Huawei's "Intelligent World 2035" report, in its Trend 9, projects 9 billion people and 900 billion AI agents by 2035 — roughly one hundred agents per person. The report also forecasts that these agents will handle 90% of AI token traffic and drive a 100-fold increase in handled data volume.
Key figures and sources
- Source: Huawei, "Intelligent World 2035", Trend 9.
- Population and agents: 9 billion people, 900 billion AI agents (by 2035).
- Traffic and data: agents expected to carry 90% of AI-token traffic and cause a 100× rise in data volume.
- Corroboration: Reuters reported in the same week that agents will move more than 90% of AI token traffic by 2035.
- Industry investment: hyperscalers have already committed $725 billion of capex for 2026, a 77% year-on-year increase.
Why this matters
Comparing the projections and the investment commitments highlights the relationship between deployed capacity and anticipated usage patterns. The reported capital expenditures are already committed and the depreciation clock starts when capacity is put in place. That implies industry spending is not solely predicated on increased human demand; instead, it anticipates machine-to-machine demand and activity that would consume the built capacity.
Operational logic: humans versus machines
The expected capacity requirements should be weighed against human usage limits — for example, human reading and attention impose practical ceilings on how much information a person can consume. If returns on infrastructure investment relied only on human consumption, those physical limits would constrain feasibility. An alternative scenario in which investments can pay off is one where agents transact with other agents, buying compute and data from machines — making machines the primary customers driving demand.
Implications for the industry
- The scale of investment ($725 billion for 2026 and 77% annual growth) indicates hyperscalers are building capacity to handle anticipated machine-driven traffic.
- If agent-driven traffic becomes dominant, infrastructure design, pricing and operations will prioritize machine-to-machine calls, automated scaling and transactional patterns between agents.
- Human-centered applications and user experiences will remain important, but the business models underpinning current capex appear to assume significant activity from machine actors rather than only expanded human usage.
What to watch next
- The 2035 horizon is long; technological advances, regulation and market dynamics can alter the path.
- Track the actual split of traffic between human and machine consumers over the coming years to assess whether the forecasted agent-dominant scenario emerges.
- Monitor how committed investments are executed and how resulting depreciation and utilization affect pricing and service architectures in the short and medium term.
Conclusion
Huawei's "Intelligent World 2035" Trend 9 and a contemporaneous Reuters report both anticipate a future where AI agents account for the vast majority of token traffic and drive massive increases in data volume. Significant industry capex commitments suggest the cloud and infrastructure sector is already positioning for such a scenario, implicitly betting on machine-to-machine demand alongside — or in some respects instead of — direct human-driven growth.
