The sovereign cloud shift: AI scale is shaping operations

Ilkari intelligence note

In 2023, 57% of data centre operators said they would trust an AI model to make operational decisions.

That signal marks a quiet but important shift. Trust is no longer expressed primarily through vendor assurances or contractual safeguards. This is increasingly evident in the willingness to delegate day-to-day control to automated systems. As AI workloads scale, this shift is starting to surface across operations. It shows up in how infrastructure is managed and in decisions about where sensitive workloads can run.

What is emerging is not a debate about technology choice. It is a reassessment of control, accountability and sovereignty at the operational level.

Context: the forces at play

AI workloads are scaling faster than many infrastructure models anticipated. At the same time, power and energy constraints are becoming more visible, and geopolitical accountability pressures are increasing. Together, these pressures are shifting sovereignty away from abstract framing and into the design and operation of data centre infrastructure.

As a result, decisions that were once treated as technical or contractual are now operational. Jurisdiction and control are becoming constraints that shape where and how workloads can run. Long-standing cloud abstractions are starting to break under physical limits, and power availability is emerging as a gating factor rather than an assumption.

Pressure points shaping sovereign cloud decisions

Pressure point 1: Trust is becoming operational

Trust is no longer an abstract condition. It is exercised through systems.

In 2023, 57% of data centre operators said they would trust an AI model to make operational decisions. That shift reflects a change in how confidence is established. Trust is increasingly demonstrated through performance under real operating conditions rather than through vendor assurances, documentation or contractual safeguards.

As automation expands, trust is no longer something organisations claim to have. It is something they test, observe and rely on in day-to-day operations.

Pressure point 2: Sovereignty is becoming an operational constraint

Sovereignty is no longer confined to policy or compliance discussions. It is showing up directly in operational decision-making.

Data residency requirements, jurisdictional accountability and regulatory exposure are increasingly shaping where workloads are allowed to run and how infrastructure is configured. A growing share of organisations are already using sovereign cloud environments, with high reported satisfaction suggesting these choices are being made for practical, not symbolic, reasons.

Sovereignty is no longer something organisations can retrofit later. It is becoming a condition that must be designed into operations from the outset.

Pressure point 3: Control is being reasserted

After years of rapid cloud adoption, many organisations are reassessing how much control they have ceded.

Security concerns, cost unpredictability and jurisdictional risk are prompting a shift away from blanket public cloud reliance. Nearly half of enterprises running workloads in public cloud environments are now considering moving some of those workloads into private or sovereign environments.

This does not signal a rejection of cloud models. It reflects a more selective approach, bringing sensitive or critical workloads closer to environments with clearer lines of control and accountability.

Pressure point 4: Openness and flexibility are becoming risk mitigators

As operational pressure increases, rigidity is becoming a liability.

Organisations are placing greater value on systems that can be inspected, adapted and moved without excessive friction. Open technologies are being prioritised less for philosophical reasons and more for their ability to reduce vendor lock-in, support auditability and preserve long-term optionality.

In this context, flexibility is not about choice for its own sake. It is about maintaining control as conditions change faster than planning cycles.

Pressure point 5: Security is treated as an environmental condition

Security is no longer framed as a feature to be added. It is an operating condition.

As AI workloads expand and systems become more autonomous, security expectations are being pushed deeper into infrastructure and operations. End-to-end protection, encryption and verifiable controls are now treated as baseline requirements rather than enhancements.

When infrastructure underpins critical services, security becomes inseparable from operational continuity and trust.

Signals to watch

Operational trust thresholds are moving faster than governance

In 2023, 57% of data centre operators said they would trust an AI model to make operational decisions. That willingness to delegate control is advancing more quickly than formal governance frameworks, creating pressure to redesign how trust is operationalised.

Sovereignty is becoming formalised, not optional

Fewer than ten percent of multinational organisations had digital sovereignty strategies in place only a few years ago. Analysts now expect that figure to exceed fifty percent by 2029, signalling a shift from exploratory interest to structured operational planning.

Selective workload repatriation is accelerating

Nearly half of enterprises currently running workloads in public cloud environments are considering moving some of those workloads into private or sovereign environments, driven by security, cost predictability and jurisdictional exposure.

Power shaping geography

As AI workloads consume more energy, power availability and energy density are becoming gating factors. Infrastructure decisions are increasingly constrained by the availability of reliable power, not just by network proximity or latency.

Regional powerhouses emerging as reference points

Sovereign infrastructure is beginning to cluster in markets that combine regulatory clarity, energy availability and geopolitical alignment. Countries such as Colombia are increasingly viewed not as peripheral options, but as viable locations for sovereign, AI-ready infrastructure.

Verification replacing rhetoric

Certifications, auditability and inspectable systems are being used as shorthand for trust. High satisfaction rates among organisations already using sovereign cloud environments suggest these choices are being validated in practice.

The shift toward sovereign cloud is not being driven by ideology or branding. It is emerging from operational pressure.

As AI workloads scale, organisations are being forced to reconcile trust, control and accountability with physical, regulatory and geopolitical constraints. Sovereign cloud is becoming relevant not because it is new, but because it reflects how infrastructure decisions are now being made — and where they can realistically be sustained.

• AI