Beyond Cheap Electricity: World Bank Reveals What Really Attracts AI Data Centre Investment

The World Bank study finds that cheap electricity alone is not enough to make countries competitive in AI compute exports, as institutional quality, regulatory stability, financing and geopolitical trust play a far greater role in attracting investment. It recommends that governments, development partners and private investors prioritise governance reforms, affordable financing, reliable digital infrastructure and strong data governance alongside energy investments to build sustainable AI economies.

Artificial intelligence is creating a new global race for computing power, but a new World Bank Policy Research Working Paper suggests that cheap electricity alone will not determine the winners. While many developing countries hope to leverage abundant energy resources to attract AI data centres, the research finds that the real drivers of competitiveness are institutional quality, regulatory certainty, affordable financing and international trust. Based on a trade model covering 85 countries, the study argues that governments seeking to build AI industries must look far beyond energy policy if they want to become global hubs for AI compute services.

Cheap Electricity Is Only Part of the Equation

The rapid expansion of artificial intelligence has triggered an unprecedented demand for computing infrastructure. The report notes that the computational power used to train AI models has been doubling roughly every six months since 2010. Data centres accounted for around 1.5% of global electricity demand in 2024, with that share expected to exceed 3% by 2030. Meanwhile, global data-centre foreign direct investment crossed US$310 billion in 2025, while annual cloud-computing exports surpassed US$9 billion, making AI infrastructure one of the fastest-growing digital investment sectors.

Despite these opportunities, the study finds that low electricity prices alone rarely translate into export competitiveness. One of the main reasons is that hardware accounts for nearly 90% of total AI compute costs, while electricity, cooling and construction together contribute only a small share. As a result, production costs differ by only 12–20% across countries, leaving little room for cheap power alone to create a lasting competitive advantage.

This finding has important policy implications. Governments investing heavily in power generation to attract AI infrastructure also need to strengthen digital governance, improve regulatory stability, modernise electricity networks and create investment-friendly business environments.

Strong Institutions Attract More Investment Than Low Costs

The research demonstrates that institutional quality often outweighs production costs in determining where global technology companies invest. After adjusting electricity prices to remove government subsidies, countries such as Kyrgyzstan, Ethiopia, Kosovo, Canada and Tajikistan emerge as the world's lowest-cost AI compute producers. However, many of these countries have attracted very little data-centre investment.

For example, Kyrgyzstan ranks as the cheapest producer under cost-recovery pricing but has only 5 MW of installed data-centre capacity. By contrast, countries such as Malaysia, India and Brazil, which have only moderate electricity costs, have attracted billions of dollars in AI infrastructure investment because they offer stronger regulatory systems, more reliable infrastructure and greater investor confidence.

The study argues that investors prioritise stable regulations, predictable electricity supplies, transparent permitting processes, enforceable contracts and trusted legal systems over marginal savings in electricity costs. This suggests that governments hoping to attract hyperscale cloud providers should view AI infrastructure as an institutional development challenge rather than simply an energy investment opportunity.

Financing and Trust Could Decide the AI Race

Another major finding concerns access to finance. Because expensive graphics processing units dominate project costs, borrowing costs have a much larger impact on competitiveness than electricity prices. According to the report, a 10-percentage-point increase in financing costs raises production costs by around US$0.29 per GPU-hour, roughly four times greater than the electricity-cost differences separating many leading AI compute locations.

The paper also introduces the idea of a "sovereignty premium." Companies processing sensitive data often prefer countries with trusted legal systems, strong data protection laws and stable geopolitical relationships. Even if electricity is inexpensive, concerns about cybersecurity, contract enforcement or political risk can discourage businesses from locating AI infrastructure in certain markets.

For policymakers, this means improving regulatory credibility may generate greater investment than offering subsidised electricity. Measures such as internationally recognised data governance standards, transparent regulations, sovereign guarantees, blended finance and partnerships with multilateral development institutions can significantly improve investor confidence.

For development partners including the World Bank, regional development banks and donor agencies, the findings suggest future support should extend beyond financing power infrastructure. Greater emphasis on regulatory reforms, digital governance, institutional capacity building and financial risk reduction could help developing countries become more competitive participants in the global AI economy.

Building AI Economies Requires Long-Term Policy Reform

The report concludes that AI compute exports should be viewed as a sophisticated digital industry rather than simply another energy-intensive business. Countries with abundant renewable energy, hydropower, or natural gas certainly possess an important advantage, but success ultimately depends on combining those resources with strong institutions and a stable investment environment.

For governments, the recommendations are clear. AI strategies should integrate energy planning with digital economy policies, strengthen legal and regulatory frameworks, improve access to affordable capital and develop internationally trusted data governance systems. Rather than relying primarily on electricity subsidies, countries should focus on creating predictable business environments capable of attracting long-term private investment.

Private-sector stakeholders also have clear lessons. Technology companies, infrastructure developers, renewable energy firms and cloud providers should evaluate future investment destinations not only on operating costs but also on financing conditions, governance quality, regulatory stability and geopolitical alignment. Countries that combine these strengths are likely to become the next generation of global AI compute hubs.

Ultimately, the World Bank study concludes that the future AI economy will be shaped less by who has the cheapest electricity and more by who offers the most trusted institutions. For developing economies, this means the path to becoming an AI powerhouse lies not only in building more power plants but also in strengthening governance, improving financial systems, and enhancing international credibility.