The rapid development of artificial intelligence raises a fundamental question for electricity systems: who should finance the grid investments required to connect data centers, and who should bear the costs if planned projects never materialize?
According to the International Energy Agency (IEA), global investment in data centers exceeded USD 550 billion in 2025, while their electricity consumption increased by 17%. Demand is expected to double over the next five years.
These findings are presented in the IEA report Powering Artificial Intelligence – Strategies for Integrating Data Centers into Electricity Systems, which examines regulatory approaches across more than 34 jurisdictions representing approximately 90% of global data center capacity.
Yet behind the rapid expansion of the digital economy lies a major infrastructure challenge. Data centers require substantial and reliable electricity supply, and connecting them often involves building new substations, transmission lines and reinforcing existing networks.
Who should pay for electricity infrastructure?
The IEA examines two principal approaches to allocating grid connection costs.
Under the shallow connection charging model, developers pay for their direct connection to the grid, while broader network reinforcement costs are recovered through regulated network tariffs. Under deep connection charging, developers also contribute to additional network investments attributable to their projects.
This distinction is critical. When connecting a large data center requires substantial grid upgrades, recovering those costs through general network tariffs may effectively shift part of the financial burden onto households and other businesses.
At the same time, new infrastructure may deliver benefits to the wider electricity system. Regulators must therefore distinguish between investments required exclusively for a particular project and those that provide broader benefits to other network users.
The central issue is not whether electricity networks should expand to accommodate artificial intelligence, but how the associated costs and risks should be allocated.
Data center developers should finance their connections and the additional grid investments required exclusively for their projects, while providing appropriate financial guarantees for their commitments. Network operators, meanwhile, should continue investing in infrastructure that benefits the wider electricity system, subject to regulatory oversight and cost justification.
Such an approach would support the growth of the digital economy without transferring private investment risks to households and other businesses.
Greece introduces specific financial obligations for data centre developers
Greece provides one of the latest examples of regulatory action addressing this challenge.
Under legislative provisions approved in November 2026, data center developers are required to cover their grid connection costs, including additional investments in network expansion and reinforcement necessitated exclusively by their projects.
This establishes an important principle: infrastructure costs directly attributable to a private investment should not automatically be passed on to other electricity consumers.
The Greek framework also introduces a bank guarantee of EUR 200,000 per megawatt of maximum electricity demand. For a 100 MW data center, this would amount to a EUR 20 million guarantee.
The measure aims to discourage speculative connection applications, prevent the reservation of grid capacity by projects lacking sufficient investment readiness, and reduce the risk of unnecessary infrastructure expenditure.
The Greek approach is particularly significant because it links access to grid capacity with the developer’s financial responsibility.
The financial risks of projects that never materialise
The challenge extends beyond financing network expansion. What happens if a data centre is never built or ultimately uses significantly less capacity than originally contracted?
In some regions, connection requests from prospective data centers already exceed the electricity system’s existing peak demand. However, a connection request does not guarantee that the project will proceed.
If network operators invest in infrastructure based on uncertain development pipelines, they risk creating expensive, underutilized assets whose costs may ultimately be recovered from other consumers.
Virginia, in the United States, has adopted measures addressing this risk. From 2027, the tariff framework for large electricity consumers provides for long-term agreements, minimum payments linked to contracted capacity, and financial guarantees.
Although Greece and Virginia apply different mechanisms, they pursue a common objective: preventing the financial risks associated with large private developments from being unfairly transferred to other network users.
Can growing electricity demand also stimulate new energy investment?
Data centers do not necessarily have to represent an additional burden on electricity systems. They can also stimulate investment in new power generation, battery storage and network infrastructure.
According to the IEA, data center operators signed long-term power purchase agreements (PPAs) covering approximately 45 TWh of electricity in the first half of 2026, nearly 20% more than during the same period in the previous year.
Another emerging approach is Bring Your Own Power, under which developers secure additional generation capacity to support their electricity requirements. In Ireland, certain new data centers are required to secure dispatchable generation capacity capable of responding to system needs.
Flexible grid connections offer another potential solution. Under such arrangements, electricity consumption may be temporarily curtailed during periods of network congestion.
The IEA estimates that these approaches could potentially unlock an additional 750–900 GW of grid capacity to accommodate electricity demand globally.
However, flexibility comes at a cost. Not all computing workloads can be interrupted or postponed, and the economic viability of flexible operation depends on the nature of the services provided and the conditions within the electricity system.
Could data centres help address surplus solar generation in electricity networks?
The rapid expansion of solar photovoltaic capacity raises another important question: could new data centers be located in areas where solar generation exceeds local electricity demand?
This issue is particularly relevant to North Macedonia. As solar capacity expands, periods increasingly occur when electricity flows from the distribution network back into the transmission system because local generation exceeds local consumption.
This raises the question of whether some new electricity demand could be developed closer to generation facilities, rather than transmitting the entire surplus to more distant consumption centers.
Data centers could contribute to greater local utilization of solar electricity, particularly where certain computing workloads can be shifted towards periods of high renewable generation.
A European scientific study published in 2026 suggests that flexible data center operations could reduce renewable energy curtailment in a modelled electricity system.
However, data centers also require electricity during evening and nighttime hours, when solar facilities are not generating. This creates additional challenges involving generation adequacy, energy storage, network infrastructure and overall system costs.
Could data centers, through appropriate siting and operational flexibility, become part of the solution for integrating renewable energy rather than simply adding another burden to electricity systems?
There is no universal answer. Nevertheless, these questions highlight why the development of artificial intelligence must be considered alongside the planning of electricity generation, consumption and network infrastructure.
Viktor Andonov, PhD

