Cross-Border AI/Data Centre Development
Options for the Trebinje–Dubrovnik Region

Summary

The Trebinje–Dubrovnik region, spanning Bosnia  and  Herzegovina (Republic  of  Srpska) and Croatia, has the rare combination of renewable hydropower, strategic geography, and digital connectivity needed to host a world‑class AI/Data Centre. This proposal compares two fully viable development pathways: a cross‑border, EU‑aligned facility in Dubrovnik powered by the shared hydro system, and a sovereign, cost‑efficient Trebinje‑only model powered solely by Trebinje‑1. Both options are technically feasible, financially strong, and capable of generating significant employment and regional impact. The choice depends on strategic priorities: EU integration and long‑term hyperscale growth favour the Dubrovnik model, while speed, sovereignty, and cost efficiency favour the Trebinje‑only path.

Introduction

The rapid acceleration of artificial intelligence, high‑performance computing, and cloud services is reshaping how economies grow, how research is conducted, and how societies function. At the core of this transformation lies a new class of infrastructure: large‑scale AI/Data Centres capable of delivering massive compute power, secure data storage, and resilient digital services.

The Trebinje–Dubrovnik region is uniquely positioned to host such infrastructure. It combines:

  • Abundant renewable energy from a shared hydroelectric system centred on Bileća Lake, Grančarevo Dam (Trebinje 1 HPP), and the Dubrovnik Hydroelectric Power Plant.
  • Strategic geography at the interface between the European Union and the Western Balkans, with direct access to EU markets and regional partners.
  • Existing physical interconnection between Bosnia and Herzegovina and Croatia through hydropower assets that already operate as a functional cross border system.
  • Proximity to Adriatic fibre routes and submarine cables, enabling low latency connectivity to Italy, the Mediterranean, and wider global networks.
  • Untapped regional demand for AI, cloud computing, and digital infrastructure, creating a first mover advantage for whoever builds the foundational platform.

This proposal evaluates two distinct but complementary development pathways:

  • Cross Border AI/Data Centre in Dubrovnik. A flagship EU aligned facility located in Dubrovnik, powered by the combined hydroelectric output of Trebinje 1 and Dubrovnik HPP, and positioned as a regional AI hub for the EU and Western Balkans.
  • Trebinje Only AI/Data Centre. A sovereign, cost efficient facility located in Trebinje, powered solely by Trebinje 1 hydropower, optimised for speed of deployment, lower operating costs, and local economic concentration.

Both options are technically feasible, economically attractive, and strategically meaningful. They differ in governance complexity, funding pathways, regulatory alignment, and long‑term scalability. This document presents a balanced, evidence‑based comparison, including:

  • Detailed energy modelling and hydropower surplus analysis
  • Data centre energy usage breakdown (compute, cooling, infrastructure)
  • Job creation and skills development for each model
  • Financial performance, including revenue, OpEx, CapEx, and margins
  • CapEx vs OpEx structure and long term investment implications
  • Cross border management risks between an EU and a non EU country

The goal is not to declare a single “winner”, but to clarify what each model is best suited for—and under what strategic priorities each becomes the preferred choice.

Strategic Context

The Trebinje–Dubrovnik region stands out for several structural advantages.

Shared hydroelectric infrastructure

Bileća Lake feeds the Grančarevo Dam (Trebinje‑1 HPP), which in turn supplies water to the Dubrovnik Hydroelectric Power Plant via a derivation tunnel. This creates a rare, deeply integrated cross‑border energy system where:

  • Water resources are shared
  • Energy flows are physically interconnected
  • Both countries already depend on the same hydrological and infrastructural base

This existing interdependence is a natural foundation for shared digital infrastructure.

Strategic geographic position

The region sits at the interface between:

  • The European Union (via Croatia), and
  • The Western Balkans (via Bosnia and Herzegovina)

This dual positioning allows an AI/Data Centre to serve:

  • EU based clients requiring EU jurisdiction and regulatory certainty.
  • Western Balkans clients seeking affordable, regionally hosted compute and storage.

Renewable baseload power

Hydropower provides stable, predictable baseload energy—ideal for AI workloads that run continuously and require high reliability. Unlike intermittent renewables, hydro can support:

  • 24/7 GPU clusters
  • High density compute
  • Long duration training runs

Adriatic connectivity

Dubrovnik’s coastal position offers:

  • Access to submarine cables
  • Low latency links to Italy and the Mediterranean
  • Integration into wider European and global networks

This enhances the value of any data centre located in or connected through Dubrovnik.

Data Centre Energy Requirements

To assess feasibility, we first need to understand how much energy a modern AI/Data Centre consumes and what that energy is used for.

Compute load

The core of an AI/Data Centre is its GPU cluster.

  • Modern AI GPUs (e.g. NVIDIA H100) consume 0.7–1.0 kW each under load.
  • A facility with 31,500–47,100 GPUs requires approximately 22–47 MW of continuous compute power.

This energy is used for:

  • Training large language models and other AI systems
  • Running inference workloads for applications and services
  • Supporting research, simulation, and high performance computing tasks

Cooling systems

Cooling is the second‑largest energy consumer.

  • Cooling systems typically account for 25–35% of total data centre energy use.
  • In Dubrovnik, seawater cooling can reduce cooling energy by 30–40% compared to conventional air based systems.

Cooling energy is used for:

  • Seawater pumps and heat exchangers
  • Backup chillers
  • Air handling units and circulation systems

Supporting infrastructure

The remaining 10–15% of energy is used for:

  • Networking and storage
  • Security systems and access control
  • Lighting and building services
  • Power conversion and distribution losses

Total annual energy consumption

Across both models, the data centre is expected to consume 118–176 GWh per year. This figure is central to all subsequent energy and financial modelling.

Hydropower Surplus Analysis

The key question: Is there enough energy to power such an AI/Data Centre in Dubrovnik or in Trebinje?

Hydropower output

  • Trebinje 1 (Grančarevo Dam) Installed capacity: 180 MW Annual output: 370–420 GWh
  • Dubrovnik HPP Installed capacity: 252 MW Annual output: ~766 GWh
  • Combined system (Cross Border) Installed capacity: 432 MW Annual output: 1,136–1,186 GWh

Energy only comparison table

The table below summarises installed capacity, annual hydro output, data centre demand, surplus energy after data centre load, percentage of hydro used by the data centre, and peak load can capacity margins.

Cross Border model (Dubrovnik)

  • Data centre demand: 118–176 GWh
  • Hydro output: 1,136–1,186 GWh

Surplus after powering the data centre: 960–1,068 GWh remain available for other uses.

Implications:

  • The data centre uses only 10–15% of total hydro output
  • There is substantial headroom for: future GPU expansion, additional cloud services and disaster recovery and backup facilities.
  • The system is highly resilient to seasonal variation.

Trebinje only model

  • Data centre demand: 118–176 GWh
  • Hydro output: 370–420 GWh

Surplus after powering the data centre: 194–302 GWh remain available for other uses.

Implications:

  • The data centre uses uses 28–47% of Trebinje 1 output
  • The model is feasible, but with a tighter energy margin
  • Future expansion may require solar integration, battery storage, and additional hydro optimisation or new capacity

Conclusion

Both models can be powered reliably. The cross‑border model has a much larger energy surplus and scalability potential; the Trebinje‑only model is feasible but more constrained.

Option 1: Cross‑Border AI/Data Centre in Dubrovnik

A flagship EU‑aligned digital infrastructure project.

Strategic advantages

1. Access to EU funding. A Dubrovnik‑based facility is eligible for:

  • Digital Europe
  • Horizon Europe
  • Interreg IPA CBC
  • EU Green Deal and related programmes

This can significantly offset CapEx and support research, innovation, and cross‑border cooperation.

2. Large energy surplus and scalability. With 1,136–1,186 GWh of hydro output and only 118–176 GWh required for the data centre:

  • The facility can scale GPU capacity over time
  • Additional services (cloud, backup, research clusters) can be added
  • The site can evolve into a regional hyperscale hub
  • EU Green Deal and related programmes

3. EU regulatory stability. Operating within the EU provides:

  • GDPR compliance and data protection certainty
  • EU level cybersecurity frameworks
  • High investor confidence
  • Direct access to EU digital markets

4. Adriatic connectivity. Dubrovnik’s coastal position offers:

  • Submarine cable access
  • Low latency links to Italy and the Mediterranean
  • Integration into European and global networks

5. Diplomatic and regional integration value. A joint facility:

  • Strengthens EU–Western Balkans cooperation
  • Demonstrates practical cross border integration
  • Supports regional stability and shared prosperity

Financial performance

The table below covers: annual revenue and its components, operating costs and their breakdown, net income and profit margins, energy pricing and provider revenue, CapEx estimates and ROI timelines:

In summary:

  • Annual revenue: €344M–€527M
  • Operating costs: €9M–€20M
  • Net income: €324M–€518M
  • Profit margins: 92–95% (driven by low energy costs and high demand)

Job creation and skills development

Key points:

  • Total jobs: 350–500
  • High skill roles: 120–180 (AI, data science, cloud, cybersecurity)
  • Technical operations: 100–150 (cooling, electrical, network, DC technicians)
  • Administrative and compliance: 60–100
  • Local services: 70–100

CapEx and OpEx structure

Highlights:

  • CapEx: €350M–€500M
  • Higher land, labour, and regulatory setup costs in Dubrovnik
  • OpEx: €9M–€20M per year
  • Higher compliance and cross border administrative costs

Despite higher CapEx and OpEx, strong EU funding eligibility and high revenue potential make the cross‑border model highly attractive for institutional investors and public‑private partnerships.

Cross border management challenges (EU vs non EU)

The cross‑border model also introduces real operational challenges:

  • Dual regulatory compliance (EU + Bosnia and Herzegovina)
  • Slower decision making due to bilateral governance
  • Labour mobility constraints (work permits, social insurance coordination)
  • Political risk linked to EU accession dynamics and bilateral relations
  • Cybersecurity complexity in aligning EU and non EU frameworks

These factors increase governance complexity and must be explicitly addressed in any implementation plan.

Option 2: Trebinje‑Only AI/Data Centre

A streamlined, cost‑efficient, sovereign digital infrastructure project.

Strategic advantages

1. Faster deployment. Single‑jurisdiction permitting and governance can reduce timelines by 6–12 months compared to a cross‑border project.

2. Lower costs. Trebinje offers:

  • 30–50% lower land costs
  • 25–40% lower labour costs
  • Lower regulatory and legal overhead

This translates into lower CapEx, lower OpEx and higher net margins.

3. Full energy sovereignty. Trebinje retains full control over:

  • Reservoir management
  • Hydropower generation
  • Grid connection and pricing

This ensures predictable long‑term energy contracts and keeps energy revenue local.

4. Higher net profit margins with lower costs:

  • Net income: €329M–€520M
  • Profit margins: 93–96%
  • Strong capacity to offer competitive GPU pricing

5. Local economic concentration. All benefits—jobs, taxes, energy revenue, procurement—are concentrated in Trebinje, maximising local development impact.

6. Strategic positioning for the Western Balkans. Trebinje becomes:

  • A regional AI hub
  • A low cost compute provider
  • A sovereign digital asset for Bosnia and Herzegovina

Financial performance

The same financial table applies, with different cost structures:

Key differences:

  • CapEx: €280M–€420M (lower than cross border)
  • OpEx: €7M–€15M per year (lower ongoing costs
  • Slightly higher net income due to reduced expenses

Job creation and skills development

Key points:

  • Total jobs: 30–450
  • High skill roles: 100–150
  • Technical operations: 100–150
  • Administrative and compliance: 40–70
  • Local services: 60–80

While slightly fewer jobs than the cross‑border model, the Trebinje‑only option concentrates all employment locally, amplifying its impact on the city and surrounding region.

CapEx and OpEx structure

Highlights:

  • Lower land, labour, and permitting costs
  • No cross border coordination CapEx or OpEx
  • Stronger cost efficiency and simpler financial management

Integrated Comparison

To support clear decision‑making, a combined comparison table brings together the key differences across all major dimensions — including energy metrics, financial performance, job creation, CapEx and OpEx, governance and risk factors, and overall strategic positioning.

Balanced Strategic Conclusion

Both models are strong, but they serve different strategic priorities.

Cross Border model is strongest when the priority is:

  • Deep EU integration
  • Access to EU funding and programmes
  • International visibility and diplomatic value
  • Long term hyperscale expansion enabled by large hydro surplus
  • Symbolic and practical regional cooperation

Trebinje Only model is strongest when the priority is:

  • Speed of deployment
  • Cost efficiency and higher margins
  • Operational simplicity and reduced bureaucracy
  • Local economic benefit and energy sovereignty
  • Lower political and regulatory risk

Both options are feasible, profitable, and strategically valuable. The choice is not about right vs wrong, but about alignment with the region’s long‑term vision:

  • If the goal is integration and scale, the cross border Dubrovnik model is the natural flagship.
  • If the goal is sovereignty and efficiency, the Trebinje only model is the more focused and streamlined path.

If you’re interested in discussing this concept, please get in touch

Share this page

Licence: All ideas and concepts shown on this website are shared under the Creative Commons Attribution 4.0 International Licence (CC BY 4.0) . You are free to use, adapt, and build upon them, provided you give appropriate credit to Dr. Patrick Reynolds and include a link to this website.
© 2026 Patrick Reynolds