A two-seat sports car lifted off a ramp near Waco, Texas on Thursday night, turned in the air and came back down under its own control. Nobody was inside. The flight lasted seconds. And for a few minutes, the oldest promise in the automotive imagination stopped being a slogan and became something you could watch on a phone.
What follows is not a review. It is an argument about what just changed, what it could become, and why the answer is bigger than the car.
What actually happened#
Tesla unveiled the second-generation Roadster at a SpaceX test facility in McGregor, Texas. The car was first shown as a prototype in 2017, when Tesla said deliveries would begin in 2020. This time the company said production would start in 2028. It did not announce a price; in 2017 it quoted 200,000 dollars as a base figure.
The demonstration: the car drove up a curved ramp, stopped halfway, then climbed and held itself in place. It drove upside down along the ramp using a suction system, performed a sideways roll, and landed with the help of built-in thrusters. Later it jumped over a truck and a row of other vehicles while rotating in the air. The vehicle engineering chief, Lars Moravy, said the Roadster accelerates from zero to sixty miles per hour in under a second and reaches more than 230 mph. Musk gave 0.81 seconds for the demonstration car and said the production version would be quicker.
"This is the culmination of incredible work by Tesla and SpaceX engineers working closely together to make something that you haven't seen in any movie."
Then he told the audience that success was not guaranteed, that the car ran without a driver in order to make sure nobody got killed, and that the Roadster is, in his words, not necessarily a safety optimized car. He also said it should become Tesla's only manually driven vehicle in the long term.
How the trick works: ten thrusters and a vacuum#
Top Gear, whose reporter watched the thing in person, counted ten compressed-gas thrusters: four at the rear facing outward, two at the front facing outward, and one at each corner facing downward. The rear four add thrust during acceleration, which is why the car can reach sixty miles per hour in under a second without vaporising its tyres. The front two help it brake and rotate. One at each corner gives the four downward-facing units that lift the car, hold it steady and set it back down.
Nothing here burns. A cold gas thruster is a pressure vessel and a valve: inert gas escapes through a nozzle and the recoil does the work. It is the same family of hardware SpaceX uses to adjust the attitude of a booster, and it leaves no flame, which is a useful property when the thing is standing on asphalt in front of a crowd. It is also, and this matters later, the most primitive form of propulsion humans have ever flown. No combustion. No oxidiser. No plume of fire. Just stored gas, let go.
What the coverage said#
The arithmetic that decides what this is#
You can work out the ceiling on this technology with a sheet of paper. The numbers below are my own estimate, and I show the arithmetic so you can check it rather than take my word for it.
A car of this class weighs roughly two tonnes, so holding it up needs about 19,600 newtons. A nitrogen cold gas thruster delivers a specific impulse of 50 to 75 seconds; taking 65 seconds, the mass flow needed is 19,600 / (65 x 9.81), or about 31 kilograms of gas per second. Three seconds of hovering therefore consumes around 92 kilograms of nitrogen. Compressing nitrogen to storage pressure costs about 0.2 to 0.25 kilowatt hours per kilogram, so the hop stores and spends something like 20 kilowatt hours. An efficient electric car uses 15 to 18 kilowatt hours per 100 kilometres. The jump in Waco cost about as much energy as an efficient car uses to travel well over a hundred kilometres.
That number settles one argument and opens another. It settles the argument about transport: this is not how anyone will get to work, because the physics sets the duration at seconds and no amount of engineering moves that. Lifting two tonnes against gravity for three seconds costs more energy than rolling it for a hundred kilometres. A car that flies is a car that can afford to fly, and the affording is the hard part.
But it opens the argument that actually matters. Thrusters give a vehicle an axis of control that wheels cannot provide: force without waiting for tyre grip, correction without steering angle, an answer in the third dimension. Read that way, the stunt in Texas is not a claim about aviation. It is a demonstration of authority over a heavy object in three axes, and authority of that kind is precisely what a machine needs before you let it move people.
Nine years, and the arithmetic of promises#
The Roadster has a history that should be printed next to the fireworks. Announced in 2017 with a 200 kilowatt hour battery, a claimed 620 mile range, a 1.9 second sprint and a 2020 delivery date. Teased in 2018 with what Musk then called a SpaceX option package of about ten small rocket thrusters arranged seamlessly around the car, which means the idea being celebrated this week is eight years old. Called a small potato situation in 2023, in his own words, a car that would not move the needle financially. Then years of silence, and a stunt.
What is new is not the ambition but the demonstration. In 2017 the company showed a picture and a specification sheet. In 2026 it showed a car in the air. The gap between those two things is nine years of engineering, and the gap between the car in the air and a car in a customer's garage is still called 2028, which is the same kind of date as 2020 was.
Now the bull case, because a fair reading owes it and because this is where the evening stops being a car launch.
The bull case: what this makes possible#
Start with the most obvious thing, the one every engineer has been trained to dismiss: a mass that can be pushed in any direction is a mass that can be rescued in any direction. A car that can lift a corner of itself can lift a corner of itself off a muddy verge, out of a flooded underpass, out of a snowdrift, off a kerb it should never have hit. The Roadster spends twenty kilowatt hours to hover for three seconds. A stranded ambulance spends far more than that standing still, and money has not yet been invented that buys back the minutes before a road is cleared. Nobody will sell you a flying ambulance this decade. But the hardware that holds a two-tonne object in the air is the same hardware that shoves it sideways off an obstacle, and the second application is boring, cheap and enormous.
Next, the direction the company is actually moving. Tesla stopped selling the Model S and Model X this year to make room for Optimus robots, and it launched the Cybercab, a robotaxi without a steering wheel, pedals or mirrors. Musk says the Roadster will be the last Tesla a human drives by hand, and calls it not necessarily a safety optimized car. Put those two sentences side by side and you have the whole story: the last car built for a person, made by a company that has already decided cars are no longer for people. The stunt is the farewell party, and the fireworks were real.
Then the hard part, the part a cautious piece leaves out and that I think is the actual argument. Consider what has to be true for a car to hold itself level in the air, outdoors, with wind moving over it, in front of a crowd, and then set itself down. Sensors have to know the attitude to a fraction of a degree. Software has to decide in milliseconds. Actuators have to deliver exactly the commanded force and no more. All three have to work as one system, every time, because a mistake does not produce a warning light. It produces a car falling on people. That is a solved problem now. Not in theory, not in a paper: on a Thursday night, in Texas, in front of witnesses, with the chief executive admitting beforehand that it might not work.
A machine that can hold two tonnes perfectly still in the air can hold anything still, anywhere. That is the sentence people will remember from this week, and it has almost nothing to do with cars.
From there the applications write themselves, and they are mostly not about flying. A vehicle that can command force in three dimensions does not need to be airworthy to be useful: it can right itself, brace itself, catch itself, and hold its position against something stronger than its own tyres. An autonomous fleet made of such machines is a fleet that recovers from its own mistakes instead of asking a driver to. Add cheap energy and you have the precondition for everything the twentieth century promised and never delivered. Not flying cars. Machines that a wet leaf on a hill cannot stop, or a patch of ice, or the moment a human stops paying attention.
And yes, the honest version of the flying part. The air is the last unbuilt road. Free of traffic, free of tolls, free of the entire apparatus of asphalt, and barred to ordinary people for one reason: flying was always a specialist's act. Too expensive, too dangerous, too regulated. What changed on Thursday is not that cars can fly. It is that stabilising a heavy object in the air has become a software problem. Software gets cheap. Software gets copied. Software turns a stunt into a feature, and a feature into a commodity. That is the door the Roadster just pushed at. Not that you will fly to work, but that what kept you from flying has stopped being physics and started being paperwork. Paperwork moves.
And the cold water#
The sceptics have the better case if the claim is that cars will fly next year. The vehicle ran unmanned, inside a cordon. The chief executive said in advance that success was not guaranteed, and Tesla itself says the flight functions are not approved for public road traffic. Top Gear's own review of the evening ends with two questions nobody answered: what does it cost, and where on earth would you be allowed to do it. Tesla announced three versions, and only the third, the Rocket, with extra thrusters for upward movement and rotation, will be available by invitation only. A car you cannot buy, using a feature you cannot use, on a road where it would not be legal, is not a product. It is a demonstration, and Tesla is careful to call it one.
The comparison that keeps honest is this one. EHang's EH216-S holds the first type certificate ever issued for an electric vertical take-off aircraft, in China. Joby Aviation, generally considered the leader in the West, flew its first FAA-conforming aircraft in early March, and as of this month neither Joby nor Archer has passed FAA certification. So the aircraft builders have spent a decade proving safety to regulators and are still waiting, while Tesla spent a Thursday evening proving a point to a camera. Both are real work. Only one of them has a certificate.
The fairest verdict is that the sceptics are right about the product and wrong about the engineering. Nothing here will be sold as a flying car, and nothing here will be legal on a motorway. But holding a two-tonne object level in the air and putting it down gently is a control problem, and solving it with gas thrusters means the software, the sensors and the actuators all worked under conditions where a mistake is expensive. That is transferable. A paint job is not.
What it says about the future of the species#
The flying car is a hundred-year-old promise, and it did not arrive in the twentieth century for a reason that had nothing to do with wings. The ground was cheaper. Roads are built once and shared by everyone; the air requires a machine, a pilot, a regulator and a great deal of energy for every trip. What changed is not the physics of flight. It is the cost of control. Computing, sensing and thrust vectoring have become cheap enough to stabilise a car in mid-air for a few seconds, and that is the same force that is making machines drive themselves, land boosters and walk on two legs.
That is the honest glory of Thursday night, and it is glory, if you allow the word. A car flew because its maker could afford to show that it can, and because the ability to command a heavy object in three dimensions has stopped being a hardware miracle and become a software discipline. Every industry that runs on moving atoms should be paying attention, because the discipline transfers: to the crane that must not drop its load, the tractor that must not roll on a slope, the delivery robot that must not fall down the last step, the aircraft that must not lose its attitude in a gust. None of them will look like the Roadster. All of them just got a proof of concept they did not ask for.
The Roadster touches the two questions that decide this century and answers neither. It spends the energy of a long drive to hover for three seconds, and it demonstrates the control authority that autonomy needs while being, by its maker's own description, the last car he expects a human to drive. A company turning itself into a robotics and artificial intelligence business built a flying sports car and told the world it will not need a steering wheel much longer. That is not a contradiction. It is a farewell, delivered at 230 miles per hour.
And the part that is easy to miss because it sounds like marketing: this is what a frontier looks like when it reopens. For fifty years the car industry competed on refinement. Quieter, safer, more efficient, marginally better. That is the competition of a mature art, and mature arts are about the last two percent. Thursday was not refinement. It was a company doing something with no obvious use, no legal path and no price, in public, with a straight face, because it could. Civilisations do not advance by optimising what exists. They advance when somebody builds the thing that does not yet make sense, so that twenty years later it can become boring. The Roadster is not the future of transport. It is the sound a door makes when it opens.
What we still do not know#
- Price, mass and range of the production car as it will be sold. None of the three was mentioned at the event; the figures in circulation are from 2017.
- Whether customers will receive the thruster package at all, or whether it stays a stunt for events.
- Whether any regulator anywhere will permit gas thrusters on a road vehicle, and under what conditions. Tesla's own statement that the flight functions are not approved is the only official word so far.
- How much of the demonstrated behaviour survives into a car built to be sold and serviced, including the part where a pressure vessel has to fail safe every day of its life.
- Whether 2028 means anything. This programme has already missed one delivery date by eight years.
Sources#
- Thao Nguyen, Forget zero to 60. Tesla Roadster flies, flips with SpaceX thrusters, USA Today, 16 October 2026
- Jack Scarlett, Wow: the new Tesla Roadster can actually fly, Top Gear, 16 October 2026
- Maurus Elias Moritz, Hier kommt der neue Tesla Roadster geflogen, Auto Bild, 16 October 2026
- Watch: Did Tesla really just unveil a flying car?, BBC News, 16 October 2026
- Tesla's new Roadster makes a flying debut, Fox Business, 16 October 2026
- SpaceX thrusters just made a Tesla Roadster fly, Space.com, 2026
- Tesla Roadster (second generation), Wikipedia
- Tesla stock rises 5% on Roadster reveal, Invezz, 16 October 2026
- Joby and Archer FAA certification status, October 2026
- Cold gas thruster specific impulse and Cold gas thruster
Method note: everything factual is quoted from the sources above, including every figure attributed to Tesla. The energy estimate is my own calculation and is labelled as such, with mass, specific impulse and compression cost stated so it can be checked. The thruster positions in the illustration follow Top Gear's description. The illustration is my own drawing; the photographs and screenshots belong to their publishers and are reproduced here for review. The section titled the bull case is an argument, not reporting: where it reasons past what the sources say, it says so.
Comments0
No comments yet. Start the conversation.