Huawei has made a bold prediction about the future of artificial intelligence, forecasting that approximately 900 billion AI agents could be active worldwide by 2035. If the projection becomes reality, autonomous AI systems could outnumber the global human population by roughly 100 to one, potentially transforming how the internet operates, how businesses function, and how digital infrastructure is built.

The prediction points to a future in which artificial intelligence moves beyond answering questions and generating content to become an active participant in everyday digital activities. AI agents could independently process information, reason through complex tasks, make decisions, interact with software tools, and coordinate with other agents with limited human intervention.

However, the scale of this projected growth also raises important questions about computing capacity, cybersecurity, privacy, accountability, and the growing geopolitical competition over AI infrastructure.

AI Agents Could Dominate Global AI Traffic

According to Huawei’s forecast, autonomous AI agents could account for more than 90% of global AI token traffic by 2035.

Unlike traditional chatbots that primarily respond to individual user requests, AI agents can potentially perform extended sequences of tasks. They may search databases, analyse documents, write and execute code, communicate with other systems, and repeatedly use AI models to complete objectives.

For example, an AI agent working for a business could monitor market conditions, analyse customer behaviour, prepare reports, identify operational problems, and recommend actions. Multiple agents could collaborate to manage different stages of a workflow.

These activities require repeated exchanges of information between AI models and software systems. As a result, widespread adoption could significantly increase demand for AI processing, data storage, networking, and cloud computing.

If Huawei’s prediction proves accurate, the internet of the future may carry substantially more machine-generated activity than human-generated requests.

Why 900 Billion AI Agents Would Require Massive Computing Power

Operating hundreds of billions of AI agents would present a major infrastructure challenge.

Each agent would require access to some combination of computing resources, AI models, memory, data storage, and network connectivity. Agents performing complex or continuous tasks could consume significant computing resources, particularly when they interact with other agents or process large volumes of information.

This could accelerate investment in AI data centres, specialised processors, high-speed networking, energy infrastructure, and distributed computing systems.

Technology companies would also need to improve the efficiency of AI models so that agents can complete useful tasks without unnecessarily consuming computing resources.

The challenge is not simply creating more AI agents. It is ensuring that these systems can operate reliably, securely, and affordably at a scale far beyond today’s typical AI deployments.

Security and Privacy Could Become Major Concerns

As AI agents gain the ability to make decisions and interact with external tools, controlling their behaviour will become increasingly important.

An AI agent connected to business databases, financial applications, communication platforms, or critical infrastructure could potentially cause serious problems if it is compromised, misconfigured, or given excessive permissions.

Security risks could include unauthorised access to sensitive information, automated cyberattacks, data exposure, malicious instructions, and the rapid spread of errors across connected systems.

A compromised agent could also interact with other agents, potentially allowing a single security weakness to affect multiple systems.

To address these risks, organisations will need stronger identity verification, access controls, continuous monitoring, audit trails, and safeguards that restrict what agents can do without human approval.

Privacy will be equally important. AI agents that continuously process personal, commercial, and operational information must be designed to protect sensitive data and comply with applicable privacy requirements.

The central challenge will be finding a balance between autonomy and oversight: allowing AI agents to work efficiently without giving them unrestricted control over important systems.

Huawei Sees Physical AI as Part of the Road to AGI

Huawei also believes that bringing AI into the physical world will be an important part of the journey towards artificial general intelligence, commonly known as AGI.

AGI generally refers to a hypothetical form of AI capable of learning and performing a broad range of intellectual tasks at a level comparable to, or potentially exceeding, human capabilities. There is no universally accepted test for AGI, and its development remains uncertain.

Physical AI involves applying artificial intelligence to systems that interact with the real world, including robots, industrial machinery, autonomous vehicles, and other intelligent devices.

These systems must do more than process text or generate digital content. They may need to perceive their surroundings, understand changing conditions, plan movements, and respond safely to unexpected situations.

For example, AI-powered robots could support manufacturing, warehouse operations, logistics, agriculture, and healthcare. In industrial environments, autonomous agents could help coordinate equipment, identify faults, and optimise production processes.

Although physical AI could expand the practical applications of artificial intelligence, it does not guarantee that AGI will emerge by a particular date.

Huawei’s AI Ambitions and the Geopolitical Competition

Huawei’s prediction also has implications for the global technology race, particularly the competition between China and the United States.

Huawei has become a major focus of US national security and technology policy. The company faces extensive US restrictions, including limits affecting its access to certain American technologies and its participation in US telecommunications markets.

As AI agents become more capable and widely deployed, the companies that supply the underlying infrastructure could gain considerable economic and strategic influence.

That infrastructure includes AI processors, cloud platforms, networking equipment, data centres, operating systems, and security technologies. Countries that develop competitive domestic capabilities may seek to reduce dependence on foreign suppliers and strengthen control over critical digital systems.

For the United States, China, and other major technology powers, the issue extends beyond building better AI models. It also involves determining who supplies the systems on which businesses, governments, and essential services depend.

If autonomous AI agents become widespread, questions about infrastructure ownership, cybersecurity standards, data governance, and international technology dependence could become even more consequential.

However, Huawei’s projection should be understood as a forecast rather than an established outcome. Whether the company or its competitors ultimately dominate this market will depend on technological progress, regulation, economics, security requirements, and global adoption.

What Could This Mean for Businesses and Everyday Internet Users?

A world increasingly powered by AI agents could bring significant benefits.

Businesses could automate repetitive administrative work, improve customer service, accelerate research, and coordinate complex operations. Individuals could use personal agents to organise schedules, compare services, manage digital tasks, and retrieve information from multiple sources.

Industries such as banking, education, healthcare, manufacturing, and telecommunications could also use specialised agents to improve efficiency and support decision-making.

At the same time, widespread automation could disrupt some jobs, increase demand for new technical skills, and create new forms of dependence on AI platforms.

There are also questions about accountability. If an autonomous agent makes a costly mistake, leaks confidential information, or takes an action that harms someone, determining who is responsible may become more complicated.

The long-term impact will therefore depend not only on how many agents are deployed but also on how they are designed, governed, and integrated into society.

The Internet Could Become an Environment Run by Machines

The prospect of 900 billion AI agents represents a possible shift in the internet’s fundamental operating model.

Today’s internet is largely built around human users accessing websites, sending messages, watching videos, and using applications. In an agent-driven environment, software systems could increasingly communicate with other software systems, negotiate tasks, exchange information, and execute transactions automatically.

People may eventually spend less time navigating individual applications and more time instructing AI agents to complete tasks on their behalf.

This could make digital services more efficient, but it could also create an internet that is harder for humans to monitor. Automated systems might generate enormous volumes of activity at speeds that exceed traditional methods of supervision.

The result could be a more productive and intelligent digital ecosystem, provided that the necessary security, transparency, and governance mechanisms develop alongside it.

Conclusion: A Bold Forecast With Major Implications

Huawei’s projection of 900 billion AI agents by 2035 highlights the scale of the transformation that autonomous artificial intelligence could bring to the global economy and digital infrastructure.

If the forecast approaches reality, AI agents could become major consumers of computing resources, reshape how businesses operate, and intensify competition over the technologies that power the digital world.

Yet the number of agents alone will not determine the future. Their capabilities, levels of autonomy, resource consumption, security protections, and real-world usefulness will matter just as much.

The central question is no longer simply how intelligent AI systems can become. It is also how society will manage billions of interconnected autonomous systems and ensure that their benefits outweigh the risks.

What do you think? Would a world with 900 billion AI agents make the internet more powerful and productive, or would it introduce new security and control risks that humanity is not yet prepared to handle? Share your thoughts in the comments.