Huawei Publishes 2035 AI Infrastructure Outlook, Highlighting Power and Cooling Constraints
Редакция ЦОД:Главное
Huawei has published its “Intelligent World 2035: From Vision to Action” study, forecasting a 100,000-fold increase in global annual token consumption by 2035. The company identifies larger compute clusters, supernode-based architectures, and power and cooling solutions for AI data centers as core development priorities.
- Huawei forecasts a 100,000-fold rise in global annual token consumption by 2035.
- Agentic AI is projected to account for more than 90% of token traffic by 2035.
- The study calls for a 100-fold expansion in compute-cluster scale and greater use of supernode-based architectures.
- Power supply and cooling are identified as constraints to scaling larger AI data centers.
- GDII 2026 forecasts 19.14% CAGR in global digital and intelligent infrastructure investment, exceeding US$4 trillion by 2030.
Huawei presented its “Intelligent World 2035: From Vision to Action” study during an online presentation on September 16, alongside the Global Digital and Intelligent Index (GDII) 2026, produced with the Institute of Economics at Tsinghua University.
Huawei forecasts that global annual token consumption will increase by 100,000 times by 2035, with agentic AI accounting for more than 90% of token traffic. The study also sets out a 100-fold expansion in the scale of compute clusters and a transition toward supernode-based architectures.
For data-center operators and infrastructure suppliers, the central message is the physical infrastructure required to support that scale-up. Huawei names the resolution of power-supply and cooling challenges for larger AI data centers among its ten development directions. The report therefore frames electrical capacity, thermal management, network capacity and compute architecture as interdependent constraints in the expansion of AI infrastructure.
Separately, GDII 2026 projects a 19.14% CAGR for global investment in digital and intelligent infrastructure, exceeding US$4 trillion by 2030. That investment category is broader than the data-center market and should not be read as a forecast for data-center investment alone.
Huawei’s projections are not evidence of commissioned capacity, funded data-center development or announced site-specific projects. However, if the anticipated move toward agentic AI occurs, demand could shift further from conventional cloud workloads toward high-density, highly scalable AI clusters, increasing the importance of energy-efficient compute architectures, network throughput, power systems and cooling in operator and supplier investment decisions.
The publication is a strategic outlook rather than a data-center project announcement, but it points to potentially much higher requirements for compute, networking, cloud infrastructure, electrical systems and thermal management. If the projected shift toward agentic AI materializes, high-density, scalable AI clusters could take greater weight in data-center investment decisions.
- TechNews 科技新報 / КитайскийПервоисточник →
Материал подготовлен редакцией ЦОД: Главное на основе указанных первоисточников.