A new approach to building AI computing infrastructure has arrived in China, with the launch of what developers describe as the world’s first prefabricated computing power hub in Qingdao, Shandong Province. According to CCTV News and the Global Times, the facility, which is developed by Qingdao TGOOD Electric, promises to cut electricity costs per token by 30 percent, reduce construction costs by nearly 80 percent, and enable 100 percent green energy consumption, addressing three of the most pressing constraints on China’s rapidly expanding AI data center industry.
The facility was reported by CCTV News and the Global Times on June 7, with additional details from Science and Technology Daily. Pan Helin, a member of the Information and Communication Economy Expert Committee at the Ministry of Industry and Information Technology (MIIT), provided technical commentary on the design.
What Makes It Different from a Conventional Data Center Hub
Traditional computing power hubs, the energy infrastructure that powers data centers, are built on-site using conventional construction methods, a process that is slow, expensive, and difficult to standardize. The Qingdao hub takes a fundamentally different approach: all construction details are finalized through virtual simulation before a single physical component is assembled, and the facility is then built with prefabricated modules integrated on-site.
This method shortens the construction cycle to approximately five months, compared with the much longer timelines typical of conventional builds. It also reduces land use by more than 30 percent and cuts overall costs by 20 percent. The near-80 percent reduction in construction costs specifically refers to the hub infrastructure itself, the power supply and energy management systems, rather than the full data center build.
“This approach shortens the construction cycle, accelerates computing power deployment, and reduces costs, enabling large-scale rollout,” Pan Helin told the Global Times. “Through scientific planning, prefabrication, and integrated installation, a mature investment model has been established, significantly easing the token supply-demand imbalance.”
Green Power and Reliability at the Core
The hub’s most commercially significant feature is its direct green power connection, which enables 100 percent renewable energy consumption. By integrating energy storage for peak load shifting, computing-power collaboration, and intelligent operation and maintenance, the system achieves the 30 percent reduction in electricity cost per token that TGOOD has highlighted as its headline figure.
Zhou Jun, executive president of Qingdao TGOOD Electric, explained that the hub ensures each device is connected to three independent power sources, providing what he described as “excellent electricity reliability.” The design is engineered to withstand external grid fluctuations and absorb the load impacts of GPU computing, ensuring stable operation even in the event of equipment failure.
For AI data centers, power continuity is existential rather than merely desirable. A training run that takes days to complete can be irretrievably lost in a momentary outage; inference workloads serving millions of users can be disrupted by fluctuations that a conventional office building would never notice. The hub addresses this by implementing multiple layers of redundancy at the circuit level, so that no single point of failure can interrupt power to the computing load.
Timing: A Market Hungry for Cheaper Computing
The launch comes at a moment when China’s AI computing market is undergoing rapid repricing. Several companies have recently announced permanent price reductions for computing power, with DeepSeek and Xiaomi MiMo among those slashing prices, in some cases by nearly 99 percent. China’s three major telecom operators have already launched token packages for individual consumers, signaling that AI compute is moving toward a commodity model.
The prefabricated hub fits directly into this dynamic. By reducing the capital and operating costs of building new computing capacity, it lowers the floor price at which new entrants can offer competitive token rates, accelerating the broader trend toward cheaper, more accessible AI infrastructure.
The hub has already been connected to an enterprise’s self-built data center and is expected to be deployed in national-level data center clusters and regional computing power centers in the second half of 2026, providing, as TGOOD describes, an efficient, low-carbon, and rapidly replicable model for computing hubs at various scales.
Infrastructure as a Strategic Lever
The Qingdao hub is part of a larger infrastructure push reshaping China’s AI compute landscape. China has built 42 intelligent computing clusters at the 10,000-card scale, and the country’s cities are now racing to establish space-based computing infrastructure as the next layer of the stack. The prefabricated hub model, if it scales as TGOOD projects, could become a standard template for rapid, low-cost data center deployment across China’s national computing power network, and potentially for export to markets in the Global South where AI infrastructure investment is accelerating.