AI Data Centre Power Wait Could Drop to 2 Years via Flexible Grids: BCG
BCG reports that flexible grid connections can slash AI data centre power wait times from 10 years to just 2-4 years.
The explosive growth of artificial intelligence is hitting a massive roadblock: electricity. According to a new report by Boston Consulting Group (BCG), the waiting period for AI data centre power connections, which currently stretches between five to ten years or more, could be slashed to just two to four years. The secret lies in a shift toward "flexible grid connections" and a move away from total reliance on guaranteed power supplies.
The Power Crisis Facing AI Expansion
Power availability has evolved into one of the most significant hurdles for the expansion of data centres globally. In several markets, connection queues have become so long that they threaten to stall the deployment of next-generation AI infrastructure. The scale of the problem is evident in the projected demand: BCG estimates that global data centre power needs will skyrocket to approximately 210 gigawatts by 2030, a massive jump from the 84 gigawatts expected in 2025.
How Flexible Connections Work
Traditionally, data centres wait for "firm" connections—guaranteed, uninterrupted power supplies that often require major reinforcements to the existing electricity grid. This infrastructure work is what causes the decade-long delays.
The proposed alternative is the "non-firm" or flexible electricity connection. Under this model, data centres would agree to have their power supply reduced during periods of high grid congestion. To ensure they don't go dark during these dips, data centres would implement a hybrid power strategy including:
- On-site power generation: Producing their own electricity to fill gaps.
- Battery storage systems: Storing energy for use during grid constraints.
- Flexible computing workloads: Adjusting processing tasks to match available power.
The Billion-Dollar Incentive
While the technical shift is complex, the financial incentive is staggering. BCG modeled a 150-megawatt AI inference data centre in Northwest Europe and found that connecting just three years earlier could unlock an estimated USD 1.28 billion in additional profits, even before factoring in the costs of the on-site power assets.
The Trade-offs and Costs
However, speed comes at a price. The report warns that moving toward an off-grid or highly flexible setup could increase annual power supply costs by 70 to 230 per cent compared to a standard firm grid connection, depending on how the assets are financed and accounted for. Such setups are most viable where flexible connections are unavailable or where the cost of grid reinforcement is prohibitively high.
A Strategic Shift for Governments and Operators
The analysis, which focused on Northwest European markets—including the UK, Germany, France, Belgium, and the Netherlands—for the 2030-2035 period, suggests that data centres don't have to be a burden on the system. "Grid-positive" data centres can actually help spread fixed network costs across a larger consumer base and reduce the overall need for new infrastructure.
To make this a reality, BCG recommends several urgent actions:
- National Strategies: Governments should create dedicated data centre strategies to guide projects toward areas with existing grid capacity.
- Faster Approvals: Streamlining the permit process for on-site power generation assets.
- Standardization: Grid operators should make flexible connections a standard offering rather than a rare exception.
Ultimately, the report emphasizes that whether a data centre strengthens or strains the energy system depends entirely on where it is located, how it connects, and how it is operated once it goes online.