Smart Cities, Hard Questions: What Central Asia's AI Ambitions Mean for Europe's Public Sector

The Astana Blueprint: Ambition Meets Infrastructure

Astana's smart city ambitions centre on a single institution: Alem.ai. This eight-storey complex in the capital's downtown is the physical manifestation of Kazakhstan's AI strategy. The ground floor serves as a public-facing museum and education centre. Levels two and three house creative tech courses for teenagers. Floors four and five deliver adult AI schooling. The sixth floor hosts a startup incubator, the seventh dedicated R&D laboratories, and the eighth floor houses the AI government operations centre.

What makes Alem.ai noteworthy is not just the hardware; it is the deliberate emphasis on technology transfer. Presight Kazakhstan, the state-backed AI institute, employs over 50 specialists explicitly tasked with moving knowledge from theory into local industry application. Kazakhstan's authorities have mandated that 60 per cent of suppliers for the smart city project come from domestic sources, a protectionist move that reflects both economic nationalism and the genuine absence of sufficient local AI talent at scale. European policymakers debating domestic AI procurement rules under the EU AI Act will recognise the dilemma immediately.

Behind Alem.ai sits Alem.Cloud, Central Asia's first supercomputing cluster. Equipped with NVIDIA H200 processors and ranking 86th on the TOP500 global list, the system is designed to support AI applications across education, healthcare, and agriculture. For context, the European High Performance Computing Joint Undertaking (EuroHPC JU) has been pursuing a comparable strategy across the continent since 2018, distributing supercomputing capacity across sites including the Leonardo system in Bologna and the LUMI system in Kajaani, Finland. Kazakhstan has, in a single facility, concentrated what Europe spent years distributing.

The infrastructure push extends beyond the capital. Kazakhstan has commissioned three new data centres with a combined capacity of 12.9 megawatts, part of a broader push to achieve 100 per cent 5G coverage in twenty major cities. The government has already reached 99 per cent internet coverage nationally, meaning the connectivity layer is largely in place. Even remote border crossings and transport hubs are set to receive Direct-to-Cell satellite internet.

Uzbekistan's Parallel Path: MyID and the Digital State

Uzbekistan is taking a different trajectory, centred on biometric identification rather than centralised data hubs. The MyID system has enrolled 14.5 million citizens in a unified biometric identity framework. This system operates as the backbone of Uzbekistan's digital state: mandatory biometric verification for mobile SIM cards, automated OneID registration, and integration across government services.

Where Astana emphasises computational prowess and public education, Uzbekistan emphasises identification and integration. These are two genuinely different visions of what "smart" means in practice. European readers will note the contrast with the EU's own digital identity framework, eIDAS 2.0, which is attempting to create a voluntary, privacy-preserving wallet system across 27 member states. The Uzbek approach is faster, more centralised, and considerably less concerned with the consent architecture that European regulators have spent years designing.

Věra Jourová, the former European Commission Vice President for Values and Transparency, has repeatedly stressed that digital identity systems must embed fundamental rights protections from the outset, not retrofit them after deployment. The MyID model, with its mandatory enrolment and government-controlled verification, sits at the opposite end of that design philosophy. That is not automatically a failure; it is a deliberate choice. But it is one European procurement officials should understand before drawing comparisons to their own programmes.

The Reality Check: Promise Versus Delivery

Here is where the narrative begins to fracture. Astana and New Tashkent are building genuinely impressive infrastructure. The numbers are real. The processing power exists. The networks are deployed. Yet none of this guarantees that a bus passenger in either city will experience noticeably better public transport, that a farmer in rural Kazakhstan will see improved yields from AI-driven agricultural advisory, or that a doctor in Tashkent will have diagnostic tools that materially improve patient outcomes.

Both nations are attempting to compress decades of digital transformation into a single decade. They are building the hardware and governance layer without yet proving that the software layer, the actual applications citizens interact with, will function at scale. This is not a problem unique to Central Asia. Professor Virginia Dignum of Umea University, one of Europe's leading voices on responsible AI in public services, has consistently argued that technology deployment in government settings fails most often not at the infrastructure stage but at the integration stage, when systems meet the messy reality of public administration and uneven user capability.

The biometric emphasis across both nations also raises a governance question that smart city marketing materials rarely address. When a government controls not only the ID system but also the AI systems that analyse that data, the potential for mission creep is substantial. Central Asia does not have the independent data protection authorities, ombudsman structures, or civil society pressure that provide at least partial accountability in EU member states. The General Data Protection Regulation (GDPR) is imperfect, but it exists for precisely this reason.

What Success Actually Looks Like

For Astana and New Tashkent, "smart city" status cannot rest solely on data centre rankings or biometric enrolment numbers. Genuine success requires demonstrable improvements across a specific set of citizen outcomes:

• Public transport efficiency and passenger experience
• Healthcare diagnostic accuracy and treatment outcomes
• Agricultural productivity and farmer income
• Government service delivery speed and transparency
• Energy efficiency and grid optimisation
• Public safety, without mission creep or overreach

These are the same criteria that European smart city evaluations use, or should use. The City of Amsterdam's responsible AI register, which publicly documents every algorithmic system deployed in municipal operations, represents one model for how infrastructure investment can be paired with accountability. Tallinn's long-running e-governance programme represents another: citizen-facing services built on interoperable digital identity, developed incrementally over two decades rather than unveiled in a single grand infrastructure moment.

The two Central Asian capitals are worth watching precisely because they are not attempting incremental digital change. They are trying to leapfrog, building 21st-century governance infrastructure in nations where baseline digital maturity remains uneven. That same boldness explains both why their ambitions are impressive and why the risks of failure are proportional. Infrastructure without application is expensive storage. Application without accountability is a different kind of risk entirely.

European public sector AI teams have spent considerable energy debating procurement frameworks, algorithmic impact assessments, and transparency requirements under the EU AI Act. Central Asia's approach is a useful stress test of the opposing hypothesis: that speed and centralisation, unconstrained by those frameworks, can produce better outcomes faster. The results, in two to five years, will be instructive for both sides.