation, the thing that turns a parked Model Y into an earning asset and a Cybercab into a fleet, sits on a different regulatory track. Europe does not treat that as a software update you push from California. It treats it as vehicle type approval, driver responsibility law, insurance, and city-level permission. Until that stack moves, Tesla can sell supervised FSD in the EU and still be blocked from the product Part Deux actually described.

That is why I keep saying European regulation has held Robotaxi growth back. Not because Tesla has no European customers. Because the business model of autonomy is utilisation, and utilisation needs unsupervised operation in dense, high-fare cities. London, Paris, Berlin, Amsterdam, Milan. Those markets are enormous. They are also exactly the markets where a camera-only robotaxi is a political object, not just a technical one. Every quarter the EU stays on Level 2 is a quarter Waymo, Tesla, and everyone else cannot harvest the European TAM.

The 2016 plan already told you this would be the fight. Six billion miles was never a training-data flex. It was a regulatory forecast.

If unsupervised Robotaxi does get through that wall, the economics are ugly for incumbents in a way car-unit forecasts cannot capture. Ride-hailing today is a labour business with a software skin. Remove the driver, keep the fleet, and the contribution margin starts looking like a network, not a taxi company. ARK and others have put the robotaxi TAM in the many trillions. You do not need their most aggressive 2030 number to be right. You need Tesla to take a serious share of unsupervised miles in the US, China, and eventually Europe. Europe is the piece that is late, and it is late for legal reasons, not because the cars forgot how to steer.

Optimus, and why the TAM argument is not a meme if you sit with it#

Musk has said Optimus could be north of $10 trillion in revenue and that roughly 80% of Tesla's value will eventually be the robot. He has talked about a unit cost under $20,000 to $30,000 at scale, internal factory use first, then selling the surplus to everyone else, including rivals.

Those are founder numbers. Treat them as an upper bound, not a model.

The independent numbers are still large. Morgan Stanley has put the humanoid market around $5 trillion a year by 2050, with on the order of a billion units in use, most of them industrial and commercial rather than "robot butler in every kitchen." Citi has been in a similar zip code. Goldman has been far more conservative, with some coverage still talking about tens of billions by 2035, not trillions. That spread should make you humble. It should not make you shrug.

Here is the arithmetic that I find hard to dismiss. Global labour is the biggest market that exists. A huge fraction of it is still repetitive, dangerous, or dull physical work in environments built for human bodies: factories, warehouses, care, agriculture, logistics. A humanoid is not the only robot that can do those jobs. It is the robot that fits the world we already built, which is why Tesla is bothering with bipeds instead of a fleet of specialised arms. If you can make a general-purpose body, train it on Tesla's real-world video stack, and manufacture it like a car, you are not competing with Boston Dynamics for demo videos. You are competing with wages.

Tesla's edge, if it has one, is not that Optimus walked on a stage. It is the combination of:

a factory culture that already stamps, welds, and assembles at automotive volume
a vision and inference stack already paid for by Autopilot and FSD
a plan to use the first robots to build more robots, which is how unit cost actually falls

The delays are real. Musk talked about thousands of useful Optimus units in Tesla factories on timelines that did not happen. Milan Kovac, who led the programme, left in 2025. Rare-earth export controls hit the motors. In Q2 2026 Tesla dropped the "volume production" language that had been in the prior letter, said the first-generation lines were being installed in Fremont and Texas, and sent early units to an internal academy for training data rather than to customers. Musk told investors it would be the hardest product Tesla has ever scaled, because there is no existing supply chain and every part is new.

That is not a hidden bear case. That is the Q2 letter. I am not asking you to ignore it. I am asking you to put it next to 2006. The Roadster was also the hardest car they had ever scaled, because there was no existing supply chain and every part was new. The question is whether the company that learned to make cars and batteries can learn to make bodies.

If Optimus only ever becomes a useful factory tool inside Tesla, that is still a margin story. If it becomes a product other companies buy, the TAM stops looking like "robotics" and starts looking like "labour as a service." You do not need Musk's $10 trillion. You need Tesla to take a single-digit share of a market Morgan Stanley already puts in the trillions by mid-century. At car-company multiples that is enough to argue about number one. At software-and-fleet multiples it is enough to win the argument.

SpaceX is the other half of the stack#

Tesla without SpaceX is already a strange company: cars, grid batteries, robots, a training cluster measured in buildings, a Supercharger network other manufacturers now use. Tesla with SpaceX is something the market does not have a template for.

Look at what SpaceX has actually done, not the Mars posters.

Reusable Falcon 9 turned launch into a manufacturing problem. Commercial Crew put NASA astronauts on American rockets again. Starlink turned a constellation into a cash-flow business, with reporting around the IPO putting subscribers above 10 million and connectivity as a real segment rather than a science project. Starship is the heavy-lift bet that either opens the cost curve for mass to orbit or eats years of capital. In February 2026 SpaceX combined with xAI, folding Grok, X, and the Memphis compute build into the same house. On 12 June 2026 the combined company listed on Nasdaq as SPCX, sold hundreds of millions of Class A shares, and raised on the order of $85.7 billion with the overallotment. That is the largest IPO on record, at a pricing valuation around $1.8 trillion.

None of that is Tesla's revenue. All of it is Tesla's industrial neighbourhood.

The overlap is already in the products. Grok runs in Tesla vehicles. Cybercab is expected to lean on Starlink for connectivity. Tesla's Q2 2026 accounts showed a SpaceX equity investment unrealised gain. SpaceX has been buying Tesla Megapacks to power the Memphis training cluster. Musk has talked about the businesses converging: inference in the car and the robot, connectivity from orbit, batteries and solar into the data centres that train the models.

A full merger is the speculative part, and it should stay labelled as such. Bloomberg reported in January 2026 that SpaceX had considered combining with Tesla or with xAI, and then xAI happened first. Morningstar's take after that deal is the grown-up version: the xAI combination made a near-term Tesla acquisition less likely, SpaceX going public may actually make a later Tesla deal easier because both shareholder bases can vote, and Musk has said he sees the businesses converging, so they do not rule it out. Tesla holders could still hate the terms. SpaceX, after the IPO, has often traded at a richer story than Tesla's car multiple. A merger at the wrong ratio is how you dilute the people who already took the 2010-2021 ride.

I am not predicting a deal tomorrow. I am saying the industrial logic is not fan fiction. Launch plus constellation plus frontier models plus cars plus robots plus grid batteries is one of the few combinations that can look Nvidia-sized without having to win the GPU war. Nvidia is the training pickaxe. A Tesla-SpaceX combination would be trying to own the physical world the models act in: wheels, hands, watts, and the network between them.

If you want a picture of number one that is not "Tesla sells 20 million cars," that is the picture. One listed entity, or two listed entities with the same controlling mind and a thickening web of offtake, is how you get energy, mobility, labour, launch, and connectivity under one gravity well.

Why number one is possible, not magical#

As of writing, Tesla is not close to the top of the league table. Nvidia and Apple still sit in a different altitude, around the four to five trillion mark. Tesla is a trillion-plus company that the market still argues about as if it were GM with a neural net.

Number one is a market-cap statement. Market cap is a story about cash flows people believe will arrive, discounted. Tesla gets there if enough of the following become believed, then real:

Cars stay a cash engine even if unit growth is lumpy. The fleet is the data. The factories are the robot plants in waiting.
Energy keeps doubling in slow motion. Megapack is the proof that Tesla can sell something other than a driver's seat, at infrastructure scale.
Unsupervised Robotaxi escapes the city-by-city crawl, including through European approval that does not currently exist. Software on a paid-off fleet is how you jump from auto multiples towards network multiples.
Optimus becomes a manufactured product, even if the first useful years are Tesla's own shops. Labour TAM is the only TAM in this essay that can dwarf cars without anyone stretching the metaphor.
SpaceX stays on the same team, whether that is a merger, a holding company, or a decade of tightly coupled offtake. Starlink in the cab, Megapacks in the training cluster, Grok in the dashboard, solar on the catch.

You can knock any one of those over. You cannot knock all five over and still call the 2006-2025 sequence an accident. The company that said it would use a sports car to fund a family car, then said the family car would drive itself, then said the grid was the market, then said the robot was the product, has a habit of being early, loud, and eventually roughly correct.

That is the whole case. Not that Tesla is the world's number one company today. That the evidence it can be is already in the documents, the factories, the unsupervised miles, and the launch cadence next door. The missing pieces are regulation, especially in Europe, and whether Optimus can be built like a car rather than demonstrated like a prototype.

I would rather be early on that than wait for the car-unit crowd to notice the master plan was never about the car.

What would make this wrong#

A serious argument needs the kill-shots.

Safety. One bad unsupervised incident, filmed well, and the US cities that are open can close. Robotaxi is a permissioned business. Permission is withdrawn faster than it is granted.

Europe stays Level 2. If FSD Supervised is the ceiling for a decade, Tesla can still sell cars in the EU. It cannot run the Robotaxi model there. That is a large, wealthy TAM left on the table, and it is the scenario I think is under-discussed.

China and BYD. The car cash engine can be competed down to a commodity. Tesla has cost and brand. It does not have a monopoly on making EVs. If automotive gross margin keeps compressing while autonomy stays small, the spend on Optimus and Cybercab gets politically harder inside the company.

Optimus never leaves the academy. A thousand training units is a research programme. A million shipped units is a market. Tesla has not shown the second one. Supply chain, actuators, batteries, and the social question of labour displacement are all real.

A merger that destroys value. Combining Tesla and SpaceX at the wrong price, or under a voting structure public shareholders will not accept, is how you turn a stack into a fight. Governance is not a footnote when one person sits at the centre of both.

Dates. Musk will miss more of them. If you need calendar certainty, this thesis will make you miserable. The evidence is the sequence, not the keynote.

I still think the sequence is the tell. Four master plans, each one bigger than the last, each one mocked, each one partly shipped. Robotaxi is live in America and legally smaller than its TAM in Europe. Optimus is behind and pointed at the largest market humans have. SpaceX already did the impossible thing in launch, listed, and plugged Tesla batteries into its own training cluster. Number one is a stretch. It is not a fantasy.