Kazakhstan Bets on Becoming a Regional AI Centre

Estimated reading time: 6 minutes · Last updated: 21 September 2026

Kazakhstan is positioning itself as a regional AI centre by combining a national strategy, new laws and rapid skills training. As first reported by Al Jazeera, the government declared 2026 the “Year of Digitalisation and Artificial Intelligence”, created a Ministry of Artificial Intelligence and Digital Development in 2025, and approved an action plan that runs through 2029. Public and private hubs are scaling course pipelines: Astana Hub reports 25,000 applications to its programmes and strict selection, and the Tomorrow School has trained 500 people eighteen months into the programme. Kazakhstan’s approach mixes national models, tax incentives for startups and state-backed AI pilots such as SKAI on the Samruk-Kazyna board.

How Astana Hub and the Tomorrow School scale talent

Astana Hub is central to Kazakhstan’s skills push. The cluster runs free programming and generative-AI classes for schoolchildren and adults, and it operates the Tomorrow School for intensive retraining. Valeria Tio, Astana Hub’s managing director, told Al Jazeera the hub has received 25,000 applications and that selection is strict: organisers take less than 2 percent of applicants.

Selection begins with logic and memory testing, followed by a four-week course called the “Pool” that selects a cohort for two years of training. The programme serves students aged “18 to over 50”, and eighteen months into the Tomorrow School it had trained 500 people. This month pools opened in 12 Kazakh regions, extending the pipeline outside the capital.

Astana Hub also runs an incubator that offers zero tax for residents and takes 1 percent of income to support the hub. Over eight years the cluster helped incubate Higgsfield; Astana Hub identifies Higgsfield as Kazakhstan’s first unicorn, a privately held company with a valuation exceeding $1 billion. Valeria Tio said the hub hopes to produce five unicorns by 2030.

Homegrown models and infrastructure: MagniSoz, Oylan and a supercomputer

University research groups are building local models to reflect Kazakh language and culture. The Institute of Smart Systems and Artificial Intelligence at Nazarbayev University developed MagniSoz, a model launched in December 2024 that works with text and audio in 11 languages, and an updated model, Oylan, that also handles images in English, Russian and Kazakh. Yerbol Absalyamov, the institute’s executive director, said the models were deployed on local servers and released open source to Kazakhstani users.

The researchers began the work after 2019 when mainstream assistants failed to recognise Kazakh; during the coronavirus year of 2020 they gathered voice recordings from local speakers to train the models. The institute framed its work around two principles: “AI for good” and “AI for Kazakhstan”, and it emphasises data residency by running the models on local infrastructure.

The institute says it was the first academic entity in Central Asia to operate a supercomputer, which supports training and inference workloads. That combination of local models, compute and open-source release is positioned as a way to preserve culture and keep data within the country while building practical tools for government and businesses.

State strategy, corporate pilots and tax incentives

The government has pitched AI as part of a modernisation and diversification agenda. President Kassym-Jomart Tokayev declared 2026 the “Year of Digitalisation and Artificial Intelligence” to accelerate adoption across public services, healthcare, taxation and energy. In 2025 Kazakhstan became the second jurisdiction after the European Union to pass a law governing AI, and the government approved an action plan that runs through 2029.

State-owned and private actors are testing governance and transparency models. Samruk-Kazyna’s experiment with SKAI (Samruk-Kazyna Artificial Intelligence) put an AI programme with voting rights on its board in late 2025; the fund says the move is intended to support transparency and governance. Companies in Astana Hub pay zero tax while resident startups contribute 1 percent of income to the hub’s development, a fiscal incentive intended to attract and retain teams.

Constraints: power, transparency and rights

Infrastructure and oversight are the main limits to the strategy. CAPS Unlock programme coordinator Tlegen Kuandykov pointed out that Kazakhstan still imports power at peak times and that much of its grid and plants are old. The plan to build large data centres has prompted questions because data centres need stable electricity and cooling; Kuandykov noted the ministry said cooling would use immersion systems rather than water but that no public assessment had been released.

Human-rights risks are the other constraint. Kuandykov said the new AI law contains protective language but that enforcement and public oversight were unclear. He warned biometric identification and facial-recognition technologies are being rolled out rapidly, a deployment track that raises privacy and rights questions without clear independent oversight.

How Kazakhstan’s AI push could play out

The bottom line

Kazakhstan’s AI push pairs top-down policy with bottom-up skills and university research. The combination of a 2025 AI law, a ministry, tax incentives for hub residents and local models such as MagniSoz creates a credible pathway to regional relevance. That pathway remains conditional: it depends on reliable power and transparent infrastructure planning, enforceable safeguards around biometric and identification systems, and whether the training pipeline and incubator can convert talent into exportable companies. The government’s action plan through 2029 and Astana Hub’s target of five unicorns by 2030 will be concrete milestones that determine whether the country becomes a lasting regional AI centre.

What to watch

• monitor delivery against the government’s action plan that runs through 2029 and any published progress reports in 2027–2029; 2029 is the plan’s end year.
• watch whether Astana Hub reaches its stated goal of five unicorns by 2030, as Valeria Tio has set the target for that year.