Hong Kong is accelerating plans to expand its artificial intelligence computing capacity as leading Chinese chip executives warned that the rise of autonomous AI agents will trigger an unprecedented surge in demand for processing power. At an AI summit in the city, Moore Threads founder Zhang Jianzhong described token consumption in the era of AI as “far beyond our imagination,” underscoring industry concerns that current infrastructure may struggle to keep pace.
What Happened
At a high-profile AI summit in Hong Kong, senior figures from China’s chip sector sounded an alarm over rapidly growing computing needs driven by new AI technologies. Zhang Jianzhong, founder and CEO of graphics processing unit designer Moore Threads, said token consumption in the era of AI is “far beyond our imagination.” The summit highlighted forecasts from industry leaders that autonomous AI agents — systems that can act, learn and make decisions with limited human oversight — will cause a sharp rise in demand for GPUs and other specialist processors.
Responding to those warnings, Hong Kong officials and private-sector actors are moving to expand data-centre capacity and other AI computing infrastructure in the city. While organisers and executives at the summit emphasised the scale of the coming demand, details on specific projects or timelines were not disclosed in the report.
Background
The past several years have seen rapid advances in large language models, generative AI and other machine-learning systems. More recently, attention has shifted to autonomous AI agents — software systems that can manage tasks, interact with users or other systems, and pursue goals with varying degrees of autonomy. These agents often require sustained processing over long sequences of tokens, driving up consumption of compute, memory and specialized accelerator resources such as GPUs and tensor-processing units.
Graphics processing units, originally developed for graphics, are now central to AI model training and inference because of their parallel-processing capabilities. Companies like Moore Threads design GPUs tailored for such workloads. As models grow in size and as use cases multiply across industries, demand for datacentre capacity, power, cooling and high-bandwidth networking has increased alongside the need for chips and accelerators.
Hong Kong has sought in recent years to strengthen its technology and innovation profile, leveraging its financial markets, connectivity and proximity to the Chinese mainland. Expanding AI computing capacity is consistent with that goal: more local compute infrastructure can support research, startups, financial services using AI, cloud services and partnerships with regional technology firms.
Why It Matters
The warnings from chip executives and the resulting push in Hong Kong matter for several reasons. First, an accelerated global demand curve for AI compute can create pressure on chip supply chains, potentially influencing availability and pricing for GPUs and related components worldwide. That, in turn, affects cloud providers, enterprises and research institutions that rely on scalable access to accelerators.
Second, the race to build AI-capable infrastructure is becoming a strategic factor for cities and regions that want to host AI development, startups and data-intensive industries. Hong Kong’s move to expand capacity aims to keep the city competitive as firms weigh where to locate compute resources and where to run latency-sensitive applications.
For Panama and Latin America, the development is relevant as part of a broader global reshuffling of cloud and data-centre investment. If demand for AI compute concentrates in particular hubs, multinational cloud and hardware providers may prioritize capacity and partnerships in those regions, which could influence service availability and costs for Latin American businesses relying on global cloud platforms. Moreover, heightened competition for chips and accelerators can affect procurement timelines and prices for local research institutions and companies seeking to deploy advanced AI services.
Finally, rapid growth in AI compute underscores practical challenges: building and powering large data centres requires significant electricity, cooling and real-estate commitments. Cities that attract such investments will need to manage infrastructure, environmental and regulatory implications as they expand capacity to meet new technological demands.
As industry leaders continue to signal a steep increase in computing needs, Hong Kong’s push to scale up capacity highlights the broader, global race to prepare for an AI-driven future — a race with implications for technology supply chains, regional competitiveness and where advanced AI workloads will run.