Decentralise for Resilience: Three Energy Security Lessons for All Nations

The age of concentrated risk

Throughout the 20th century, the power sector followed one rule: bigger is better. Large generating plants, long transmission lines and a national grid dispatched from a single control centre delivered economies of scale. They also carried a cost that was rarely discussed: concentrated risk.

In recent years that cost has become impossible to ignore. Armed conflict has turned power plants into military targets. Geopolitical shifts can cut off a fuel supply overnight. Extreme weather drains hydropower reservoirs, overloads grids and brings down lines. And a new wave of demand, from AI data centres to high-tech factories, is packing huge loads into a few points on the map.

Three short documentaries from the International Energy Agency (IEA), on Ukraine, Moldova and AI data centres, tell three very different stories. Yet they point to one conclusion that applies to every country: energy security today depends on how decentralised the power system is.

Three stories, three kinds of vulnerability

Ukraine – when big plants become big targets. Before 2022, Ukraine had about 38 GW of dispatchable capacity. It lost 19 GW in the first year of the war; after the spring 2024 attacks only about 12 GW remained. Between March and September 2025 alone, attacks caused more than 3,100 disruptions in the power sector. The IEA recommends that Ukraine increase and decentralise supply, because scattered wind, small gas plants, solar and batteries are harder and costlier to destroy than a central plant.

Moldova – when a whole country hangs on one plant. For years, 70–90% of Moldova’s electricity came from a single Russian gas-fired plant in the breakaway region of Transnistria. When the gas stopped on 31 December 2024, Moldova had to import up to about 75% of its electricity from Romania, household tariffs rose 65–75% within days, and Transnistria faced rolling blackouts of up to eight hours a day.

AI data centres – when concentrated demand outruns the grid. The IEA estimates that data centres used about 415 TWh in 2024 (1.5% of global electricity) and could exceed 945 TWh by 2030. A typical AI-focused data centre uses as much electricity as 100,000 households, and they grow in clusters: nearly half of US data centre capacity sits in five regional clusters. Meanwhile, a new transmission line takes four to eight years, and waiting times for transformers and cables have doubled in three years.

Story Point of concentration Shock Decentralised response
Ukraine Large plants, backbone substations Deliberate attacks Scattered small sources, batteries, municipal cogeneration
Moldova One plant, one gas source Fuel cut off overnight Diverse sources, domestic renewables, grid interconnection
AI data centres Giant load clusters Grid and supply chains cannot keep up On-site generation, storage, flexible demand

Why centralised systems are vulnerable

A centralised power system can have a single point of failure at four different layers. Every country carries at least one of them; they differ only in degree.

  1. Concentrated generation. A few plants account for a large share of total capacity. A sudden trip of a 1,000 MW unit can trigger widespread load shedding; an attack or technical failure there can paralyse a whole region.
  2. Concentrated fuel. One fuel, one supplier or one shipping route dominates supply. The risk may be geopolitical (as in Moldova), price-driven (imported LNG or coal) or climatic (drought for hydropower).
  3. Concentrated transmission. Generation is far from demand, linked by a few long backbone lines. A failure at one key substation or corridor can split the system.
  4. Concentrated demand. New large load clusters (data centres, industrial parks, megacities) need capacity and grid infrastructure faster than conventional investment can deliver.

What the four layers share: risk does not disappear as a system grows; it accumulates. The fewer elements a system relies on, the greater the damage when one fails.

What decentralised systems offer, and their limits

A decentralised system does not eliminate risk; it divides it. Instead of collapsing all at once, it fails gracefully: it degrades in parts and recovers in parts.

Key benefits:

  • Thousands of small sources (rooftop solar, cogeneration, biomass, batteries) are hard to knock out at once. Microgrids can island themselves and keep hospitals, water supply and telecoms running when the main grid goes down.
  • Fuel independence. Sun, wind, biomass and waste are domestic resources that do not pass through another country’s straits or pipelines.
  • A rooftop system or a battery can be installed in weeks, while a large plant and its lines take years. When demand grows fast, speed is security.
  • Mobilising private capital. Households, businesses and local governments invest alongside the state, easing the burden on public budgets and state utilities.
  • Grid relief. Generating close to consumption cuts losses and defers the need for new lines.

Limits to acknowledge:

  • Distributed renewables depend on the weather and need storage and flexible resources alongside them.
  • Millions of small sources require far more advanced distribution grids and dispatch systems.
  • Some heavy industrial loads still need stable, continuous baseload supply.
  • A strong transmission grid is still needed for regional links and cross-border trade, as Moldova’s reliance on Romania showed.

So the right question is not “centralised or decentralised” but “what mix”: centralised sources provide the base, while resilience must come from a sufficiently dense decentralised layer.

Where each group of countries should start

There is no single formula. The starting point depends on the kind of concentration each country carries.

Country group Typical concentration risk Decentralisation priorities
Developing countries with fast-growing demand Supply–demand mismatch, long backbones, limited capital for large plants Industrial rooftop solar, substation storage, direct power purchase agreements
Fuel-importing countries Dependence on LNG, coal or oil via a few sea routes Domestic renewables, cogeneration, biomass, energy efficiency
Small countries reliant on one plant or neighbour One source dominates output Diverse sources, interconnection in several directions, on-site generation
Island states and remote areas Long lines, costly diesel, natural disasters Renewable microgrids with storage, less diesel dependence
Countries exposed to conflict Plants and backbone substations as military targets Scattered small sources, passive protection of substations, equipment stockpiles
Countries attracting data centres and high-tech industry Giant load clusters, connection bottlenecks Require large loads to bring their own generation, storage and flexibility

Most countries fall into several groups at once, so a practical roadmap usually combines these priorities.

Seven policy principles

  1. Measure resilience, not just capacity. Power development plans should test scenarios that remove each region’s largest plant, largest fuel source or most critical backbone line, and set a target share of local generation for every load centre.
  2. Cap dependence on any single source. No plant, fuel or supplier should exceed a safe share of supply without a backup plan.
  3. Open the door to small producers. Simple connection procedures, transparent prices for surplus power, and direct power purchase mechanisms so households, businesses and local governments can invest.
  4. Treat storage as strategic infrastructure. Batteries at substations, industrial parks and buildings both shave peaks and provide backup when the grid fails.
  5. Protect critical services with microgrids. Hospitals, water supply, telecoms and emergency control centres should be able to run on their own for hours to days.
  6. Large loads must bring their own supply. Data centres and other major consumers should be encouraged or required to invest in their own generation and storage and to operate flexibly when the grid is stressed.
  7. Digitalise the distribution grid. Distributed energy resource management systems (DERMS), virtual power plants (VPPs) and smart metering let millions of small sources operate as one coordinated system.

Conclusion

Ukraine learned the lesson of decentralisation under bombardment. Moldova learned it after losing its gas supply overnight. The leading AI economies are learning it through grid connection queues that stretch for years. Other countries have the chance to learn it before they pay the price.

In the 21st century, energy security is no longer measured by how many large plants a country has, but by how much power the system can still deliver when its largest plants cannot run. Investing in a decentralised, layered and intelligent power system today is buying insurance for tomorrow’s development.

References