AI Data Centers: Britain's Environmental Bill Slips Through the Cracks

Britain's AI and data center policy was shaped through consultations with 137 companies in the sector. Over the same period, not a single formal meeting was held with any environmental organization. According to a review by a corruption watchdog, dozens of ministers, senior civil servants and advisers moved into positions at AI companies while the "AI growth zones" plan for data centers was being drawn up.

What happened?

The non-profit organization Spotlight on Corruption found that a "significant revolving door" of traffic shaped the government's AI policy. The review documents a large number of individuals who moved from public service into AI companies during the policy's development period. Environmental groups, meanwhile, were never brought to the table. In other words, the side that would have argued for items like energy and water consumption was absent from the process.

Why does it matter? The invisible infrastructure bill of data centers

The AI debate is largely conducted in terms of model capabilities and investment figures. Yet behind every large language model lies a physical infrastructure: data centers that house thousands of GPUs, run around the clock and draw serious amounts of electricity. These facilities also use water to cool the chips as they heat up. With the wave of generative AI, this burden has grown rapidly.

The problem is not unique to a single country. In recent years, major cloud providers have admitted that they are struggling to meet their carbon targets; AI-driven energy demand has pushed those targets back. In some regions, data centers have strained the local grid so much that new connection requests have been deferred. Electricity and water are not infinite resources; every megawatt that goes to a data center is a demand taken from another user.

This is precisely why it matters who the policy was written in consultation with. If the only party at the table is the one that wants to scale up investment, the invisible part of the cost — the environmental burden — does not get properly factored in. The revolving door adds to this: the fact that some of the people who drafted the policy later found jobs in the same sector legitimately raises the question of whose interests the decisions served. In academic literature, this is called "regulatory capture" — when the regulator adopts the priorities of the very sector it is supposed to regulate.

Who is affected, and how?

The picture clarifies the question of who wins and who carries the bill.

The way benefits and costs are distributed across different parties is a classic sign of imbalance in public policy. Investment is centralized and measurable; the environmental cost, by contrast, is diffuse, delayed and hard to measure. This asymmetry leads to the cost being overlooked in any process where the environmental side is not represented by a strong voice.

What does it mean for Turkey and other countries courting data center investment?

This debate will not stay confined to Britain. Many countries looking to attract data center investment face a similar dilemma: whether to bring investment in quickly, or to build a framework that accounts for energy and water costs from the outset. For countries with high energy costs and water stress — including Turkey — this balance is even sharper. Investment may arrive on the back of cheap land and incentives; but if a facility's lifetime energy and water needs clash with local resources, the bill falls on society.

The UK example can be read as a "how not to do it" case. The lesson to draw is not to oppose AI infrastructure; it is that when writing policy, the table should include not just the investor but also the energy planner and the environmental side.

What's next?

There are several threads to watch. The first is how the government will respond to this review and whether it will tighten the revolving-door rules. The second is the energy source for the AI growth zones: will the new capacity be fed by renewables, or will it pile extra load onto the grid? The third is whether regulation will emerge requiring data centers to report their water and energy consumption on a mandatory basis.

The growth of AI is now an infrastructure question. Who decides the cost of that infrastructure, through what process, and by listening to whose voice, will be at least as decisive as the models themselves. If a party is missing from the table, the outcome will be incomplete too.