Huawei predicts AI agents will dominate AI traffic by 2035

Huawei expects autonomous AI agents to account for more than 90% of global AI token traffic within the next decade, pointing to a sharp increase in the computing and networking capacity needed to support an agent-driven AI economy.

The Chinese telecom giant published the estimate in a research report released on September 16, 2026. The forecast comes as chipmakers, cloud operators, and enterprises assess how much infrastructure they may need to support the growth of AI agents, while U.S. regulators and executives debate whether the most powerful AI systems are advancing too quickly.

The forecast behind the headline

The projection comes from “Intelligent World 2035: Turning Vision into Action,” one of two reports Huawei unveiled during an online event themed “Act Today. Shape Tomorrow.”

In AI systems, a token is a small unit of text or data that a model processes as it reads and generates information. Huawei argues that AI agents could fundamentally change how token demand grows. Rather than a person clicking through an app, software agents can continuously monitor their surroundings, reason, make decisions, and call external tools, with each step using tokens at a much higher frequency.

As that usage scales, the demand for tokens could rise sharply. Huawei projects annual token consumption will increase 100,000-fold by 2035, with AI agents accounting for most of that growth.

A future of 900 billion agents

The traffic estimate is based on another Huawei projection that about 900 billion autonomous AI agents could be running simultaneously all over the world by 2035. To put this figure into perspective, it would amount to about 100 AI agents for each person expected to be living on the planet at the time.

Supporting that scale of AI agents is the focus of the report’s 10 “key directions,” which outline the engineering challenges Huawei expects to face in the agent-driven era. The company calls for computing clusters to scale 100-fold while cutting the cost of an agent task by 1,000 times, a shift it says will require moving beyond chip-level improvements toward what it calls SuperPoD computing systems.

The other priorities include storage and memory, an “Agent OS,” intelligent driving, data center power and cooling, and a chip-design approach Huawei calls the Tau Scaling Law. Security and privacy also feature among the 10 directions.

Deliberate contrast with the US debate

Huawei’s position is notable for what the report leaves unsaid. Some U.S. researchers and tech executives have called for a slower pace of AI development after cases of agents behaving unpredictably and warnings that they could take control of equipment or access resources on their own.

The Huawei report views the spread of AI agents as a certainty and focuses on how to make them secure, instead of looking toward a pause in development. The company is treating this as an engineering challenge, and not as a reason to slow down.

Beijing wants to accelerate AI adoption across the economy while developing technical standards to address the risks that come with it. Huawei is already working on the security side, having launched an autonomous platform for security operations centers in April that uses monitoring and response agents to tackle cyberthreats.

What Huawei’s leadership is signaling

Guo Ping, Huawei’s supervisory board chairman, described AI as Huawei’s biggest opportunity and said its computing infrastructure should become China’s answer to Nvidia.

David Wang, Huawei’s Deputy Chairman of the Board and Rotating Chairman, said in the report that agentic AI is a key part of the transformation ahead and that, while the direction is clear,32 turning that vision into reality will require concrete action.

Huawei also sees moving AI beyond software and into the physical world as a key step toward artificial general intelligence (AGI), which refers to AI systems that can handle a wide range of tasks at a level comparable to or beyond humans.

The company’s other report, the Global Digitalization and Intelligence Index 2026, surveyed 90 countries and estimated that AI could generate $27 trillion in economic value over the next five years.

If you're reading this, you’re already ahead. Stay there with our newsletter.