Pay a tax bill in Kazakhstan, and you might be done in thirty seconds.

Kazakhstan’s government platform has transformed how citizens interact with the state while fintech super-apps handle utility payments, money transfers, and tax returns in a few taps. 

For anyone from a country still reliant on in-person queues and paper forms, the contrast is striking.

Kazakhstan was included in the world's top 10 for the Online Services Index alongside South Korea, Estonia, and Denmark, and rose four positions to rank 24th globally on the 2024 UN E-Government Development Index, outperforming Germany, China, and Australia. 

Living in the US, I notice this acutely. Many of the digital conveniences that Kazakhstanis now take for granted remain unavailable in countries that consider themselves technological leaders. 

But while Kazakhstan has been successful in the digital era, the new era of AI-enabled government is just beginning. President Kassym-Jomart Tokayev has officially declared 2026 the "Year of Digitalization and Artificial Intelligence," and in 2025, the government established a dedicated Ministry of Artificial Intelligence and Digital Development, elevating AI from a sub-portfolio to a central pillar of national governance. 

On November 17, 2025, President Tokayev signed Law No. 230-VIII "On Artificial Intelligence," which entered into force on January 18, 2026, making Kazakhstan one of the first countries in the region with a comprehensive AI statute. 

Meanwhile, Freedom Holding Corp. has signed an agreement to develop a $2 billion Sovereign AI Hub powered by NVIDIA exascale infrastructure at a site with 100 MW of available power, and 165,000 educators are being given free ChatGPT Edu licenses through a partnership with OpenAI and Freedom Holding. 

Ambition for AI capability is not the same as inclusion, and AI raises the inclusion bar even higher than electronic government did.

This is a remarkable acceleration. But ambition for AI capability is not the same as inclusion, and AI raises the inclusion bar even higher than electronic government did. 

The next chapter of Kazakhstan's digital story will be defined less by whether the frontier moves forward and more by whether it reaches everyone.

What follows is a map of what needs to be in place for that to happen.

Energy: from powering services to powering compute

None of this starts without electricity, and AI changes the arithmetic. 

Kazakhstan's current domestic generating capacity falls short of fast-growing demand, with total capacity at about 26.8 gigawatts at the outset of 2026. The country's energy deficit during the 2025–2026 autumn-winter period was expected to reach up to 330 million kilowatt-hours per month, a volume that would have to be purchased from Russia and Uzbekistan, with the Ministry of Energy projecting that the gap will widen to more than 5 billion kilowatt-hours annually by 2029.

AI infrastructure demands extensive electricity resources.

According to the International Energy Agency, electricity demand from data centers soared by 17% in 2025, and that of AI-focused data centers climbed even faster, well outpacing growth in global electricity demand of 3%; electricity consumption from data centers is set to double by 2030, and power use from those focused on AI is poised to triple. The Sovereign AI Hub alone is planned for a site with 100 MW of available power, and Kazakhstan's "Data Center Valley" in Ekibastuz envisions four AI-focused data centers, each 50 MW, with initial Greenfield capacity of up to 100 MW.

To fill that gap, the government is leaning on whatever it can. The proposed Data Center Valley, developed in cooperation with the Pavlodar regional akimat (local executive administration/mayor's office), is expected to be powered by the Ekibastuz coal basin, one of the largest in the country. And Kazakhstan has selected Russia's Rosatom (Russian state-owned nuclear conglomerate) to construct its first nuclear power plant, with operations expected by the early 2030s. 

As Kazakhstan invests in AI infrastructure, who gets priority when compute and electricity compete for resources?

The choice frames the question sharply: as Kazakhstan invests in AI infrastructure, who gets priority when compute and electricity compete for resources? Major tech companies are expected to invest nearly $400B in cloud infrastructure by 2025, and Kazakhstan's $1.9 billion data center project is meaningless if the country can't generate enough electricity to run it. The government has acknowledged this by making the project timeline contingent on resolving the deficit.

A recent CAPS Unlock report, "Opportunities for Small and Medium-Sized Enterprises in Kazakhstan's Green Economy," found that the country has yet to unlock the potential of its SMEs in driving the green transition; expanding household and business uptake of renewable energy that would ease the structural deficit from the demand side, at the same time, the supply side is being asked to power exascale GPUs.

Devices: from "laptops matter" to "is AI even usable from here?"

According to the International Telecommunication Union, basic smartphones are now within reach of most people globally.

Handsets costing around $50 are widely available. But laptops are a different matter. In 2021, the most recent year for which data is available, only 53% of Kazakhstani households owned one. In Almaty city, 72% of households had a laptop; in Almaty province, 38%. In Kyzylorda province, the figure dropped to 28%, in Shymkent to 31%. More than 70% of computer equipment in Kazakhstan is imported, which means prices track exchange rate fluctuations.

While the lack of devices for government was already challenging, meaningful use of generative AI, such as drafting documents, analyzing data, and building anything beyond a chat, depends on devices with keyboards, the ability to run multiple windows at once, and the ability to render long outputs at usable speeds. 

The productivity benefits that the technology promises are not evenly distributed across form factors. 

In an experimental study of computer programmers, the use of ChatGPT increased the number of lines of code written by 56%, benefiting the least-skilled workers the most. A real-world deployment with call center agents similarly found that using a generative AI-based conversational agent increased overall productivity and reduced productivity gaps between novices and experts.

But those gains accrue to people working on full computing environments, not on the entry-level smartphone screens that most rural Kazakhstanis own.

In short, in cities, roughly one in five people has a laptop. In the countryside, it is one in nine. If the policy ambition is to put 165,000 ChatGPT Edu licenses in educators' hands, the question that needs to come first is which device they will use them on, and whether the student sitting in front of them has anything comparable at home.

Internet: from connectivity to AI capability

Around 16% of people in rural Kazakhstan, roughly 1.7 million, have no access to 4G mobile internet. A further 630,000 live entirely beyond mobile coverage. Four providers serve the entire national market, a concentration that tends to produce poor service and high prices. 

Around 16% of the population cannot afford 10GB of mobile data at the standard affordability threshold of 2% of monthly income. Among the poorest households, that figure reaches 40%. 

Connecting 48,000 unserved households to fixed internet would cost roughly $46 million. Extending 4G to rural areas would require $650 million. Reaching the 630,000 people in areas with fewer than one person per square kilometer would require satellite solutions costing over $1 billion.

The government is investing to make this plausible. A national program, Accessible Internet, running from 2024 to 2027, aims to extend coverage to 100% of the population; more than 1 trillion tenge ($2 billion) has been invested in telecoms infrastructure over the past three years. Of Kazakhstan's 6,179 villages, 84% now have mobile internet. By 2027, the government aims for 4G coverage across 92% of settlements and 5G in 20 cities.

But the shift to AI services raises the bar for what "being connected" actually means. The Alliance for Affordable Internet's concept of "meaningful connectivity" — a smartphone or better device, speeds of at least 4G, enough data to go online every day, and a reliable connection at home, work, or school — was already a more demanding standard than the binary of "online or not." AI tightens it further. 

The ITU's Connecting Humanity Action Blueprint warns that technologies such as multi-gigabit fiber, 5G, and artificial intelligence are being deployed extensively in more advanced markets, further widening the gap between the digital "haves" and "have-nots." 

Useful AI requires stable bandwidth (the conversation breaks if the connection drops mid-prompt), affordable data (each query is a round-trip to a cloud server), and low enough latency that the assistant actually feels like one. 

The millions of Kazakhstanis who are nominally online — counted as "connected" in the headline figures — are not yet meaningfully AI-connected.

By that measure, the millions of Kazakhstanis who are nominally online — counted as "connected" in the headline figures — are not yet meaningfully AI-connected. Intermittent or expensive data is not just an inconvenience; it limits the ability to use AI tools at all.

From digital literacy to AI literacy

Access to power, a device, and a connection are still only part of the picture. The harder piece is knowing how to use them. What AI introduces here is not a continuation of digital literacy but a new layer on top of it.

The International Computer and Information Literacy Study (ICILS), the major international benchmark testing eighth-graders, found that Kazakhstani students score below the OECD average. Kazakhstan's average in the 2023 ICILS cycle was 407 points, against a top score of 540 for South Korea.

Almost every second Kazakhstani eighth-grader still struggles with tasks most adults would consider basic, such as clicking a hyperlink, adding an image to a presentation, or understanding what it means to copy someone into an email. 

There is progress to report: the gap between city and rural students narrowed by 33 points between 2018 and 2023, girls outperformed boys, and students with fewer devices at home are slowly catching up. But these are gains from a low base, and Kazakhstan recorded one of the widest gaps between its highest and lowest performers of any country in the 2018 study: 347 points between the top 5% and the bottom 5%.

If basic digital literacy is uneven, AI literacy starts further behind. UNESCO describes the emergence of an "AI divide" — unequal access to, benefits from, and opportunities in AI technology across regions, communities, and socioeconomic groups, with the most marginalized communities bearing the brunt. 

The deepening of generative AI literacy mirrors and exacerbates current digital literacy divides, primarily due to unequal access to necessary technological resources and educational opportunities; GenAI literacy involves understanding how generative models are created and trained, how they generate new content, the limitations and potentials of this technology, and the unique ethical considerations they entail, such as authorship, authenticity, and misinformation. 

Evaluating outputs, writing useful prompts, knowing when an answer is hallucinated, understanding bias in training data — these are not optional add-ons; they are the difference between a tool that empowers and one that misleads. 

A new empirical study of generative AI use in Italy found that GenAI is supplanting other technologies to become a primary information source, including for sensitive tasks like emotional support and medical advice, despite low user digital literacy, posing a risk as users struggle to recognize errors or misinformation.

Kazakhstan's response to this is taking shape. In January 2026, the country joined OpenAI's "Education for Countries" initiative alongside Estonia, Singapore, the UAE, Jordan, and Greece, with 165,000 ChatGPT Edu licenses provided free of charge — 100,000 for preschool, secondary, technical, and vocational educators, 62,800 for faculty members and administrative staff, and 2,200 for the Astana Hub ecosystem. 

Minister of Science and Higher Education Sayasat Nurbek framed it bluntly: "Kazakhstan is not training users of artificial intelligence — we are training its creators." And in the Kazakh language specifically, the country launched its first national supercomputer in 2025 with a capacity of 2 EFlops and 512 NVIDIA H200 GPUs, now powering KazLLM and AlemLLM — sovereign large language models that operate in Kazakh and Russian. 

This matters because mainstream models like ChatGPT, Gemini, and Qwen are built with massive resources aiming for general intelligence, while Kazakh LLMs were created in small teams with relatively small budgets to ensure that Kazakh speakers have AI tools tailored to their language and cultural context. 

AI literacy in a language you do not natively speak is itself a barrier; an AI that speaks Kazakh poorly excludes Kazakh speakers from genuine fluency with the tool.

One structural barrier to basic digital literacy was already the shortage of teaching materials in Kazakh; the Connect-Ed Foundation, an organization I founded, developed free courses in both Kazakh and Russian to begin addressing it. The AI-literacy version of this problem is larger, not smaller.

The frontier and the foundations

Kazakhstan has built something genuinely impressive: government services that work, fintech that functions, a technology sector that draws international attention, and now an AI policy stack — ministry, law, supercomputer, sovereign models, hyperscale data centers, and a national rollout of AI in education — that few mid-sized economies can match. 

The country is positioning AI as a new national development platform, framed in official discourse in terms similar to how e-government was once framed during Kazakhstan's earlier wave of public-sector modernization — but with far higher stakes, in a global environment defined by intense geopolitical competition, technological concentration, and unresolved questions around trust, ethics, and sovereignty.

The country’s digital story has two chapters, and the AI chapter is being written before the first one is finished.

The paradox is that, alongside all of this, a significant share of the population lives in regions without reliable internet access, where laptop ownership is low and digital skills are even lower. The country's digital story has two chapters, and the AI chapter is being written before the first one is finished. 

What was true for government — that being technically connected is not the same as being meaningfully so — is more sharply true for AI. Access alone does not translate into agency.

What happens when countries move from digital government to AI-enabled government is that every existing inequality in the system gets amplified by the new capability layered on top.

Closing the gap means unglamorous work: 

• investing in electricity infrastructure in regions that still import power and now must also feed data centers; 

• introducing real competition into a telecoms market dominated by four providers when the new use case demands stable bandwidth; 

• making devices affordable on real incomes when "device" increasingly has to mean something with a keyboard; 

• building teaching materials in the languages people actually speak, including for AI literacy specifically. 

• treating AI policy not as a tool for innovation rankings, but as a question of who gets left behind in the next phase, not just the last one.

Kazakhstan has demonstrated that it can move quickly when it chooses. The more important question now is not how to reach the frontier of AI, but whether the foundations exist for the frontier to reach everyone.