Equinix plans a large multi stage data center build on a former AkzoNobel paint factory site in Slough, but power constraints mean the full build may take up to 20 years. In parallel, another developer is seeking approval for a separate Slough data center plus battery storage on green belt land, with the final decision escalated to the UK government
My Analysis:
This is what AI infrastructure buildout looks like when it hits the limits of the power grid. Equinix has land, a motivated council, and a proven hub in Slough. What it does not have is enough power to light the whole thing up this decade. Phased build over potentially 20 years is a clear signal that grid capacity, not capex or demand, is the gating factor.
Slough is already one of Europe’s core data center clusters. Adding more facilities here is not about generic cloud. It is about concentrating low latency access to major AI and digital workloads close to London and key network routes. That “power only available in coming years” language is doing a lot of work. It tells you utilities and regulators are rationing capacity and sequencing which projects get connected when.
The second site with a dedicated Battery Energy Storage System is another important tell. Developers are trying to buffer around grid limits with storage and on site generation. Not because it is trendy, but because it is becoming mandatory to get big projects approved. We will see more AI and data center sites with integrated BESS as part of the planning story: “we are not destabilizing your grid, we are helping manage it.”
From an AI perspective this is exactly the kind of region where GPU dense clusters will land. Existing fiber, ecosystem, and skills are there. But if power for later phases only shows up closer to 2038, that caps how much sovereign AI or private AI capacity the UK can pack into Slough without major transmission upgrades. That opens the door for neoclouds and regional providers to look at alternative UK and EU locations where power and planning come faster.
The long timeline also matters for enterprises. You cannot assume “just put it in London” will be an option at scale. Capacity is being pre allocated by hyperscalers and big colo like Equinix. If you need GPU racks near major metros, you will need to plan multi year commitments, accept secondary metros, or look at sovereign / regional providers that can move quicker with local authorities.
The Big Picture:
This hits several macro trends at once.
AI data center construction surge:
Slough is already a hub and still getting more proposals, including on former industrial land and even green belt. That is classic infill and edge expansion behavior when a region becomes a strategic compute cluster. Everyone wants in, but the easy builds are long gone.
Energy and grid constraints:
The 20 year horizon is the loudest signal. You do not stretch a project like this over two decades because you want to. You do it because the transmission build out and substation capacity cannot be delivered on AI’s current hype cycle timelines. Expect more “build in stages as power arrives” designs across Europe and the UK.
NIMBY vs YIMBY for AI buildouts:
You see the tension in the second project. Green belt use, BESS, generators, and a decision escalated to the national housing and communities secretary. Local willingness to host data centers is now a national policy question in the UK. That will directly shape where AI clusters can be built and how “sovereign AI” capacity gets distributed.
Vendor ecosystem dynamics:
Equinix is cementing its role as neutral interconnect and AI colo backbone. Councils under financial pressure are selling strategic land to players like Equinix instead of using it for housing. That shifts long term land use away from residential and toward long lived digital infrastructure, and it gives large colo providers even more control over where AI workloads can practically live.
Sovereign AI and neoclouds:
If Slough and similar hubs are power constrained, governments and sovereign AI initiatives will start looking at alternative cities and regions where the grid is less saturated. That is fertile ground for neocloud providers focused on specific countries or regulatory regimes. They can partner with smaller councils eager for investment and willing to be YIMBY for data centers in exchange for tax base and jobs.
This story is not about one more data center. It is a case study in the new bottleneck for AI infrastructure in Europe. Power first, land second. The operators that can secure both, and manage the politics, will control where the next decade of AI compute actually runs.
Signal Strength: High
