The token value chain shows how the rise of agents forces a reengineering of communication and computing networks

Technology markets treat the AI race as an exclusive battle to stack GPUs in data centers, but the rapid shift from traditional language models to autonomous agents has begun to expose cracks in that view. In a session on global digital infrastructure at the AI for Good 2026 summit organized by the International Telecommunication Union in Geneva, the senior representative for the international ecosystem at ZTE, Samer Chen, presented a vision that redefines AI infrastructure as a basic utility that must be made available to everyone like water and electricity, stressing that the growing computing demand driven by the rise of agents requires the telecom sector to go beyond merely providing connectivity toward integrating networks with processing capabilities directly.

Chen breaks down the AI ecosystem into a three-stage value chain: producing digital tokens, transporting them, and then consuming them. The production stage includes data centers and computing facilities that run the models, while the transport stage concerns the networks that carry AI-generated data between computing sites and users, leading to the consumption stage that encompasses devices, applications, industrial systems and public services.Building massive data centers remains of limited value without telecommunications networks capable of distributing their services efficiently and reliably.Likewise, connection speed loses its meaning if users lack suitable devices or the financial means to access them.

The production of tokens poses a sharp challenge for global power grids, as computing centers demand massive electrical densities that not all geographic regions can supply. According to the summit presentation, reducing emissions and improving energy efficiency is not a supplementary option but a critical operational pillar that includes liquid cooling, the design of more efficient power systems, and achieving a joint, integrated improvement that brings together chips, hardware and software. Chen notes that merely adding more processing units inside a center does not solve the crisis, pointing to the green-energy project developed by ZTE in partnership with Tencent to cut electricity consumption, while warning that renewable sources such as wind and solar, despite their importance, face difficulty in meeting the steady, continuous demand of hyperscale data centers.

The equation worsens at the transport stage given the uneven global distribution of computing capacity, with the bulk concentrated in China, the United States and Europe, while large regions suffer from digital access gaps. Addressing this gap requires a hybrid transport strategy that blends fiber optics, fixed wireless networks, mobile communications infrastructure and satellite solutions to suit desert, rural and island environments, a model the Chinese company has deployed in African markets to expand coverage and ensure communities do not fall behind in digital transformation.

The circle completes at the consumption stage, where intelligent capabilities move into AI-enhanced glasses, dedicated smartphones, and embedded systems in industrial production lines for process monitoring and operational efficiency improvement.Integrating AI into the core of industries requires digital security, compliance and ethical controls to become an intrinsic structural component of the hardware.This mirrors the historical experience of the telecommunications sector, which built its global systems by standardising specifications and frequency bands under the coordination of organisations such as the International Telecommunication Union and the United Nations.

This shift carries direct strategic implications for engineering and investment decision-makers in our region. In the Gulf markets, which have poured massive investments into purchasing processors and building computing complexes, the chain forces a reallocation of budgets toward interconnection networks and edge exchange points; investment in swarms of smart agents for banking or logistics will falter if networks continue to suffer from fluctuating latency, and the harsh climate makes the adoption of liquid cooling and joint software optimisation a urgent economic necessity to control cooling costs and high energy bills. In Egypt and the Levant, where capital constraints and electricity networks limit the hosting of giant complexes, this model enables technical teams to focus on token-consumption architecture by building applications and agents that rely on integrated models and low-bandwidth edge processing, thereby reducing dependence on costly cross-border cloud calls and lowering actual operating costs.

Infrastructure is not a hidden technical layer; it is the actual benchmark that determines where AI can be deployed and who benefits from its economic returns. Building this foundation requires moving beyond the logic of purchasing individual processors toward constructing integrated, reliable systems that link computing with networks and balance energy efficiency with equitable access.