Summary
- BCG says connection queues in some markets now extend five to ten years as large data centre loads compete for grid capacity.
- Its model combines non-firm connections with flexible workloads, behind-the-meter generation, and storage.
- The concept shifts data centre power strategy from securing maximum firm capacity towards operating around grid constraints.
Boston Consulting Group has proposed a grid-positive model for Northwest European data centres built around non-firm connections, flexible computing loads, behind-the-meter generation, and energy storage.
The consultancy argues that connection queues in many markets now run for five to ten years, making energisation speed a decisive factor in where new AI and cloud capacity can be developed.
Its central proposal is that data centres should not always wait for an unrestricted firm grid connection at their preferred location. Instead, developers could accept a connection that allows the network operator to curtail demand at periods of grid stress, while using local generation, storage, and workload management to keep computing services available.
The idea challenges one of the basic assumptions behind conventional data centre electrical design: that the site should be able to draw its contracted load from the public grid whenever required. That approach supports a simple resilience model but becomes difficult where hundreds of megawatts of new demand are seeking connections to already constrained transmission and distribution systems.
BCG says the alternative could allow capacity to energise years earlier in some Northwest European markets. The trade-off is that the data centre has to become much more active in managing when and how it consumes electricity.
Flexible computing provides one part of that mechanism. Some AI workloads, particularly training and batch processing, can be scheduled around power availability more readily than low-latency customer services. Work can potentially be moved in time or between locations, reducing demand at a constrained site when required.
Behind-the-meter infrastructure provides the second part. Batteries can cover short-duration constraints and provide fast response, while onsite generation can support longer periods when grid import is restricted. The correct mix depends on curtailment frequency, duration, fuel or energy costs, emissions requirements, and how much of the compute load can actually move.
BCG’s analysis suggests significant behind-the-meter capacity would be necessary in congested Northwest European markets. That raises capital costs and introduces additional planning, permitting, fuel, emissions, maintenance, and operational requirements. The consultancy argues that those costs may still be outweighed by the commercial value of bringing a data centre online several years earlier.
The concept has obvious limits. Many data centre loads cannot simply disappear during a grid event. Colocation customers buy contracted capacity with availability expectations, cloud services include latency-sensitive workloads, and critical systems cannot be treated as interruptible industrial demand without a replacement power source.
Large onsite generation also moves the constraint rather than eliminating it. Gas engines or turbines require fuel infrastructure and emissions permitting. Batteries need substantial capital and physical space, and they provide stored energy rather than continuous generation. Renewable generation is variable and may require its own grid connection or storage.
The regulatory model would also have to change. BCG notes that non-firm connection is often an option rather than a requirement, leaving developers able to remain in the queue for firm capacity. Network operators and governments would need commercial incentives or connection rules that reward sites capable of responding to system conditions.
Even so, the proposal reflects the direction of travel in European data centre power strategy. Grid access is becoming less binary. Developers are increasingly considering staged connections, private networks, batteries, onsite generation, demand response, and workload flexibility as parts of the same electrical architecture.
A grid-positive data centre is therefore less a single technology than an operating model. The difficult part is proving that compute flexibility, generation, storage, and the connection agreement work together while preserving the availability customers expect.
