• Huawei allocates $2.5bn for smart-driving R&D in 2026, half earmarked for computing infrastructure.
  • Compute capacity becomes the competitive bottleneck in China's smart-driving market.

What happened

Huawei has announced a plan to invest more than $10 billion over five years to expand computing power dedicated to smart-driving development, according to the company. The focus is on infrastructure used to train autonomous driving models, which are increasingly central to advanced driver-assistance systems.

The company also said it will spend around 18 billion yuan in 2026 on smart-driving research and development. About 10 billion yuan of that will be directed specifically to computing infrastructure.

The announcement was made in Beijing ahead of a major auto industry showcase. Senior vice-president Jin Yuzhi confirmed the investment plan at the event.

Huawei also presented 38 vehicle models equipped with its smart-driving and intelligent cockpit systems. These include collaborations with automakers such as Audi and Toyota-linked projects. The firm continues to expand deployment of its Qiankun ADS driver-assistance platform across partner vehicles.

Why it's important

The investment highlights a shift in where competitive advantage is forming in smart mobility. The bottleneck is moving from vehicle hardware to the computing systems that train and run driving models. By scaling compute infrastructure at this level, Huawei is not only supplying automotive technology but also shaping the pace at which autonomous driving systems can improve—embedding itself in the infrastructure layer that underpins model development and deployment.

At the industry level, this reflects a broader transition in China's auto sector, where competition is increasingly defined by software capability and AI performance rather than electrification alone. The companies that control compute capacity gain an advantage in iteration speed, system integration, and long-term platform influence.

Also read: Verizon brings private 5G to remote reality TV production

Also read: DeepSeek V4 runs on Huawei chips