Global data center ambition is sky high. Projects are attracting billions of dollars in investment around the world, and emerging companies, such as Nscale, are raising rounds at massive valuations on the promise of bringing AI infrastructure from concept to reality. But despite ambitious targets and the funding to match, data center plans across the globe are about to hit a critical roadblock: the grid.

Energy infrastructure is under significant strain. Our rapidly electrifying societies are putting more and more pressure on a creaking grid that wasn’t designed to handle this level of demand. In recent years, we’ve seen its impact on clean energy in Europe; thousands of so-called “zombie projects” are languishing in queues, from Germany to the Netherlands, while projects in the UK face waits of up to 15 years for a connection. There simply isn’t enough capacity to go around.

With peak AI power demand set to hit 50GW in the next four years, global data center ambitions risk running headlong into this grid-constrained reality. If we plough ahead and build now without thinking about the wider infrastructure, we’ll pay a high price later.

Building blindly

One of the biggest pitfalls when it comes to effectively planning new data center grid connections is a lack of visibility. Developers are drawing up plans without sufficient critical insights into the state of existing infrastructure and the constraints at play. Without these insights, it isn’t possible to create a clear roadmap for where, how, or even when new infrastructure should be built.

Instead, developers are building blindly. New data centers are being given the green light without sufficient planning or optimization measures to determine how successfully they can be connected to the grid or the impact they will have on local energy supply.

We’re already seeing the impact of poor planning begin to play out. In the UK, “phantom investments” appear to be paying little more than lip service to AI ambitions, and across Europe, grid delays are fueling connection queues, with waits of up to seven years. Meanwhile, large-scale projects funded by overseas investment – such as OpenAI’s Stargate UK – are threatening to withdraw due to concerns over capacity and cost.

Getting eyes on the grid

To stop global data center dreams from going up in smoke, we need to open our eyes to the grid, and fast. A clearer, more accurate understanding of existing infrastructure is essential to ensure that these ambitious plans are grounded in reality. The right combination of data, leveraged effectively, can give the visibility needed to shape new developments in a way that accounts for the physical constraints of the grid and helps allocate investment more efficiently.

Ironically, AI could be part of the solution. Large spatial models are enabling us to more accurately map real infrastructure and simulate the impact of new developments before shovels hit the ground. By coordinating multiple data sources and grounding them in physics-backed structural reality, these models can show how infrastructure will behave under real-world conditions more precisely than ever before.

One of the most critical ways this can support data center development is through assessing grid capacity. Physics-enabled modeling can be used to accurately pinpoint specific areas of the network where there is sufficient space for new connections to be made. Where networks are already operating at capacity, it can help safely identify where restrictions could be raised to create additional space.

For example, our recent analysis found that three-quarters of the UK and Ireland’s energy networks could be running at higher capacity. Insights like these can be used to drive more strategic siting, allowing developers to build with confidence in areas where the grid can reliably support new data center infrastructure.

Another way technology is primed to help strengthen data center planning is through modeling flexibility. Simulating dynamic power consumption could help ensure that data centers are designed in a way that enables their activity to be managed in response to fluctuating consumer energy demand. This balance will be crucial to reducing the impact of new data centers on grid capacity and local energy supply. Understanding exactly how this will work in practice, before projects get underway, could help to speed up timelines, reduce costs, and mitigate public opposition to new infrastructure builds.

Fostering system-wide collaboration

Increased visibility into the state of existing infrastructure is one part of the puzzle. Greater system-wide coordination is needed to translate this into tangible delivery. Right now, too much infrastructure planning is conducted in siloes, leading to stalled delivery and wasted investment. We need all stakeholders around the table from the word ‘go’ with shared access to critical grid data. All the insight in the world won’t move the needle unless it is leveraged collaboratively and fed into every stage of data center design.

Data center development will only be successful if it is shaped by the reality of our existing infrastructure, rather than swept along on a wave of market enthusiasm. To ensure developers can deliver on their economic promises, we must invest in more strategic, grid-centered planning, harnessing real, data-led insights to inform collaborative decision-making and optimize development. It’s an additional step, but a critical one. Without it, the AI infrastructure bubble is sure to burst, and there will be a high price to pay as a result.