China's Automakers Turn to AI and Autonomous Driving as Domestic Sales Slump

A crowded Beijing show reveals a changing car business

China's passenger vehicle sales fell 17% in the first quarter of 2026, while vehicle exports rose more than 60%. At the Beijing Auto Show this month, manufacturers presented their response to that widening gap: cars that can interpret spoken instructions, software that connects vehicles with homes, and autonomous systems intended to carry passengers and freight.

Hundreds of manufacturers gathered across 380,000 square metres on the outskirts of the Chinese capital. More than 1,000 vehicles were displayed, and hundreds of thousands of enthusiasts streamed through the exhibition halls. Yet hardly anyone was behind a wheel. At the world's largest automotive exhibition, the attention had shifted from driving a car to what a car might do for its occupants.

Chinese manufacturers have already cornered their domestic electric vehicle market. They now see autonomous driving and artificial intelligence as the next frontier, at a time when government subsidy programmes are being phased out and competition is squeezing the money earned from each vehicle.

BYD, a bellwether for the sector, has reported seven consecutive months of declining sales. That pressure helps explain why manufacturers are looking beyond the initial purchase toward software subscriptions, intelligent services and overseas customers.

As of April 24, 2026, the industry's ambitions stretch well beyond passenger cars. The Beijing exhibition also featured autonomous delivery trucks, plans for humanoid robots and a 10-seat air taxi. Alongside those promises, recent crashes and robotaxi failures have pushed Chinese authorities to narrow the conditions under which vehicles can take control.

What Xpeng and Xiaomi want cars to do

Xpeng unveiled an AI model that allows a driver to give conversational instructions, such as asking the vehicle to park near the entrance to a shopping centre, instead of selecting precise coordinates on a map. The aim is to make the vehicle understand a person's intention rather than require a detailed technical command.

Xiaomi, the appliance and smartphone company that has entered car manufacturing, demonstrated an operating system that detects a driver's stress levels. It can adjust lighting and music at home before the vehicle reaches the driveway, connecting the journey with the environment awaiting the occupant.

Xiaomi's vehicle can also make restaurant reservations, compile notes during a journey and order coffee through voice integration. These functions illustrate how manufacturers are presenting the car as a platform for everyday services, alongside its role as transport.

Tu Le, managing director of Sino Auto Insights, a consultancy focused on China's automotive market, says intense competition has made domestic passenger vehicle sales an unviable business model on their own. He describes a market in which intelligent driving has become a widespread competitive feature.

The fact that almost every automaker has some version of intelligent driving makes it different to almost any market in the world.

Le says manufacturers are competing through extra benefits and new sources of income beyond the vehicle sale. Leasing software powered by AI offers one route: a customer pays to keep using a service, giving the manufacturer recurring revenue after the car leaves the showroom.

Huawei and Pony.ai expand the technology push

Huawei, the telecommunications company that has expanded aggressively into automotive technology, announced plans to invest up to 80 billion yuan, approximately £8.7 billion, over the next five years. The investment will support autonomous driving software and the computing infrastructure needed to run it.

The scale of that planned spending shows how central software has become to the industry's strategy. A vehicle's intelligent functions require both programs that interpret conditions and computing equipment capable of processing the information.

Pony.ai, a Beijing company developing autonomous driving technology, presented what it described as the world's first fully compliant Level 4 light duty truck capable of autonomous operation. Developed jointly with battery manufacturer CATL, the vehicle is intended for urban deliveries and freight journeys between cities.

Level 4 describes autonomous operation within defined conditions or areas. The operating limits matter because a vehicle designed for a particular delivery route or service area faces a different task from one expected to travel on any road.

Pony.ai also announced that its fourth generation heavy duty autonomous trucks have entered mass production. Cargo operations are scheduled to begin later in 2026, extending the company's ambitions from exhibition demonstrations to commercial freight services.

Robotaxi failures bring safety into focus

Wuhan became a warning about the practical risks of autonomous passenger services in March. Several Baidu Apollo Go robotaxis suffered system failures and stopped in traffic, leaving passengers stranded for hours on expressways while trucks passed nearby.

Police received multiple reports of immobilised vehicles blocking lanes. Videos circulating on social media showed one vehicle stopped in an active traffic lane rather than on the shoulder, with a passenger visible inside as heavy vehicles passed.

In Zhuzhou, a Hello autonomous vehicle struck two pedestrians, leaving both hospitalised in intensive care. Bystanders collectively lifted the vehicle to free one victim who was trapped underneath it.

The Hello vehicle had been operating under a testing programme approved only months earlier. It was subsequently suspended from service. Together, the Wuhan failures and the Zhuzhou collision put the safety of passengers, pedestrians and other road users at the centre of the deployment debate.

Chinese authorities have responded with an approach that puts safety first, tightening rules and seeking clearer responsibility before permitting wider use. Manufacturers' technical promises are now being assessed alongside how vehicles behave when something goes wrong.

A fatal Xiaomi crash and tighter marketing rules

A fatal crash in March involving a Xiaomi SU7 equipped with assisted driving technology added to that scrutiny. The vehicle struck a concrete barrier on a highway in Anhui Province, killing all three occupants.

According to Xiaomi's statements, the car was travelling at 72 miles per hour when it detected lane closures related to construction and issued warnings. The driver took control one second before impact.

The crash drew unusual public scrutiny in a sector where such incidents are typically suppressed by censors. It also prompted regulatory action over how companies describe assistance features to buyers.

In April, the Ministry of Industry and Information Technology banned misleading terms such as "autonomous driving" and "hands-free driving" in marketing for Level 2 vehicles. The action followed evidence that drivers were relying too heavily on assistance functions.

Level 2 technology assists the person driving, who remains responsible for controlling the vehicle. Understanding that responsibility is central to the ministry's marketing restrictions: a customer needs to know whether the car is helping them drive or has permission to take over under specified conditions.

The recent sequence shows how quickly commercial ambitions have met safety concerns:

• March: Several Baidu Apollo Go robotaxis stalled in Wuhan, stranding passengers and blocking traffic lanes.
• March: A Xiaomi SU7 crashed into a highway barrier in Anhui, killing three occupants.
• April: The Ministry of Industry and Information Technology restricted misleading marketing language for Level 2 vehicles.
• April 2026: Manufacturers showcased new autonomous systems and AI services at the Beijing Auto Show.

Why Level 3 approvals remain narrow

The Ministry of Industry and Information Technology has granted limited Level 3 permissions to just two of nine applicants: Beijing Automotive Group and Changan Automobile. The approvals reflect a cautious approach to allowing vehicles greater control.

Level 3 allows a vehicle to handle driving under specified conditions, with a human available to take over when required. The approvals in China are tightly bounded: the vehicles may operate only on designated highway sections in the manufacturers' respective hometowns, and computer controlled lane changes are prohibited.

On any other road, those vehicles must remain under human control. The permitted geography and restrictions on manoeuvres make the initial deployment much narrower than the broad automation presented in many product announcements.

Bill Russo, an electric vehicle consultant based in Shanghai, says the industry's earlier expectations ran ahead of the arrangements needed to support them.

What looked like an imminent L3 rollout was, in hindsight, a marketing-led acceleration running ahead of governance, insurance frameworks and public trust.

Russo sees the government's decision as formalising a pause intended to slow progress, narrow its scope and establish guardrails, rather than ending development. For manufacturers, that means commercial plans must advance at the pace of the operating permissions they receive.

Exports grow as Chery looks to Britain

China's vehicle exports rose more than 60% in the first quarter of 2026 as manufacturers accelerated overseas expansion. The United Kingdom, Canada and selected European countries are among their targets, particularly where entry is less restrictive than in the United States or European Union.

The contrast with the 17% fall in domestic passenger vehicle sales explains the attraction of foreign markets. Manufacturers are seeking customers abroad while their home market becomes harder to serve profitably.

Chery, China's largest car exporter, has become one of Britain's fastest growing brands since launching there in August 2025. It sold 13,500 vehicles between September 2025 and March 2026.

The company announced a target of 10 million annual global sales by 2030, compared with 5 million in 2025. Achieving that goal would mean doubling the stated 2025 figure.

Farrell Hsu, Chery's UK country director, described Britain as a strategically important contributor to that ambition.

This exceptional growth underlines Chery UK's position as a key contributor to the overall business growth by 2030.

Industry professionals describe Britain as culturally agnostic toward Chinese electric vehicles compared with countries that have blocked entry on national security grounds. That reception helps explain why manufacturers regard the British market as an important part of their expansion.

Robotaxi ambitions cross borders too

Baidu has announced partnerships with Lyft and Uber to deploy Apollo Go robotaxis in London, Germany and other European cities by 2026. Autonomous technology partnerships are following the same overseas routes as vehicle exports.

Geely plans to deploy thousands of driverless taxis globally in 2027 through Caocao, its subsidiary providing ride services. The proposals extend competition beyond selling vehicles to operating transport services.

Tariffs in major Western markets remain an obstacle for Chinese companies. Overseas autonomous operations also require them to navigate rules covering safety standards and data sovereignty, meaning where information is stored and which country's rules govern it.

WeRide, another company developing autonomous driving technology, has opened a compliance office in Frankfurt. It has pledged to store all European user data locally in response to privacy concerns.

Those commitments show that overseas expansion involves more than shipping a vehicle or installing software. Companies must also organise their services around the legal and data requirements of the markets they enter.

BYD challenges Tesla on price and hardware

BYD intensified the competition earlier in 2026 by offering its "God's Eye" driver assistance package at no additional cost. Tesla's Full Self-Driving software, by comparison, retails for nearly $9,000 in China.

Taylor Ogan, an American investor based in Shenzhen who has driven vehicles from both companies, argues that BYD's system is more capable in Chinese urban environments. His assessment concerns the systems' performance in those driving conditions.

The companies also take different technical approaches. Chinese systems typically combine information from several kinds of sensors, including radar and lidar, while Tesla relies exclusively on cameras and artificial intelligence. Combining sensor information is known as sensor fusion: the software brings several views of the surroundings together to guide its decisions.

Analysis by A2MAC1, a benchmarking firm based in Paris, puts the hardware cost of a middle tier God's Eye system at approximately $2,105, compared with $2,360 for Tesla's approach using cameras alone. On those figures, BYD's hardware costs $255 less.

Sensors cost roughly 40% less in China than comparable devices in Europe and North America, while lidar units carry a 20% discount. Economies of scale created by China's electric vehicle boom help explain those lower costs: producing equipment in large quantities spreads manufacturing costs across more units.

BYD's pricing puts pressure on competitors trying to charge separately for assistance software. It also shows why software revenue is a competitive challenge even as manufacturers look to it as a new source of income.

Foreign brands turn to Chinese partners

Cadillac announced its first vehicle with driver assistance for China, the VISTIQ luxury SUV, using software developed jointly with Momenta, a company specialising in autonomous driving. Hyundai's new IONIQ V incorporates similar technology from the same partner.

Volkswagen presented four new models incorporating AI from Tencent, Alibaba and Baidu. They included the ID. UNYX 09, developed jointly with Xpeng in just two years.

These partnerships show foreign manufacturers drawing on Chinese technology providers to compete in the domestic market. Competition between national brands is accompanied by cooperation between companies supplying vehicles, batteries and software.

At the exhibition, Volkswagen Group chief executive Oliver Blume was seen interacting with CATL chairman Robin Zeng at the Porsche booth. Xiaomi's Lei Jun toured displays alongside executives from Li Auto and Mercedes-Benz.

The encounters reflected the industry's close interdependence. Manufacturers competing for customers are also connected through the technology and equipment needed to build their next generation of vehicles.

Robots, air taxis and the systems still being tested

Xpeng chief executive He Xiaopeng promised that humanoid robots would follow the company's new GX SUV later in 2026. He also said flying cars would eventually enter mass production.

AutoFlight, a Chinese aviation startup, displayed a 10-seat air taxi. Long queues formed as visitors waited to inspect its interior. The exhibit reflected China's push into what officials call the "low-altitude economy," which includes drone deliveries and urban air mobility.

Industry analysts forecast that more than one million passenger vehicles with Level 3 capabilities could operate on Chinese roads by 2027, provided pilot programmes proceed smoothly. Goldman Sachs predicts that nine out of ten cars sold in China will feature Level 3 autonomy or higher by 2040.

The legal division of responsibility between driver and manufacturer remains unresolved, alongside questions about insurance pricing and infrastructure readiness. High precision positioning systems and frequently updated maps containing detailed road information are unevenly distributed across regions. Draft revisions to China's Road Traffic Safety Law propose specific chapters addressing autonomous operation.

Zhang Yongwei, president of China EV100, a Beijing industry think tank, says pilot programmes serve a purpose beyond checking the technology. They also test and improve traffic management, the processing of insurance claims and coordination between regulators.

This is not just a technical proving ground. It is a process that pushes both technological innovation and regulatory frameworks toward maturity.

China's manufacturers are pursuing software income, export growth and more autonomous vehicles at the same time. The immediate challenge is to turn those ambitions into services that can operate within the safety rules, infrastructure and responsibility arrangements being developed alongside them.

Key Points

• Domestic passenger vehicle sales fell 17% in the first quarter of 2026, while vehicle exports rose more than 60%.
• BYD reported seven consecutive months of declining sales as manufacturers sought revenue beyond the initial vehicle purchase.
• Huawei plans to invest up to 80 billion yuan over five years in autonomous driving software and computing infrastructure.
• Robotaxi failures, pedestrian injuries and a fatal Xiaomi crash increased scrutiny of safety and marketing claims.
• Beijing Automotive Group and Changan Automobile received narrowly restricted Level 3 permissions, two approvals among nine applicants.
• Chery is targeting 10 million annual global sales by 2030, while Baidu and Geely plan overseas robotaxi deployments.
• Cadillac, Hyundai and Volkswagen are working with Chinese technology providers on vehicles for China's market.
• Forecasts for wider Level 3 adoption depend on successful pilots, with liability, insurance and infrastructure arrangements still developing.