The same week Anthropic's CEO was asking the industry to deliberately slow down, Huawei published a 2035 roadmap built entirely around the opposite instinct. Not one part of it asks anyone to pump the brakes. It asks how fast the brakes can be rebuilt to handle the speed everyone's already committed to.
What Huawei Actually Released
On September 16, 2026, Huawei published two research reports at once: "Intelligent World 2035: Turning Vision into Action" and the "Global Digitalization and Intelligence Index 2026." The first is a technical roadmap; the second, produced jointly with Tsinghua University's Institute of Economics, is an economic forecast and country-by-country readiness index.
| Detail | Figure |
|---|---|
| Release date | September 16, 2026, at an online launch themed "Act Today. Shape Tomorrow" |
| Token traffic growth by 2035 | 100,000-fold increase in global annual AI token consumption |
| Agent share of that traffic | Over 90% |
| AI economic value, 5-year forecast | Over $27 trillion in cumulative economic value |
| Digital infrastructure investment by 2030 | Surpassing $4 trillion, growing at a 19.14% CAGR |
| Countries assessed in GDII 2026 | 90, sorted into Builder, Adopter, and Frontrunner tiers |
Why Huawei Thinks Traffic Explodes 100,000-Fold
The core argument is a shift from an application-centric internet to an agent-centric one. Today's AI tools are mostly reactive, per Huawei's own framing: you prompt a chatbot, it responds. Agentic AI is different, continuously perceiving its environment, reasoning and deciding autonomously, invoking tools, and interacting with people in real time, more like a "digital employee" than a responsive tool.
Every one of those steps, perceiving, reasoning, tool use, interacting, generates its own token traffic, repeated continuously rather than triggered once per user prompt. That compounding is where the 100,000x figure comes from, and it's also why Huawei frames computing, networking, and cloud infrastructure as facing demands that "far exceed what traditional Internet applications demand."
The Ten Key Directions
Rather than a single big prediction, the bulk of the report is a list of ten specific technical fronts Huawei says need to advance together.
AGI: pursued through language intelligence, embodied intelligence (physical-world interaction), and scientific intelligence together, what Huawei calls "going physical" as the essential path forward
Computing clusters: scaling 100-fold while cutting the cost of running an agent task 1,000-fold, largely through SuperPoD systems, tightly interconnected server clusters with unified memory addressing and ultra-low latency
Storage and memory: new systems built for causal accuracy and traceability, moving from raw data storage toward something closer to structured knowledge and memory
Intelligent connectivity: networking infrastructure with AI built directly into the communications layer
Chip design (the "Tau Scaling Law"): a proposed alternative to Moore's Law, using time-based rather than geometric scaling to keep improving chip density and efficiency as physical shrinking hits its limits
Agent OS: an operating system built around agents as the primary actors rather than applications, summarized in the report as (Coordination × Execution × Memory × Connectivity) raised to the power of Evolution
Device intelligence: a framework for intelligent services that adapt seamlessly across whatever device a person happens to be using
Intelligent driving: self-learning systems pushing vehicles toward becoming, in Huawei's words, "mobile embodied agents"
Data center power and cooling: addressing the physical energy and thermal ceilings that limit how much AI computing power a facility can actually run
Security and privacy for autonomous agents: the report's dedicated safety category, focused on protecting systems that act independently rather than just responding to instructions
The Explicit Contrast With The American Pacing Debate
This report landed in the same week Anthropic CEO Dario Amodei published his call to deliberately slow the pace of frontier AI capability growth, and the contrast in posture is stark. Huawei's roadmap makes no call to slow development at all. It treats expanding agent capability and deployment as a given, and frames the real challenge as building the infrastructure, and the safety and privacy protections, needed to support that expansion safely at scale.
The pattern extends beyond these two documents. Parts of the US AI industry are actively debating coordinated limits on how fast frontier capabilities should be allowed to improve. China, by contrast, has generally prioritized accelerating AI adoption across its economy, while building technical standards around emerging risks rather than proposing to slow the underlying pace of development.
What Huawei Says Should Guide All Of This
The report closes with five stated principles rather than a single grand conclusion: take a long-term approach, remain people-centric, uphold "tech for good," stay committed to openness and collaboration, and adopt a systems approach. David Wang, Huawei's Deputy Chairman and Rotating Chairman, framed the whole exercise as a matter of follow-through rather than ambition: "Visions are painted in words, but measured in deeds."
The report arrives just ahead of Huawei Connect 2026, running September 17-19 in Shanghai under the theme "Advancing the Agentic World," where the company is expected to detail specific products built around these same ten directions.
Huawei's Intelligent World 2035 Report: FAQ
It's a research report Huawei released on September 16, 2026, titled "Intelligent World 2035: Turning Vision into Action," building on trend forecasts the company has published annually since 2024. For the first time, this edition lays out ten specific technical directions meant to serve as an actionable roadmap for building what Huawei calls an agent-centric, rather than application-centric, digital world.
The report predicts global annual AI token consumption will grow 100,000-fold by 2035, with traffic generated by autonomous agents accounting for more than 90% of that total. A companion report, the Global Digitalization and Intelligence Index 2026, separately forecasts that AI will generate over $27 trillion in cumulative economic value over the next five years, with global investment in digital and intelligent infrastructure surpassing $4 trillion by 2030.
They are: building AGI through language, embodied, and scientific intelligence; scaling computing clusters 100-fold while cutting agent task costs 1,000-fold; new storage and memory systems with traceable accuracy; experience-centric intelligent connectivity; a "Tau Scaling Law" for chip design; an agent-native "Agent OS"; adaptive device intelligence across scenarios; self-learning intelligent driving; breaking power and thermal limits in AI data centers; and security and privacy protection for autonomous agents.
It's a chip design principle Huawei is proposing as Moore's Law, which relies on physically shrinking transistors, approaches its physical limits. Instead of geometric scaling, the Tau (τ) Scaling Law proposes time-based scaling, using techniques like LogicFolding to compress signal propagation delay and increase transistor density without needing to keep shrinking transistors themselves.
It's a direct contrast. The same week Anthropic's Dario Amodei was calling for frontier labs to deliberately pace AI capability growth, Huawei's report makes no call to slow anything down, it assumes agent capabilities and deployment will expand dramatically and treats scaling that expansion safely, rather than slowing it, as the central challenge. The divide reflects a broader pattern: parts of the US industry are debating coordinated limits on capability growth, while China has generally prioritized accelerating AI adoption while building technical safety standards alongside it.
Jans Bock-Schroeder
Publisher & Founder of AI Angst
Coming from the world of art, photography, and the luxury market, Jans launched AI Angst in 2025 to explore the cultural, ethical, and psychological impacts of artificial intelligence. His work bridges creative vision with critical technology analysis, offering clarity in an era of rapid technological change.
Sources and Citations
This article is based on the following sources:
-
Huawei: "Huawei's latest report proposes 10 key directions to Intelligent World 2035"
Primary source, the official release covering all ten directions, the GDII 2026 figures, and David Wang's remarks.
https://www.huawei.com/en/news/2026/9/hc-intelligent-world-2035 -
PR Newswire: "Huawei's latest report proposes 10 key directions to Intelligent World 2035"
Cross-check of the report's launch details and quotes.
https://www.prnewswire.co.uk/news-releases/huaweis-latest-report-proposes-10-key-directions-to-intelligent-world-2035-302880708.html -
Yahoo Tech / Reuters: "China's Huawei forecasts billions of agents will dominate AI traffic by 2035"
Source for the direct comparison with the US AI pacing debate and broader US-China posture differences.
https://tech.yahoo.com/ai/articles/chinas-huawei-forecasts-billions-agents-125352858.html
Published: September 17, 2026. Sources verified at time of publication. All external links open in a new tab.
