A BYD smart driving system showcased at the Beijing Auto Show in April 2026. Credit: CnEVPost
- ByteDance is reportedly exploring an entry into autonomous driving, with the project now in an early preparation stage.
- ByteDance has approached a number of top assisted-driving or autonomous driving talents, and some technical talents who are optimistic about the move plan to join.
ByteDance is reportedly exploring an entry into autonomous driving, suggesting the Chinese internet giant may extend its AI bets into a key part of vehicle intelligence.
Chinese tech media outlet 36Kr said in a Monday report, citing several industry sources, that the project is currently led by the world model team under ByteDance’s Seed team. The team’s research overlaps with the environmental understanding, prediction and planning capabilities needed for autonomous driving.
ByteDance is interested in autonomous driving scenarios including unmanned logistics, with the business falling under the auto industry line of its cloud services brand Volcengine, the report said.
Some people close to ByteDance said the company’s team had previously held business exchanges with some leading autonomous driving teams, and that the project had entered an early preparation stage.
The report also said ByteDance has approached a number of top assisted-driving or autonomous driving talents, and some technical talents who are optimistic about the move plan to join.
In response to the report, ByteDance said its frontier large-model exploration areas include physical AI, where there are many early-stage research efforts and explorations, but it has no plan to build a smart driving business.
The response shows ByteDance is still trying to define the moves as technology research, rather than a clear commercial smart driving business, 36Kr noted.
Seed is ByteDance’s fundamental research team for large models. It was established in 2023 and is a first-tier strategic department within the company.
One of the department’s core AI research and development leaders is Zhou Chang, who joined ByteDance in 2024 and is responsible for areas including multimodal large-model development.
Zhou’s management scope has since expanded. The visual generation business was placed under his responsibility last year, and the Seed Robotics team also started reporting to him this year.
That means Zhou’s current management scope covers areas including multimodal large models, world models, visual generation and embodied intelligence.
ByteDance is no stranger to the vehicle intelligence sector. The company has already entered the smart cockpit field through the Doubao large model and Volcengine.
In this push, Volcengine mainly provides automakers with AI technology, cloud services and cockpit interaction capabilities. Its role is closer to that of a technology supplier than a vehicle manufacturer.
Saidou Technology, backed by Seres Group (HKEX: 9927), announced a deep partnership with Volcengine when it launched the Aiva brand in June.
Aiva said Volcengine will provide it with core technology services including the Doubao large model and smart cockpit, to improve in-vehicle intelligent interaction.
However, Aiva’s assisted-driving solution is expected to come from Chinese startup DeepRoute, not ByteDance or Volcengine, according to a June 3 report by another local media outlet, Lanjinger.
World models are becoming an important technology direction in the autonomous driving industry, and they also provide ByteDance with a technical entry point for crossing into the sector.
Over the past year, the industry has debated the technology paths of VLA (Vision-Language-Action) and world models. But more companies are shifting their focus to world models, or to architectures close to world models, 36Kr noted.
For ByteDance, autonomous driving may also become a training ground on the path toward embodied intelligence.
Physical AI has become a key theme in the technology industry this year. Autonomous driving and robots are widely seen as the first two scenarios to be commercialized.
If ByteDance can obtain enough traffic data, it may use the data to iterate its world models, and later reuse the capabilities in embodied-intelligence scenarios such as robots, according to 36Kr.
Tesla (NASDAQ: TSLA) CEO Elon Musk has emphasized the importance of FSD autonomous driving data to the development of the Optimus robot. The core logic is that vehicle data pipelines can help robots understand the real world.
For the autonomous driving industry, ByteDance’s potential entry could bring a new competitive variable.
China’s autonomous driving industry has gone through multiple rounds of financing, mass production and reshuffling over nearly the past 10 years. Now, the basis of competition is shifting from engineering rules to AI models.
Under the AI paradigm, platform companies with computing power, talent and model capabilities may be better able to narrow the gap with traditional players.
36Kr cited an autonomous driving industry source as saying ByteDance’s advantages lie in having sufficient capital, talent and computing resources.
As of press time, Momenta’s stock price had risen by about 6%, bringing its total market capitalization to 73.6 billion Hong Kong dollars.
