Self-driving car: future or (almost) immediate present

For nearly a century, the dream of a self-driving vehicle has captivated engineers and inventors. As early as the 1920s, American Francis Houdina astonished New York with a radio-controlled car, while in 1939, General Motors envisioned highways where cars navigated themselves in their Futurama exhibition. During the 1950s and 1960s, the first tests with remote-controlled or wire-guided cars showed that the idea could become a reality.

The first major leap forward came in the 1980s , when German engineer Ernst Dickmanns enabled a vehicle to drive autonomously using cameras and image processing. Shortly after, the European Prometheus project (Programme for a European Traffic of Highest Efficiency and Unprecedented Safety) brought autonomous driving to real roads with Mercedes-Benz—following successful tests by Ernst Dickmanns and his team at the Bundeswehr University in Munich—using VaMoRs (Experimental Vehicles for Autonomous Mobility and Computer Vision, or Versuchsfahrzeug für autonome Mobilität und Rechnersehen). These tests enabled a modified S-Class to travel hundreds of kilometers on European highways without human intervention, reaching speeds of up to 130 km/h for most of the time.

From the 2000s onward, advances in artificial intelligence (AI) and sensors accelerated the race. The DARPA Grand Challenge (Defense Advanced Research Projects Agency) competition in the United States spurred the development of increasingly precise and safe systems, becoming a technological breeding ground for the major companies that would follow, firms such as Waymo (Google), Baidu, and Tesla , which are leading the way in the new era of autonomous driving.

Autonomous driving without a driver is already crossing the border between pilot tests and large-scale commercialization to become one of the most disruptive economic and social drivers of the next decade, according to the Bank of America (BofA) Institute in its report The road ahead: The future of autonomous vehicles , which highlights that three factors are key to this path: the push of generative artificial intelligence; the fall in technological costs; and the avalanche of global investment.

A revolution in progress

The transformation is profound and will only intensify. More than 120 robotaxis programs are currently operating worldwide, driven by 32 companies, according to BloombergNEF . Seven of them already offer safe, driverless commercial services, primarily in the United States and China . In cities like San Francisco, Phoenix, Shenzhen, and Wuhan, thousands of passengers travel daily in fully autonomous vehicles, while tech giants and traditional manufacturers compete to lead an industry that could be worth $1.2 trillion by 2040, according to BofA.

The paradigm shift is based on three pillars: cheaper hardware, smarter software, and scalable business models.

The cost of hardware for robotaxis in China, for example, has been reduced by more than 50% in a few years, and eliminating the driver reduces the price per kilometer by 52%, according to calculations by the Bank of America Institute, so the economic equation begins to add up.

The "ChatGPT moment"

Bank of America identifies a new phase in the development of autonomous vehicles (AVs): the leap from rule-based autonomy (AV 1.0) to generative AI-based autonomy (AV 2.0).

Until recently, systems relied on multiple neural networks that performed separate tasks—that is, perception, planning, control—and now, generative AI allows a single end-to-end model capable of directly transforming sensor data into driving decisions.

This approach, similar to the leap that ChatGPT represented for language systems, promises cars that will be more adaptive, capable of learning from changing environments and making more human-like contextual decisions.

To give you an idea, each test AV generates about 32 terabytes of data in 6 hours, and data centers need to increase their computing power tenfold to train the models. The report warns that AVs are literally becoming supercomputers on wheels.

The hardware race

The sensors - radars, cameras and lidar (laser light detection and distance measurement systems) are the eyes of the autonomous car.

The global market for sensors for autonomous vehicles exceeded $75 billion in 2024, and their cost is currently about nine times higher than that of advanced driver assistance systems (ADAS). But advances in artificial intelligence and hardware integration could narrow that gap.

The trend suggests that new high-end vehicles will already incorporate the necessary components to enable advanced autonomy without the need to add additional equipment.

Beyond cars

Although the headlines focus on cars, the biggest advances are actually taking place in trucks, buses, and industrial machinery.

AVs are already operating in mines and agricultural operations, with 30% increases in productivity and 50% reductions in labor costs.

According to Bank of America, the global shortage of drivers—3.6 million vacancies—which could double by 2028, reinforces the appeal of automation. In Japan, for example, the average age of taxi drivers is over 59; in Europe, truck drivers are around 47. With no generational replacement in sight, AVs offer a technological solution to a structural problem where autonomous trucks, in particular, are advancing rapidly.

There are 90 pilot projects, two-thirds dedicated to long-distance transport, and seven are already close to commercialization, an emerging business model that combines subscription or pay-per-kilometer services (driver-as-a-service) with fleets managed by manufacturers and logistics partners.

Investment in boiling point, paradox and new era

Bank of America estimates that between 2023 and 2024 alone, investments in autonomous vehicle companies tripled, reaching almost $9 billion. Since 2010, more than $200 billion has been invested in some 600 companies in the sector, and close to $1 trillion if associated trends such as connected, electric, and shared mobility are included. All of this reflects venture capital's focus on autonomy as the new competitive frontier of applied AI.

The urban impact of autonomy will be paradoxical. According to BloombergNEF, the global car fleet could peak in 2035 and shrink by up to 23% by 2050 if shared mobility thrives. But total kilometers traveled will increase: robotaxis and shared vehicles will generate more, longer, and cheaper journeys. Autonomous cars could thus reduce private ownership, but they will increase traffic.

The BofA report concludes that autonomous mobility will be both an economic disruption and a social transformation. From logistics to tourism, from urban transport to last-mile delivery, driverless vehicles promise to redefine time, space, and work. In the short term, they will coexist with human drivers; in the medium term, they will replace them in repetitive or risky tasks; in the long term, they could eliminate the steering wheel altogether.

The race is therefore no longer about manufacturing cars, but about teaching them to think.