Autonomous vehicles

Driving automation is at the forefront of software developments for cars today. While fully autonomous cars (Level 5 automation)) are not currently in sight, electric driverless taxis (Level 4) are already operating commercially in more than 20 cities worldwide. Moreover, automated driving systems are not limited to self-driving cars – they are also rapidly gaining importance in the form of ADAS for private vehicles.

EVs are more frequently equipped with ADAS features, and all commercial robotaxi services in operation today exclusively use EVs. High‑voltage batteries, reduced mechanical complexity, and precise torque control support the integration of automation systems with BEVs. ADAS and autonomous systems rely on extensive data collection and continuous software refinement – an area where digital-first architectures, pioneered by EV makers, provide a clear edge. The mutually reinforcing trends of software‑defined vehicles and automated driving are therefore consolidating the technological leadership of EVs.

Over the past 10 years, ADAS have become much more widespread. In 2025, around half of new cars sold globally featured systems that can automate steering and speed control (Level 2 automation), whereas 10 years earlier this was limited to very few high-end models, accounting for less than 1% of sales. However, the most commonly deployed systems, like adaptive cruise control combined with lane-keeping assist, require the driver to have their hands on the steering wheel, and only function under certain conditions, mainly on highways.

Further development of ADAS is amongst the top strategic priorities of OEMs today. Beyond compliance with regulations and increasing road safety, OEMs see driving automation as one of the main competitive differentiators, with high potential to unlock new revenue streams. The most advanced Level 2 systems in private passenger cars today allow for hands-free navigation on autopilot (Level 2+). People’s Republic of China  (hereafter, “China”) and the United States are adopting these systems most quickly; 10% and 6%, respectively, of cars sold in 2025 in these countries were equipped with Level 2+ technology. In the United States, Level 2+ ADAS are either sold as premium features or via software subscription services, whereas in China they are often included in base models. With these systems, drivers are still required to be ready to instantly take full control over the vehicle whenever requested.

Applications of Level 3 systems, in which the legal responsibility shifts to the manufacturer while the system is active, are still very rare. In 2021, Mercedes-Benz was the first manufacturer to offer a Level 3 system: its Drive Pilot system was made available as a subscription service in two high-end models, one of them battery electric. The system permits autonomous driving in dense traffic and good weather conditions at speeds up to 95 km/h on motorways in parts of Germany and the United States. Drivers can take their eyes off the road but must be ready to retake control of the vehicle with 10 seconds warning. Level 3 systems have a much higher need for redundancy, driving up cost. Furthermore, regulations and certification for Level 3 systems differ by country, complicating widespread deployment. Notably, Mercedes-Benz decided to no longer offer their Level 3 system in their latest model update in early 2026, instead offering Level 2+ systems with wider application. BMW is reportedly also planning to discontinue its Level 3 system. This is in line with a wider industry shift away from developing Level 3 systems to integrating more sophisticated Level 2+ functionality for personal vehicles. In China, however, interest is still growing, with the first local permits for two models with Level 3 systems granted in 2025.

Driverless vehicles have now become a common sight in multiple cities in the United States and China, in the form of robotaxis. Such Level 4 systems avoid the handover of responsibility while driving and legal grey zones which currently inhibit Level 3 deployment. Robotaxis operate within a geolocalised domain, such as a single city, and are deployed as centralised fleets, allowing companies to continuously monitor, analyse and improve functionality through software updates. Autonomous trucks are also gathering momentum, with the first early commercial deployments in the United States and China. Both robotaxis and autonomous trucks offer large potential for efficiency gains and cost savings, but ensuring reliability under all conditions and complying with regionally diverse regulations remain challenging.

Today, all robotaxis are BEVs, due both to their technical characteristics and cost advantages. First, operating costs are lower for electric cars, which is an important consideration for intensively used taxis. Second, software control of EVs is more mature than for ICEVs and hybrids. This can be partly explained by the lower mechanical complexity of electric cars, and partly by the recent push for centralised architectures and SDVs led by pure-play EV makers (see above). Third, battery electric cars comply with emissions standards in cities. Finally, power draw for onboard computation and sensors needs to be very stable and can be substantial – estimates span from several hundred watts to over 1 kW – which can be more easily accommodated by EV batteries and power electronics.

Robotaxi deployment saw record growth in 2025. Following several years of commercial pilots, the robotaxi fleet more than doubled to reach 8 000 vehicles spread across around 20 cities globally. Commercial services today are concentrated in China and the United States, though Dubai and Abu Dhabi also have commercial robotaxis on the road. Testing of autonomous taxis has begun in Europe, Japan and Korea.

So far, robotaxi services have been offered by only a few companies, but competition is set to increase over the coming years. Commercial robotaxi services were dominated in 2025 by Waymo in the United States and by Baidu, WeRide and Pony.AI in China. All these companies have plans to expand to more cities and countries. Meanwhile, companies offering ride-hailing services, such as Uber, Lyft and Bolt, have formed partnerships with various developers of self‑driving technology, though are not yet offering commercial services. Many of these partnerships target launch in 2026 and 2027, which will significantly increase the number of players in the market. While car makers such as Volkswagen and Nissan are now developing their own robotaxi models, some pure-play EV makers, such as Tesla and Xpeng, are intensifying R&D efforts on autonomous driving by developing their own chips and AI models.

While early robotaxis were built on existing EV models, the focus is increasingly shifting toward purpose-built, software-defined vehicle architectures in which autonomous driving capabilities are integrated at the design stage.

Using a self-driving taxi is currently more expensive than other ride-hailing services both in the United States and China, but the price gap has narrowed. The cost structure underlying robotaxi services differs markedly from other ride-hailing services. Whereas the main cost-component in standard taxi services is the driver, accounting for over 50% of total costs, robotaxis have significantly higher upfront costs for sensors and computer hardware, as well as a need for increased maintenance and vehicle control centres.

Costs for robotaxis have declined significantly over the past few years. Current designs employ fewer sensors than the first models, and the costs of electric cars, batteries and sensors have also fallen at the same time. For example, the first robotaxi to enter commercial operation, Waymo’s Jaguar I-Pace, used 40 sensors and cost more than USD 100 000. Waymo has since reduced the number of sensors and diversified to cheaper EV models, and thus reduced vehicle cost to around USD 70 000. Baidu announced in 2025 that manufacturing costs for the Apollo RT6 self-driving car had been reduced to under USD 30 000, benefiting from drastic cost reductions for lidars1 in China. Tesla is targeting similarly low costs for their robotaxis by relying solely on cameras, but has not yet gained Level 4 approval.

Driverless taxis will expand into more neighbourhoods, cities and countries over the next decade. The pace of scale-up is, however, slowed by significant costs and the need for city-by-city testing and approval, as well as fragmented regulation and customer acceptance. Estimates of global robotaxi fleet size in 2035 range from 700 000 to 3 million vehicles, which are likely to remain concentrated in 40 to 80 cities globally. While this would constitute only a small fraction of the global taxi and ride-hailing fleet, robotaxis could claim most of the ride-hailing market in hot spots. For example, in San Francisco, Waymo has risen to the second most-used ride-hailing service, overtaking Lyft’s market share in 2025. Future cost‑competitiveness with private car ownership is still uncertain, meaning that impacts beyond the ride-hailing sector are not likely to materialise before 2035.

There is an even clearer business case for autonomous freight transport, especially in regions with high wages or driver shortages. Autonomous trucks do not require resting periods, increasing logistics efficiency, and lowering the total cost of ownership. Self-driving on highways is also more mature than navigating through dense cities. Routes are also often more predictable, allowing services to narrow their operational domain. An additional challenge compared to passenger transport is the integration into logistics operations. Whereas passengers enter and exit the vehicle by themselves, autonomous trucking and logistics need to consider loading and unloading of vehicles and further integration into supply chains.

The first autonomous trucks on public roads began commercial operation in the United States in 2025. Other early commercial deployments are also underway in China, Europe and elsewhere in the United States, covering various applications from local delivery services to long-haul trucking. Hub-to-hub logistics are likely to be the first application in which autonomous operations could have a big impact. It has been estimated that up to one-quarter of truck sales for hub-to-hub logistics could be autonomous in the United States, Europe and China by 2035. By contrast, technical and regulatory hurdles increase with higher complexity and variability of routes, such as in point-to-point services and in urban distribution, which is likely to limit deployment over the next decade.

While many pilots and early deployments of autonomous trucks use traditional diesel engines, there are clear synergies between electrification and autonomy. For electric trucks, autonomous operations can increase utilisation by avoiding driver rest periods, amplifying the effect of lower running costs of electric trucks, due to lower fuel costs. This can significantly shorten the payback period for the higher upfront costs of electric trucks. To fully capture these utilisation benefits, however, autonomous electric trucks must minimise downtime, and charging time becomes the key constraint. Innovative charging solutions such as battery-swapping, MW-scale chargers or inductive charging pads placed at warehouses where trucks load and unload could therefore support the automation of electric freight trucks.