en.Wedoany.com Reported - On July 13, Kazakh Prime Minister Olzhas Bektenov reported to President Kassym-Jomart Tokayev on the progress of the "Data Center Valley" project. Kazakhstan is collaborating with global technology companies, including U.S. artificial intelligence infrastructure firm Firebird and U.S. chipmaker NVIDIA, to build an initial 250MW data center and AI computing infrastructure. The scale of related project agreements has exceeded $10 billion.
The "Data Center Valley" is planned to be located in Ekibastuz, Kazakhstan, with the data center built near the Ekibastuz State District Power Plant No. 1. Placing high-power-consuming data center clusters near large power sources can mitigate the impact of long-distance power transmission on project costs and power supply stability, providing foundational conditions for the continuous operation of servers, GPU clusters, cooling systems, and network equipment.
According to the currently disclosed construction plan, the project is not building an isolated data center but forming a scalable data center cluster. The project site currently has approximately 300MW of available power capacity, with subsequent plans to gradually increase the carrying capacity to 1GW. The 250MW mentioned in the government's report refers to the scale of data center infrastructure currently under development and does not represent the final capacity of the entire campus.
A data center cluster of this scale requires the simultaneous construction of server buildings, substations, medium and high-voltage power distribution systems, backup power supplies, uninterruptible power supplies, cooling facilities, fire protection systems, and campus fiber optic networks. In addition to high-speed networks within the campus, the project also needs to improve international communication channels connecting major digital hubs in Europe and Asia to reduce data transmission latency for cloud computing and AI operations. Kazakhstan has designated international network connectivity as a key component of the project's construction.
The project plans to deploy approximately 100,000 next-generation GPUs, including NVIDIA GB300 and Vera Rubin architecture products, to form a large-scale AI computing cluster. Before GPU equipment enters the server rooms, the campus must first complete the construction of power capacity, server rack load-bearing, high-density power distribution, cooling, and data transmission systems. Compared to standard enterprise data centers, the centralized deployment of a large number of AI servers will significantly increase per-rack power and cooling requirements. Therefore, the power supply and cooling facilities will directly determine the final amount of usable computing power the project can achieve.
In terms of engineering progress, Kazakhstan has completed the cooperation framework and main technical condition alignment with U.S. Firebird and U.S. NVIDIA, and has begun advancing the 250MW infrastructure construction. U.S. Firebird has set a project goal to launch related computing capabilities by 2027, but the government has not yet officially disclosed the floor area of the first batch of data center buildings, the number of server rooms, construction contractors, or the power-on timelines for each phase.
Simultaneously, the Kazakh government is advancing the nationwide renovation of energy and public utilities. This year, it plans to rebuild 9 power generating units, 55 boilers, and 51 turbines, allocate 384 billion Kazakhstani tenge for related projects, and complete the repair of 12,000 kilometers of engineering pipeline networks by the end of the year. It should be noted that these funds and projects are part of the national energy and public utility modernization plan. The government has not stated that all of them are dedicated to the "Data Center Valley," and they cannot be directly counted towards the scale of the data center project's construction.
The core construction milestones for the next phase of the "Data Center Valley" will focus on the phased delivery of the campus power supply facilities, main data center buildings, cooling systems, fiber optic networks, and GPU computing clusters. Whether the project can achieve its first batch of operational computing power by 2027 will depend on whether the 250MW power infrastructure can be completed on schedule, and whether the server equipment, network systems, and server room environment can be commissioned simultaneously.
