Tesla Makes the Biggest Bet in Transportation History

Tesla's massive bet on Full Self-Driving is starting to pay off. Influencers brand new to Tesla are getting in a car and coming out stunned. You can lease a basic Model 3 for about $300 a month right now, and it drives itself. Car reviewers with huge audiences are showing millions of people a vehicle that can go coast to coast in the United States without anyone touching the steering wheel or the pedals.

Tesla posted its best second quarter in company history and the third-highest quarterly total it has ever reported. The median delivery estimate was roughly 409,000 vehicles. Tesla beat that by about 20 percent.

The federal tax credit that used to take as much as $7,500 off an electric vehicle disappeared in the United States after September 30, 2025. US electric vehicle sales then fell 20 percent year over year in the second quarter. Higher gas prices helped hybrids more than fully electric vehicles in America. Tesla's global deliveries still jumped 25 percent from the same quarter last year.

So how does a brand that lost a $7,500 discount, with a CEO who has been one of the most hated people on the planet, suddenly go on a tear?

Several forces are hitting at once. Higher oil prices are pushing more people toward electric vehicles in parts of Europe and Asia. Tesla has cheaper versions of the Model 3 and Model Y, plus financing and leasing that make the monthly payment easier. And the biggest piece: Tesla's self-driving software has reached people outside the die-hard Tesla bubble. They sit in the car and realize it can actually drive them around.

Tesla ended the quarter with 1.48 million active Full Self-Driving subscribers, up 56 percent from a year earlier. More than 55 percent of new Teslas delivered in North America included an FSD subscription at delivery.

You enter a destination, press a button, and the car steers, changes lanes, follows the route, and parks. A group of Tesla owners have done more than 20,000 miles straight without touching the wheel or pedals. The current consumer version is still supervised. You remain responsible for the car. That will change, and unsupervised driving is a much bigger deal than most people are treating it as.

Independent reviews flipped the same way. In 2025, Motor Trend gave its driver assistance award to General Motors Super Cruise and described Tesla's system as exhausting and unpredictable. After testing FSD version 14, the same publication named Tesla's FSD the best driver assistance system on the market for 2026. The public may be crossing the line from treating FSD as a science experiment to treating it as a feature you would buy a car for.

Picture two electric crossovers with similar range and similar monthly payments. One gives you the usual package. The other can drive the entire trip from your driveway to a parking space while you supervise, and eventually without supervision at all. One is a car you have to drive. The other is a transportation unit that takes you from A to B while you do something else. For competitors, that is brutal, because this capability is extremely hard to copy.

Tesla's cars use cameras. The company trains a neural network on enormous collections of driving examples. Improved software goes back to customer vehicles through an online update. Inside the car, an AI computer runs the trained model: cameras stream images, the model builds a scene, predicts what may happen next, chooses a path, and sends commands to the steering, accelerator, and brakes. Training is the expensive process of teaching the model. Inference is what happens every time the trained model looks at a new camera frame and decides what the car should do.

Tesla has spent almost a decade building the data pipeline, training computers, chips, software, and vehicle hardware that loop requires. A traditional automaker cannot buy one magic part from a supplier and wake up with the same system next Tuesday. They have tried. They are nowhere close.

Cox Automotive estimates Americans bought about 247,000 new electric vehicles in the second quarter, down 20 percent from a year earlier. The average new EV price in June was over $56,000 even after average incentives of $7,500. Tesla's average transaction price was a little over $53,000. That $7,500 credit could hide a high price, a poor charging network, or half-baked software. Take it away and buyers get much harder to please.

Tesla concentrated on two cars, the Model 3 and Model Y, and those two made up more than 97 percent of global deliveries in the quarter. That scale lets Tesla spread the cost of software, factories, batteries, and service centers across hundreds of thousands of similar vehicles. The risk is obvious. If people stop wanting those two, there is not much else to sell. Cybertruck sales are still small and below what the company hoped. Model S and Model X are expensive niche products, and those production lines are being replaced with next-generation production for the Optimus robot.

Legacy automakers have a different problem. Many make good electric vehicles, some more luxurious and better built than a Tesla. They also need to protect the gas trucks and SUVs that make all of their profit. If an EV loses money, selling more of it can hurt the business that pays today's bills. Ford has lost billions of dollars funding an electric vehicle program that has gone nowhere. Dealers also make money from oil changes and engine repairs that disappear with electric vehicles. Without the tax credit, those EVs lose money, so the companies do not push them.

Gas prices gave buyers one more reason to think about value. The US average hit $4.48 a gallon in May, according to the Energy Information Administration. The war with Iran slowed oil shipments through the Strait of Hormuz. In Europe, fuel taxes make gas even more expensive. During the first half of 2026, fully electric vehicles reached 20 percent of new vehicle sales in the European Union, up from 15 percent a year earlier. In America, many people chose a hybrid. Cox Automotive thinks hybrid sales grew about 9 percent in the first half of the year while the whole new vehicle market fell. High fuel prices likely made more people look at EVs, especially outside the United States, and Tesla's scale let it match that demand. Tesla delivered 480,000 vehicles.

The real secret weapon is a car that keeps getting better at driving itself. A normal automaker gets paid when it sells the vehicle, then hopes you come back years later. Tesla can sell the vehicle once and keep selling software on the same hardware for the life of the car. If FSD gets better, an older Tesla gains a capability it did not have when it left the factory.

Each new Tesla gives the company four ways to make that product more valuable. First, it earns money when it sells or leases the vehicle. Second, the owner may pay for FSD each month. Third, vehicles on the road help Tesla find hard driving scenarios. The company looks for a useful event, saves the clip, trains the next model, and sends it back to the fleet. Tesla recently crossed 13 billion miles on FSD, significantly above 90 percent of all self-driving miles out there. Fourth, each owner shows the product to someone else. The Robotaxi network in select cities can bring that experience to people who do not know a Tesla owner. Open the app, pay something like five bucks, and a Tesla with no one in the driver's seat takes you downtown. If that rider later wants the car, Tesla does not need an ad. They already lived it.

Then comes Cybercab, Tesla's two-seat vehicle built only for autonomous rides, with no steering wheel or pedals in the planned design. Tesla says it started Cybercab production in the first half of 2026 and plans to use Cybercabs in place of Model Ys in its ride service over time. Tesla says the production line will exceed 2 million units per year once fully ramped. A self-driving car is a robot. Cars may be the first autonomous robots made in huge numbers that most people will ever sit in.

That same technology feeds Optimus, Tesla's humanoid robot. Tesla wants it to do work that is unsafe, boring, or repetitive, and wants to manufacture millions per year as early as 2030. Let's be honest: that timeline is going to slip. The company has already begun building the Optimus manufacturing line in Fremont, California, replacing the old Model S and Model X lines. Both systems use cameras or other sensors, AI, planning software, and a fast, low-power computer inside the machine. Both run the same learning loop: collect failures, train a better model, test it, send the update back out.

If self-driving helps sell Model Ys, which it very much looks like it is, Tesla will have proved that ordinary people will pay for AI that can act in the real world. That means Tesla is going to need a lot of chips, which is why it entered a joint partnership with SpaceX called Terafab to build custom chips for cars, robots, and data centers, both in space and on the ground. That is also why Elon Musk and others keep teasing a SpaceX and Tesla merger.

Check the video here.