Amazon wants to continue growing with AWS in Europe, but the plan is running into an obstacle much less glamorous than AI: getting connected to the grid. Analysts and industry insiders have long warned that building a data center is no longer the bottleneck; it’s securing reliable electrical capacity and a connection within reasonable timeframes. The problem has escalated to the point where, in some areas, wait times to connect to the grid can be as long as 7 years, a horizon that is beginning to cause projects to be “paused” or outright rethought.
According to statements from AWS officials and media reports citing industry sources, Europe faces a scenario in which there are regions where no significant new connections are expected until well into the 2030s. Not because of a lack of investors or industrial land, but because the grid — and the permits to expand it — lag behind the pace of digital demand growth.
The symptom: building is fast; connecting, not so much
A large data center can be constructed in 18–36 months if everything goes smoothly. But if the connection point is 5, 6, or 7 years away, the asset becomes an immobilized investment. This lag is forcing hyperscalers—including Amazon—to reschedule timelines, prioritize locations, and renegotiate conditions.
The congestion is especially visible in countries where, beyond cloud growth, AI infrastructure deployment and related electricity consumption are accelerating. AWS has provided striking examples of the volume of connections needed for data centers: tens of gigawatts in countries like Italy and Spain, figures that practically compete with the needs for industrialization and the country’s overall electrification.
The cause: increased demand, aging grids, and slow permits
Europe is trying to electrify transportation, industry, and climate control… while simultaneously multiplying computing capacity. The European Commission has acknowledged that the grid needs accelerated modernization: it speaks of hundreds of billions of euros in investments for electrical networks by 2030, and a system that, in part, still relies on aging infrastructure and permit processes that are too slow for current demand.
Add to this a pattern well known to operators: when capacity is scarce, the “fever” to reserve connection points spikes. The result: long and opaque queues, where solid projects coexist with speculative requests. The European Commission itself is proposing changes to shift from “first-come, first-served” to models like “first-ready, first-served” (prioritizing those truly prepared), along with improvements in transparency regarding available capacity and planning.
Ireland as a mirror: when the system hits “stop”
The Irish case is often cited because it shows the conflict in its rawest form: data center demand grew to account for 22% of national electricity consumption in 2024 (up from 5% in 2015). The regulator (CRU) ended up establishing a policy requiring new data centers to cover at least 80% of their annual demand with renewable energy developed within the country, with a multi-year pathway to achieve this, alongside requirements for generation or storage related to new connections.
This type of decision foreshadows what many fear: that access to electricity for data centers in Europe will evolve from “I ask for power and get it” to more stringent conditions (location, flexibility contribution, generation/storage, market participation, etc.).
What does this mean for companies and system teams?
For an IT team, the headline isn’t just “Amazon delays projects.” It’s a practical consequence: cloud/colocation capacity in certain regions may become strained, not due to lack of racks, but because of a shortage of megawatts. When power is the scarcest resource, cascading effects appear:
- Availability timelines for new rooms and expansions become less predictable.
- Costs include electrical risk—connection, reinforcements, local generation, PPAs, batteries.
- Architectural decisions revisit regional edge, multi-region, and hybrid strategies.
- Continuity planning: if “obtaining power” takes years, disaster recovery and overflow capacity become business issues, not just technical ones.
In Spain, for example, there are already discussions of congestion and saturation at specific nodes, prompting the consideration of alternative locations or the need for much earlier electrical planning.
The uncomfortable message: the bottleneck is no longer just silicon
During 2024–2026, the debate has centered around GPUs and accelerators. But Europe is discovering another reality: without a robust grid (and permits), data centers are useless. And if access to power becomes a multi-year race, the competitive advantage won’t just be having better chips or contracts: it will be being in the right place, with the right connection point, at the right time.
Frequently Asked Questions
Why can a connection queue last up to 7 years?
Because the bottleneck often lies in grid reinforcements, substations, lines, and permits. If the project requires upstream works and regional saturation exists, the actual slot can be delayed by several years.
What can a company do if it needs capacity within 12–18 months?
Plan alternatives: less congested regions, multi-region approach, colocation agreements with pre-energized capacity, and hybrid strategies (on-premises + cloud) while waiting for definitive power.
Does on-site generation (gas, batteries, renewables) solve the problem?
It helps but does not “eliminate” the need for grid access. Batteries provide flexibility and peak shaving; local generation can meet part of the demand and fulfill regulatory requirements, but sustained operation and backing still depend on the system and permits.
What does the “first-ready, first-served” approach proposed by the EU mean?
An attempt to clear queues: prioritizing projects that demonstrate real readiness (permits, financing, engineering) over speculative reservations, to allocate capacity to those capable of executing it.