Two reports released by Huawei ahead of Connect 2026 lay out the company's vision for the coming decade and reveal the strategic thinking behind its latest product launches. The forecasts paint a picture of explosive growth: global annual token consumption will increase 100,000-fold by 2035, with agents responsible for generating more than 90% of that traffic. Over the next five years, AI is projected to add $27 trillion in economic value. The ten strategic directions outlined in these reports function less as abstract vision and more as a detailed product roadmap.
In the foreword to Huawei's latest forecast, David Wang stated: "Visions are painted in words, but measured in deeds." The company published two separate reports days before opening Huawei Connect 2026, and together they account for the majority of what the company unveiled at the event.
The first document, titled Intelligent World 2035: Turning Vision into Action, was developed alongside the Global Digitalization and Intelligence Index 2026, which was created in collaboration with the Institute of Economics at Tsinghua University. Huawei has maintained this forecasting series for three years, with the 2024 edition mapping emerging trends and the 2025 edition identifying ten megatrends. This year's edition identifies ten strategic directions and specifies what infrastructure and capabilities must be developed to realize them.
Wang emphasized the centrality of autonomous systems to this transformation: "Agentic AI is a key variable in this transformation. The direction is clear, but bringing our vision to life demands concrete action."
The foundational projection
A single forecast underpins all others in Huawei's analysis. The company projects that worldwide annual token consumption will expand 100,000-fold by 2035, with agents responsible for generating more than 90% of this traffic.
Huawei distinguishes between agentic AI systems and the conversational assistants familiar to most users. A conventional chatbot waits for user input and responds to it. By contrast, an agent continuously perceives its surroundings, engages in reasoning, formulates plans, invokes tools, and adapts through learning—a substantially more computationally intensive operation.
This projection carries implications for power consumption as much as for software architecture. Each token represents a unit of model output, and producing it demands electrical power, memory bandwidth, and network resources.
The report characterizes a fundamental transition from an application-focused digital ecosystem to one centered on autonomous agents. In this new paradigm, software continuously monitors its environment, performs analysis, activates tools, and engages with users. Each of these operations produces tokens.
Chinese government policy is already moving in this direction. A state report from September indicated that China's AI sector is transitioning from a focus on model development to agent deployment, with computational demands shifting from the training phase to the inference phase.
Ten directions: reading the roadmap
The ten strategic directions are the most revealing component of the forecast because they function as a product strategy rather than abstract aspiration.
The broadest direction asserts that achieving artificial general intelligence requires physical embodiment and simultaneous progress in language intelligence, embodied intelligence, and scientific intelligence. A separate direction positions vehicles as mobile embodied agents.
Huawei sets ambitious infrastructure targets: computing clusters must scale by 100 times, and the expense of running an agent task must decrease by a factor of 1,000. The company argues that SuperPoD-style systems architecture is essential to achieving these goals—precisely the technology Huawei promoted throughout Connect 2026.
Another direction addresses storage and memory systems that maintain causal accuracy and verifiable provenance. This rationale supports the context memory storage cluster Huawei introduced in Shanghai.
The Tau Scaling Law, a chip design methodology Huawei revealed in May, represents a third direction. This approach proposes substituting time scaling for geometric scaling, a distinction that matters significantly for a company facing restrictions on access to cutting-edge manufacturing technology.
Two additional directions merit attention. One addresses the need to overcome power and thermal constraints in AI data centers. The other focuses on security and privacy safeguards for autonomous agents.
A final direction covers Agent OS, which Huawei expresses through a mathematical formulation: coordination, execution, memory and connectivity multiplied together, then raised to the power of evolution.
The economics projection
The second report addresses the financial dimension. Huawei and Tsinghua project cumulative AI economic value exceeding $27 trillion over the coming five years.
The forecasts also anticipate that worldwide spending on digital and intelligent infrastructure will expand at a compound annual growth rate of 19.14%, reaching $4 trillion by 2030. The index evaluates 90 countries and categorizes them into three tiers: Builder, Adopter, and Frontrunner.
The analytical framework itself is noteworthy. The report redefines the essential factors of production for a digital economy as data, ICT talent, and digital and intelligent technologies.
The analysis then contends that combining networking, computing, and storage capabilities is what enables industrial transformation. Since Huawei manufactures all three categories, this framing warrants careful consideration when interpreting the findings.
Interpreting vendor forecasts
These projections are not neutral analyses. A 100,000-fold expansion in token consumption would require vast amounts of new infrastructure, and Huawei manufactures that infrastructure.
The same skepticism should apply to every agent-economy projection currently circulating, including those from American technology companies. Previous examinations of these projections have revealed methodological weaknesses.
Vendor forecasts are most valuable when read as windows into the vendor itself. These reports reveal what Huawei believes, what it is building toward, and where it perceives its constraints.
The signals are unmistakable. The Tau Scaling Law direction addresses sanctions restrictions. The power and thermal direction responds to grid limitations. The cluster scaling direction compensates for per-chip performance limitations the company currently faces.
Europe's position
The index organizes countries into three categories and argues that no single pathway exists toward an intelligent economy. The stated goal is to enable each nation to identify its own priorities and chart a course aligned with its particular circumstances.
Europe faces a computational capacity deficit, a reality its own leadership regularly acknowledges. ECB president Christine Lagarde stated this month that Europe must develop its own AI infrastructure or face the risk of technological isolation.
Huawei is positioning itself to address this gap. The company's latest AI cluster service becomes available to markets outside China on 30 November, with its enterprise agent platform launching on 30 December.
Huawei's report concludes with five stated commitments: maintaining a long-term perspective, keeping people at the center, advancing technology for social benefit, remaining open to collaboration, and adopting systems thinking. These commitments mirror those made by virtually every major technology vendor.
The distinguishing factor is the hardware backing these commitments. Huawei stands among a small number of companies publishing a decade-spanning forecast while simultaneously delivering most of the equipment that forecast describes.
Source: The Next Web
