Inside Kazakhstan’s push to become a regional AI centre

Kazakhstan is accelerating its bid to become Central Asia's regional AI hub, with President Kassym-Jomart Tokayev declaring 2026 the Year of Digitalisation and AI and the government approving an AI action plan through 2029. Astana Hub has trained 500 people in its Tomorrow School programme, received 25,000 applications with fewer than 2% accepted, and targets five Kazakh unicorns by 2030 after helping incubate $1bn-plus Higgsfield. The country has also enacted an AI law, deployed local Kazakh-language models and is expanding data-centre capacity, but weak grid reliability, peak-time power imports, unassessed environmental effects and unclear enforcement of biometric-data safeguards present material execution and governance risks.

Analysis

The investable transmission channel is power, not local AI application revenue. Incremental data-centre load in a grid already reliant on marginal imports can raise balancing costs and force accelerated transmission investment; KEGOC (KASE: KEGC) is the clearest domestic beneficiary if regulated tariffs are reset to earn on new capex. The offset is politically constrained pricing: if authorities prioritize cheap electricity for strategic digital projects, utility returns may lag asset growth while industrial users absorb curtailment or higher tariffs.

Data-localisation requirements create a defensible advantage for incumbents with domestic enterprise distribution and trusted payment/customer-data rails, notably Kaspi.kz (NASDAQ: KSPI), Halyk Bank (LSE: HSBK) and VEON (NASDAQ: VEON, through Beeline Kazakhstan). The opportunity is not consumer-facing foundation models; it is higher-margin cloud, cybersecurity, identity verification and government/SME workflow spending over 6-18 months. Any valuation benefit requires disclosed enterprise monetisation, because subsidised training programs and startup-incubation targets do not by themselves establish sustainable revenue pools.

The contrarian risk is that sovereign AI policy becomes a power-allocation and surveillance story rather than a technology-productivity story. A binding capacity shortage, opaque procurement, or adverse data-governance event would increase country-risk premia and weaken the local fintech/telco rerating before it meaningfully affects earnings. Conversely, independently disclosed data-centre power-purchase agreements, grid-capex approvals, and auditable public-sector software contracts would validate a 1-3 month catalyst path.

There is no clean, high-conviction standalone AI trade today: most direct beneficiaries are private or have immaterial Kazakhstan exposure. The more actionable relative-value setup is to own domestic digital incumbents only against evidence of enterprise monetisation, while treating power availability and tariff policy as the gating variables rather than extrapolating global AI multiples into Kazakhstan.

Key Decisions for Investors

• Place a 1-3 month watch alert on KEGOC (KASE: KEGC) for transmission-capex approval, regulated-return/tariff guidance, and disclosed large-load connection agreements. Initiate only if allowed-return mechanics preserve real returns; avoid if strategic-load tariffs are capped without compensating capex recovery.
• Maintain KSPI and HSBK as conditional 6-18 month beneficiaries of domestic digitalisation, but do not add on AI narrative alone. Add only after quarterly disclosure shows incremental B2B, government-services, cloud, cybersecurity, or digital-identity revenue; falsify on rising compliance costs or a material data-privacy regulatory intervention.
• Monitor VEON for enterprise-cloud and cybersecurity contract wins in Kazakhstan; a small tactical long versus an emerging-market telecom basket is warranted only after contract value, margin profile, and data-centre funding are disclosed. Exit if capex intensity rises without enterprise ARPU or EBITDA conversion.
• Avoid extrapolating Kazakhstan AI infrastructure demand into Kazatomprom (LSE: KAP) near term: local grid constraints are more likely to delay new load than create incremental nuclear demand. Revisit only if funded nuclear-generation plans, long-term offtake, or material regional power-market reforms emerge.
• For country-risk management, treat a reported grid shortfall, power-import spike, facial-recognition controversy, or restrictive enforcement action as a trigger to reduce Kazakhstan digital exposure; these events could compress KSPI/HSBK multiples faster than any AI-related revenue can offset.