ASTANA – Kazakhstan is moving rapidly to position artificial intelligence at the center of its economic modernization strategy, investing in computing infrastructure, skills, startups and Kazakh-language technologies as governments worldwide compete for a place in the emerging AI economy.
Photo credit: Shutterstock
Kazakhstan has an increasingly ambitious AI agenda. The government has made AI part of its broader economic modernization strategy, betting that it can improve public services, attract investment and help the country generate more value beyond its traditional reliance on raw materials.
It has been investing heavily in computing infrastructure, training, startup ecosystems, and Kazakh-language models. At the 38th plenary session of the Foreign Investors’ Council in July, President Kassym-Jomart Tokayev explicitly called the integration of AI into state investment policy a “critically important priority.”
The government has adopted a Digital Code and a law on artificial intelligence, launched the Digital Qazaqstan strategy, established a Ministry of Artificial Intelligence and Digital Development, and designated 2026 as the Year of Artificial Intelligence and Digital Development.
This push comes at a time as AI is becoming both a major global industry and an increasingly important source of economic and strategic power. UN Trade and Development estimates the global AI market will rise from $189 billion in 2023 to $4.8 trillion by 2033. This is a 25-fold increase in just a decade.
Globally, AI push is concentrated among a handful of countries. U.S. private AI investment alone reached nearly $286 billion in 2025, more than 23 times the level in China, according to Stanford University’s 2026 AI Index.
Global corporate AI investment more than doubled in 2025, while private companies produced more than 90% of the year’s notable AI models.
For emerging economies, the concentration of AI infrastructure in a small number of countries can make this race appear expensive, not to mention that training large language models requires massive quantities of computing power, electricity, connectivity and capital.
Kazakhstan is unlikely to compete with the United States or China at the most capital-intensive layer of the AI industry nor does it need to. Experts say its opportunity lies in combining global AI technologies with domestic expertise, while building its own capabilities in areas where dependence creates economic or strategic vulnerabilities.
Strong foundation
The country’s AI push builds on years of its experience of deploying digital technologies at a scale. Advanced technologies are increasingly embedded in government services and banking.
“Right now, especially in the artificial intelligence applications in fintech, Kazakhstan is one of the most progressive countries in the world,” Huseyin Atakan Varol, dean of the School of Computing and Artificial Intelligence and director of the Institute of Smart Systems and Artificial Intelligence at Nazarbayev University, told The Astana Times.
He pointed to technologies including palm-recognition authentication and AI-powered banking assistants as examples of how quickly digital tools have entered everyday life.
International indicators tell a similar story but also expose an important weakness. Kazakhstan ranks 10th globally for online government services, according to figures from the World Intellectual Property Organization. Yet it ranks 115th for university-industry collaboration and 112th for venture-capital deal activity relative to the size of its economy.
Dr. Varol said Kazakhstan still allocates a relatively small share of its economy to research and development, with the government providing most of that funding. While spending on science has increased in recent years, spending on research and development does not reach 1%.
“Nearly all of this allocation comes from the state, very little from the private companies,” he said. “So basically, industry, academia, government interaction, the triple helix structure is not yet fully working in Kazakhstan.”
That may prove to be one of the important constraints on Kazakhstan’s AI ambitions. While the government can establish institutions, purchase computing infrastructure and train specialistics, building a durable innovation economy requires more.
Alem.AI center in Astana is intended to help close that gap by connecting education, research, startups and companies.
“We have some real and simple goals. The first one is to build a community of professionals, who not just use but understand what AI actually is. It means that we train different categories,” Alyona Slavkina, head of the Alem.AI Development Office, told The Astana Times.
The community is supposed to include schoolchildren, students, working professionals, researchers and entrepreneurs. The idea is to create a pathway from first exposure to AI to education, research and eventually the development of commercial products.
The challenge, however, is that AI is evolving faster than conventional education systems. Olzhas Urazalin, head of Tomorrow School, said technologies and development tools can change within months, while formal academic programs often require years to adapt.
“Many graduates have strong theoretical knowledge, but very limited experience working with modern AI tools, real data sets, and product development. This indeed creates a gap between learning and creation. Young specialists often learn how to use existing AI solutions, but they rarely get the opportunity to understand how these systems are built, or how to develop new ones,” Urazalin said.
Tomorrow School is built around an alternative model. Students learn primarily through projects rather than conventional lectures and work on real problems brought into the curriculum by businesses. The program is free, runs for 18 months and is open to applicants without traditional technology backgrounds.
“In my opinion, Kazakhstan must focus on developing, first of all, creators, not only consumers of AI technology,” he said, highlighting talent, infrastructure, and international collaboration as priorities.
Development pathways
Dr. Varol describes two broad pathways for technological development. One is indigenous development, which requires substantial investment, time and a long-term commitment. The other is what he calls “technological piggybacking,” adopting technologies already developed elsewhere.
For Kazakhstan, Varol said that generative AI requires elements of both approaches.
“We have been also debating and discussing a lot about what is the best strategy for Kazakhstan, especially in the development of generative AI technologies. Our internal deliberations always pointed out that we must develop these technologies indigenously as well,” he said.
Varol also stressed that developing domestic capabilities is not only about owning a model, but also about creating a pool of specialists with expertise who can later move into startups, established companies and new commercial products.
Recent events also demonstrated that access to foreign technologies can become subject to government restrictions. In June, the U.S. government temporarily imposed export controls barring foreign nationals from accessing Anthropic’s Fable 5 and Mythos 5 models, including foreign-national employees of Anthropic itself.
“Creators of these technologies can stop your access to these technologies and this creates a very big dependence risk,” said Dr. Varol.
Technological sovereignty, however, should not be confused with technological self-sufficiency. Mohamed Eissa, chief investment officer at the International Finance Corporation, emphasized that this explains why emerging markets are unlikely to compete directly at that layer of the industry.
“You are not seeing LLMs [large language models] being developed in emerging markets because you need compute, connectivity, power,” Eissa said. “Those things are big infrastructure gaps in emerging markets.”
But he draws an important distinction between building AI infrastructure and using AI. He pointed to a rapidly growing AI company in content creation that originated from Kazakhstan as an example.
“But using AI in emerging markets, you don’t need to build that to be able to use it. That is an opportunity in emerging markets: finding innovative people who have really good ICT skills, have focused on a business problem and now have the tools to compete with Silicon Valley,” he said during The Development Podcast on the World Bank Group YouTube channel.
That may be one of the most important opportunities for Kazakhstan. According to him, in emerging markets, the much larger opportunity may be the application of AI to businesses that already exist. Likewise, in Kazakhstan, potential economic impact of AI should not be measured only by how many programmers, AI startups or research laboratories the nation generates, but rather by whether this technology raises productivity gains across the wider economy.
Existing strengths
Dr. Varol said some of the applications of AI lie in sectors where the country already has an expertise. Mining, energy and logistics are among them.
“Kazakhstan is one of the richest countries with respect to mining and natural resources, but most of these natural resources are exported with a small amount of added value to other countries,” he said.
Using AI and digital technologies in those industries, he noted, could allow Kazakhstan to generate more value from resources before they leave the country.
Where dependence matters
There are areas, however, where Kazakhstan has stronger reasons to maintain domestic capabilities. Dr. Varol said questions about where information is processed and stored become significantly more consequential when AI systems are used by companies or public institutions.
“In general, I believe this is also very important with respect to the sovereignty aspects of a country,” he added.
Language presents a different form of dependence. Global AI systems are trained on enormous quantities of data, but languages are represented very unevenly in those datasets.
Dr. Varol said approximately half of publicly accessible internet content is in English, compared with only around 0.03% in Kazakh. That means developing Kazakh-language AI starts long before a model is trained.
ISSAI, which was founded in 2019, has developed the country’s first LLM in 2024. Drawing on the institute’s experience, Dr. Varol explained that researchers first need to collect texts and other data, remove poor-quality material, organize the information and convert it into formats suitable for machine learning.
Researchers faced another problem once a model has been trained on how to determine whether it actually understands the language.
“Because the internationally accepted method for doing this is called model benchmarking and all model benchmarks are in English language or Chinese language, but none were in Kazakh language. Basically, we also have created a suite of benchmarking data sets,” he said.
The development of Kazakhstan’s LLM illustrates another benefit of indigenous AI development. The process itself creates capabilities. The team had to recruit and train specialists while learning to operate multi-node AI computing infrastructure, expertise that had not previously been used at that scale. The significance of the project therefore extends beyond the model it produced. It helped build a domestic pool of engineers with experience in the infrastructure and workflows required for advanced AI development.
Kazakhstan’s AI strategy therefore does not require choosing between complete technological independence and reliance on global platforms. The more realistic task is to determine which capabilities the country needs to control, which technologies it can efficiently import and where AI can create the greatest economic value.
