Infrastructure Shifts in the Token Economy
ZTE Corporation hosted the 2026 Global Intelligent Computing Ecosystem Congress in Kuala Lumpur on September 10, 2026. This gathering focused on the transition of artificial intelligence from basic model development to large-scale deployment. As organizations move from training models to active inference, the industry is recalibrating its focus toward efficiency. The central metric for this new phase is the Token. Computing power demand now revolves around how efficiently systems process these units while maintaining performance.
Zhang Wanchun, a Senior Vice President at ZTE, opened the event by framing AI progress through the lens of Tokenomics. He argued that the old models of hardware procurement are insufficient for the current demand. Instead, the company advocates for open decoupling and system-level synergy. This approach seeks to provide end-to-end solutions for sectors like telecommunications, finance, and healthcare. For ZTE, the goal is to drive down the cost of each processed token to make AI accessible for wider industrial adoption.
The Engineering of Industrial Synergy
Reducing token costs requires more than hardware upgrades. Technical leaders at the congress emphasized that the process involves a tight integration of chips, software, and networking clusters. Guoliang Sun from MetaX noted that computing is now a systematic endeavor rather than a simple hardware trade. MetaX has introduced a 1+6+X strategy to map its proprietary GPU architecture against specific business needs in fields like energy and transportation.
Intel Malaysia provided a supporting perspective on the hardware side. Tan, Lai Teng highlighted the role of Intel Xeon processors in maintaining infrastructure that serves both traditional and generative AI tasks. The consensus among speakers was that individual components cannot solve the scaling problem alone. Collaboration between silicon providers, network architects, and software developers is the only way to manage the surge in inference demand.
Sovereignty and Local Application
Cloud integration remains a critical pillar for the future of AI. Jayson Li of Alibaba Cloud discussed how local data sovereignty acts as a major driver for modern infrastructure. By combining Alibaba Cloud Apsara Stack with ZTE systems, enterprises can run sovereign AI models that stay rooted in local markets while remaining connected to larger cloud operations. This structure allows businesses to turn raw compute assets into sustainable operational services.
Intelligent agents are the next step in this evolution. Garry Sien from Ant Digital Technologies explained that businesses are moving past simple chatbots toward systems that execute complex tasks. The Agentar 2.0 platform currently offers hundreds of modular skills for this purpose. These tools allow companies to assemble automated production lines for digital work. The congress concluded with a shared vision from participants, including the Hong Kong Generative AI Research and Development Centre, to build a global ecosystem that stays open, inclusive, and tied to tangible industry value.