On June 15, Kazakhstan took another major step in its large-scale AI push, signing a $10 billion package of agreements with Firebird and NVIDIA to build the Data Center Valley in Ekibastuz, including a planned cluster of 100,000 NVIDIA GPUs. The deal forms part of a broader strategy that began in 2025 with the creation of a dedicated Ministry of Artificial Intelligence and Digital Development, followed by the declaration of 2026 as the “Year of Digitalization and Artificial Intelligence,” the passage of Central Asia’s first comprehensive AI law, and the rollout of large-scale training programs and a National AI Platform. The country has also sought to raise its profile by hosting the region’s largest tech and AI event to date, the GITEX AI Central Asia & Caucasus forum.
Six months into the Year of AI and Digital Development, the question is not whether Kazakhstan has launched enough initiatives, but whether those initiatives are beginning to form a functional system.
Early signs of use are visible in public services, from crime forecasting to fraud detection. A presidential decree is pushing AI into secondary schools, with pilots expected this fall. However, like many digital and reform initiatives across the region, the gap between ambition and measurable outcomes remains wide. In most countries, government platforms that fail to open up to external developers and third parties in their early stages often struggle to develop into vibrant innovation ecosystems. So far, Kazakhstan has built the shell of such a system, but not yet the network effects that would make it self-sustaining.
Kazakhstan’s AI strategy is built around a centralized model. In 2025 the country created a dedicated Ministry of Artificial Intelligence and Digital Development, giving digitalization Cabinet-level status. An AI Development Council chaired by President Kassym-Jomart Tokayev provides political direction and has become a visible feature of the government’s digital agenda.
The main policy framework is the Concept for the Development of Artificial Intelligence for 2024–2029. It rests on three priorities: infrastructure, human capital, and regulation. The Data Center Valley is intended to anchor the infrastructure side and attract foreign investment and computing capacity.
This centralized model also reflects Kazakhstan’s desire to avoid fragmentation across ministries while presenting a coherent framework to international partners. At a time when China is expanding its AI and digital footprint across Central Asia, Kazakhstan is trying to capture investment without becoming dependent on any single external technology partner.
Among Central Asian countries, both Kazakhstan and Uzbekistan have AI legislation that entered into force in January 2026. Kyrgyzstan is developing its framework through a Digital Code, while Turkmenistan and Tajikistan have made significantly less progress. As Kazakhstan was the first to adopt a comprehensive framework, it has a modest lead in regulatory development. The open question is whether that lead translates into real investment and implementation or remains largely on paper.
The Kazakh AI law adopts a risk-based approach. It classifies AI systems according to their potential impact: higher-risk applications are subject to audits, while synthetic content must be clearly labeled. The law also places clear responsibility on both developers and operators. It gives formal status to the National AI Platform as a controlled environment for building, training, and testing systems before deployment. The law also creates channels for priority sectors to access government computing power and curated data libraries.
A roadmap was approved in January, and some subordinate rules including classification criteria have already been issued, but the harder implementation work is still ahead. For now, the law’s impact is mainly procedural: agencies and firms are adjusting compliance processes, and pilot projects have clearer legal ground. The real test will come with the remaining secondary legislation, which will determine how enforceable the framework actually is
However, many of the risks are embedded in the institutional design itself. While the AI law emphasizes principles such as transparency and accountability, operational mechanisms for auditing high-risk systems and ensuring meaningful oversight remain underdeveloped.
The National AI Platform is Kazakhstan’s main attempt to build sovereign AI capacity rather than depend entirely on foreign systems, and is meant to provide a secure state-controlled space for model development, training, and testing, while also managing data libraries and access to computing resources.
So far, the platform’s most visible activity has been inside the government. Several pilots are using AI tools to assist civil servants with routine tasks, and officials want to deploy around 50 such assistants by year’s end. Although the legal architecture is in place, broader access for universities, startups, and private companies is still limited. Computing capacity remains another bottleneck, and while the Data Center Valley is meant to address it, tangible results are still years away.
Kazakhstan has also begun developing its own large language models: KazLLM and Sherkala were released in 2024 and 2025, respectively, but for now both remain largely at the research and proof-of-concept stage, with limited evidence of meaningful real-world deployment.
Human capital is where Kazakhstan has moved fastest, with the AI-Sana program already introducing more than 650,000 students to foundational AI courses in its first stage. High enrollment figures do not necessarily translate into deep technical expertise, however. A second, more intensive phase targeting 100,000 participants is now under way, with a stronger emphasis on project work and AI entrepreneurship. Moreover, a new digital literacy program was launched in 2025 with the target of reaching a million citizens, although there is limited public information on its content, quality, or actual impact. In January, Kazakhstan became the first country in Central Asia to join OpenAI’s education initiative, securing 165,000 free ChatGPT Edu licenses, including 100,000 for teachers. While this represents a significant step in raising AI awareness in education, it is still better understood as a capacity-building gesture than as a structural transformation of the education system.
Physical infrastructure is also beginning to appear. The Alem.ai International AI Center in Astana has opened as a hub for research, training, and startup activity, while the Data Center Valley in Ekibastuz will, once built, significantly expand the country’s capacity for AI training and cloud computing. It’s unclear at this juncture whether broad exposure can be turned into deeper expertise, and whether those skills stay in the country rather than leak abroad.
Kazakhstan continues to face significant brain-drain challenges, with many of its top technology graduates moving to Europe, the United States, the Gulf, or other tech hubs in search of better opportunities. This is part of a broader problem across Central Asia, where much of the skilled talent is leaving because opportunities at home remain limited. In this context, large-scale training programs risk mainly preparing talent for opportunities abroad, rather than strengthening Kazakhstan’s own long-term technological development. In effect, currently the country is optimizing for participation metrics rather than capability retention, a strategy that produces impressive headline numbers but uncertain long-term returns.
Six months in, Kazakhstan has built the visible architecture of an AI ecosystem – laws, programs, and partnerships – but has not yet shown that this architecture can produce sustained, real-world impact. The strongest evidence so far is institutional rather than operational.
Infrastructure development has moved more slowly. The National AI Platform is being used for some government pilots, but it is not yet meaningfully open to universities or private developers. Major capacity projects remain in an early stage, and several secondary regulations are still pending.
Although Kazakhstan has established new rules, expanded training programs, and secured international partnerships, Astana has yet to produce clear evidence of AI-driven improvements in government operations or broader economic outcomes. The real test by the end of 2026 will be whether the National AI Platform becomes genuinely usable by actors outside the state, whether AI deployments deliver measurable gains in public administration or state-owned enterprises, and whether promised computing capacity moves from announcement to operation.
Without meaningful progress on these three fronts, Kazakhstan will have built a plausible institutional framework but not yet a functioning AI ecosystem. That would not mean the agenda has failed, but it would indicate that the country remains stronger at designing AI policy than at converting it into practical capability.
The implications go beyond technology. If progress stalls, Kazakhstan could gradually lose ground to Uzbekistan, which is also pushing an ambitious digital strategy. Real progress, however, would strengthen Astana’s position as the region’s leading technology hub and give it more influence when dealing with both regional powers and major technology companies.
