Cloud, on-prem or sovereign hosting for telco data: three options with different trade-offs

Three options in today’s Uzbek reality

When “cloud vs on-prem” is discussed for a telecom, in reality this is a three-way decision.

Public cloud (AWS, Azure, GCP). International providers with high capability and proven reliability. In Uzbekistan — limited presence (no local regions for major providers), higher latency, data residency questions.

On-prem. Owned infrastructure. Full control. Capital-intensive. Operational responsibility on the operator.

Sovereign cloud. Local cloud providers, privately operated under regulatory licences. Data residency in country. Capability usually below major public clouds, but sufficient for most use cases.

Each option has its zone of applicability, and the universal answer “everything to cloud” or “cloud does not work for regulatory reasons” is a simplification that does not match reality.

Comparison on the things that matter

Criterion	Public cloud	On-prem	Sovereign cloud
CapEx	low	high	low-medium
OpEx	high-predictable	medium	medium
Capability frontier	highest	limited by team	medium
Data residency	hard for strict local	full control	local
Regulatory simplicity	hard	simple	simple
Local skill availability	hard	easier	easier
Time to provision	minutes	months	hours-days
Vendor lock-in	high	low	medium

What drives the choice

Class of data. Customer PII, biometric data, financial transactions usually require data residency. Sovereign cloud or on-prem fit. Marketing analytics, content delivery — can live in public cloud.

Latency requirements. Real-time customer-facing needs low latency. Cloud regions in Uzbekistan are limited. If public cloud, edge presence or sovereign cloud is needed.

Skill availability. Public cloud needs expertise that is relatively scarce in the local market. On-prem skills are larger but declining. Sovereign cloud — middle ground.

Total cost of ownership. Depends on sizing, traffic patterns. Public cloud is cheaper for variable workloads, more expensive for stable high-volume.

Regulatory landscape. The Uzbek regulatory framework for cross-border data transfer (the personal data law ZRU-547) imposes restrictions on certain data classes.

A hybrid model is often optimal

A pure one-or-other rarely works. Most operators choose hybrid:

Customer-facing real-time core — sovereign cloud or on-prem (low latency, regulatory compliance).

Analytics, ML training, batch processing — public cloud (capability, scale).

Backup and DR — split (sovereign plus public).

Development and testing — public cloud (speed, flexibility).

This model gets the best of each option but operations complexity is higher.

What often goes wrong

Public cloud bill shock. Provisioning without discipline leads to unexpected costs. Twelve months in the discovery “we spend X on cloud” becomes an organisational issue.

Overestimated on-prem capacity. Buy 5-year capacity upfront, two years later demand changes, infrastructure underutilised. CapEx sunk.

Sovereign cloud capability gaps. The local provider does not have a specific service needed. Workaround (custom build) overhead higher than expected.

Hybrid without clear ownership. Without a clear architecture standard, teams use different clouds inconsistently. Fragmentation.

Vendor lock-in via specific cloud services. Building on proprietary cloud services makes future migration prohibitively expensive.

When which model fits

Public cloud first. The operator is small or mid-size without strict data residency requirements, focused on growth, with a cloud-skilled team.

On-prem first. The operator is large, established, has existing data centre investment, the regulatory environment is strict, capex is available.

Sovereign cloud first. Data residency is strictly required, local capability is adequate, capex is constrained, operations team is mid-skilled.

Hybrid. Business workloads are structurally diverse — there is real-time customer-facing AND large-scale analytics AND legacy systems hard to migrate.

Discussion points for the committee

What is the regulatory environment for cross-border data transfer relevant to the operator? This often defines the main constraints.

What is the current infrastructure inventory and age? This affects available capex for on-prem renewal.

What is the cloud skill availability in the team? And the recruitment landscape?

What is the 5-year TCO comparison for each option? Not “cloud is cheaper” — concrete numbers.

Is the organisation ready for hybrid complexity if hybrid is chosen?

How SamaraliSoft can help

Hosting Strategy Architecture — classification of workloads by requirements (data residency, latency, scale, change frequency), TCO modelling across the three options, regulatory framework analysis, hybrid architecture design where applicable, and a phased migration roadmap over 18-24 months.

Related reading

• /en/architecture/telecom-around-core-architecture/ — layer around the core
• /en/insights/telecom-data-contracts/ — data contracts
• /en/insights/telecom-event-driven/ — event-driven foundation
• /en/insights/telecom-master-data/ — master data

Sources

• Ministry of Digital Technologies — telecom infrastructure UZ
• Digital Government Portal — digital infrastructure UZ
• UzDaily — telecom market
• Central Bank of Uzbekistan — biometric verification for digital banking
• MNP.UZ — number portability