Summary
Kazakhtelecom has deployed 880 NVIDIA GPUs in Astana and Almaty, while Firebird is building an AI data-centre project and has secured a U.S. export licence covering 45,000 Grace Blackwell GPUs. The announced plans include a commercial launch target in 2027 and a possible second phase.
NVIDIA says Kazakhstan’s AI infrastructure is expanding from deployments already in operation to larger planned data-centre capacity. Kazakhtelecom has deployed 880 NVIDIA GPUs in Astana and Almaty, while Firebird has secured a U.S. export licence covering 45,000 NVIDIA Grace Blackwell GPUs for a project whose first phase is under construction.
The announcements outline several different stages of development: installed equipment, construction and future plans. They also describe infrastructure intended to support AI models and applications developed in Kazakhstan.
Deployed systems and projects under construction
Kazakhtelecom, which NVIDIA describes as Kazakhstan’s largest telecommunications company, has deployed 400 NVIDIA H200 and 480 NVIDIA B300 GPUs across Astana and Almaty. The company plans to deploy more than 1,000 GPUs at Data Center Valley in 2027, as well as more than 1,000 NVIDIA Vera Rubin GPUs in the Ekibastuz region.
Separately, Firebird has signed an investment agreement with the government of Kazakhstan. Its first project phase is being built with Kazakhtelecom at Data Center Valley, with commercial launch targeted for the third quarter of 2027. Firebird’s planned second phase would bring its total potential AI infrastructure capacity in Kazakhstan to approximately 500 megawatts.
These figures describe distinct milestones. Kazakhtelecom’s 880 GPUs are reported as deployed; Firebird’s first phase is under construction, and the 500-megawatt figure depends on a planned expansion. The export licence covers 45,000 Grace Blackwell GPUs, but the announcement does not describe those GPUs as already installed.
Building capacity for models and applications
GPUs are processors that can handle many calculations in parallel, a capability used in training and running many AI models. More computing capacity can give developers and organisations resources to train, fine-tune and serve models—the stages involved in developing systems and making them available for use. NVIDIA says Kazakhstan’s growing supply of accelerated computing is intended to support those activities at scale.
The announcements also connect infrastructure to efforts to build AI products locally. At a public discussion in Astana, Kazakhstan’s Deputy Prime Minister and Minister of AI and Digital Development, Zhaslan Madiyev, described work on Kazakh-language models designed to reflect the country’s language and cultural context. NVIDIA also says Kazakhstan-founded Higgsfield AI is expanding its infrastructure using NVIDIA Vera Rubin systems on cloud service providers, and that NVIDIA is making an investment in the startup.
Officials also discussed possible uses of AI and robotics in sectors such as mining, agriculture, transport and logistics. Those are areas identified as opportunities for future development; the announcement does not describe specific deployments there. The practical significance of the new compute therefore depends not just on GPU supply, but on how effectively local companies, researchers and institutions turn that capacity into models and useful applications.