In March, I spoke at DCD Connect in New York about a tension that is rapidly moving from background noise to a defining constraint: the collision between global demand for digital infrastructure and the very local realities of energy systems.

At one level, the story is simple. Demand is exploding. AI, electrification, and re-industrialization are combining to drive a structural step-change in electricity consumption. Hyperscale data centers – once a rounding error in grid planning – are now material loads, often measured in gigawatts. In some US states, a single campus can rival the demand of a mid-sized city.

The new shape of demand

For most of the past two decades, electricity demand in developed economies was broadly flat. Efficiency gains offset economic growth. Energy‑intensive industries moved to cheaper locations. Grid planners operated in a world of incremental change. That world has gone.

AI workloads are fundamentally different from traditional cloud. They are power‑dense, continuous, and far less tolerant of interruption. Training clusters do not flex around grid constraints; they require firm, high‑quality power at scale.

Add to this the electrification of transport and heating, and a degree of industrial reshoring, and the result is a demand profile that is both larger and less flexible than anything grids have had to accommodate in recent history.

In the US, this is beginning to show up starkly in forecasts. Utilities that were planning for low single‑digit load growth have revised projections upwards, in some cases dramatically. Interconnection queues are filling with projects that are not speculative renewables, but firm load.

Europe is following, albeit with a different flavor. The same AI‑driven demand exists, but it collides with tighter decarbonization constraints, higher power prices, and more complex market design. The result is not necessarily slower growth, but more policy‑led friction.

The grid was not built for this

Electricity systems are engineered for reliability, not speed. They are capital‑intensive, highly regulated, and designed around long-term planning cycles.

This matters because the constraint is no longer generation in the abstract. It is the physical and regulatory ability to deliver power to specific locations, at specific times, with acceptable reliability.

In the US, this plays out through a patchwork of state‑level regulation, independent system operators, and local permitting regimes. A developer may secure land, capital, and even generation, only to find that transmission upgrades take years and require multiple layers of approval. Community opposition, which is often framed around land use, water consumption, or perceived lack of local benefit, is becoming more organized and more effective.

In Europe, the constraint is even more explicit. Grid connection queues in key markets run into the 2030s or beyond. In countries such as Ireland and the Netherlands, authorities have already imposed de facto limits on new data center connections in certain regions. Italy is experimenting with prioritization mechanisms that favor projects offering system benefits, such as flexibility or co‑location with generation.

The underlying issue is the same on both sides of the Atlantic: the system was not designed to accommodate large, concentrated loads arriving at speed.

Regulation will do what it is designed to do

There is a tendency in parts of the technology sector to view regulation as an obstacle: something to be navigated, influenced, or, ideally, avoided. That is a mistake.

Regulators are not failing when they slow projects down. They are doing exactly what they are designed to do: protect system reliability, ensure fair access, and balance competing societal objectives.

A useful lens here comes from Stafford Beer, a management cybernetician who argued that systems should be judged by their outcomes, not their intentions.

The purpose of a system is what it does. If data center projects are being delayed, constrained, or rejected, it’s not because the system is broken. It is because the system is prioritising other outcomes, such as grid stability, affordability, and political acceptability, over speed of deployment.

In the US, this is beginning to surface in more assertive regulatory scrutiny. Questions around cost allocation (who pays for the grid upgrades required by large new loads) are becoming central. So too are concerns about cross-subsidization, where residential or industrial customers effectively underwrite infrastructure for hyperscalers.

In Europe, the regulatory response is more overtly policy‑driven. Data centers are increasingly being treated as strategic infrastructure, but also as controllable demand. Access to power may be conditional on providing flexibility, participating in demand response, or aligning with decarbonization objectives.

The coming reckoning

The industry is entering a phase where the limiting factor is no longer capital or demand, but the ability of local systems to absorb that demand.

This will not result in a uniform slowdown. It will create divergence. Regions that can align regulation, grid investment, and community acceptance will attract disproportionate investment. Those that cannot will see projects delayed, downsized, or redirected elsewhere.

For developers and operators, the implications are clear:

• Engage earlier and more deeply with the energy system: Waiting until late‑stage development to secure power is no longer viable.
• Treat regulators as stakeholders, not obstacles: Understanding their objectives and aligning with them will be a source of competitive advantage.
• Prioritize flexibility: The ability to modulate demand, provide system services, or integrate with local generation will increasingly determine access to power.
• Rethink site strategy: Location decisions must be driven as much by grid realities as by network latency or land cost.

For policymakers, the challenge is more delicate. There is a genuine tension between enabling digital infrastructure – now clearly of strategic importance – and maintaining the integrity and fairness of the energy system. That tension is not going away. If anything, it will intensify.

A final thought

The narrative of the past decade for technology infrastructure has been one of near‑frictionless scaling. Build capacity, deploy capital, meet demand. The next decade will be different.

Data centers are no longer just consumers of electricity. They are becoming central actors within the energy system – large, inflexible loads that must be integrated into networks designed for a different era.

The industry is not running out of demand. It is running into reality. And reality, in this case, is local.