the physical foundation. before energy, water, siting, and responsibility, you need to understand what these buildings actually are and what they consume.
You interact with data centres at least fifty times a day. Every email, video stream, banking transaction, and AI query depends on physical infrastructure: buildings full of servers, cooling systems, and power supplies. Most people have no idea what this costs, who pays for it, or what it takes from the grid and the ground.
This course exists because those costs deserve an honest account. The AI industry moves fast and talks loudly. The environmental consequences move quietly: a planning application here, a grid connection agreement there, a water abstraction licence that nobody notices until a dry summer arrives.
We believe in understanding things before adopting them. That applies to AI tools. It applies equally to AI infrastructure.
A data centre is a building, or a campus of buildings, that houses computers (servers), storage equipment, and the infrastructure needed to keep them running twenty-four hours a day, every day of the year. Power. Cooling. Security. That is it.
The earliest versions were server rooms inside offices. As demand grew, companies realised they could share buildings: multiple organisations renting space in the same facility, sharing power and cooling infrastructure while keeping their equipment separate. This model, colocation, remains common today.
Two main types now dominate. Colocation providers like Equinix, Digital Realty, and Virtus lease space to multiple customers. They typically rent from property owners on twenty to thirty year leases, converting warehouses and logistics parks into data centres. Hyperscalers, Amazon, Microsoft, Google, Meta, build and own their own facilities. Long-term investment into land, buildings, power infrastructure, and equipment. Single-organisation campuses at a scale the colocation model does not reach.
this is a condensed version of the lesson. the full version, with sources and figures, is being rewritten in this format. if you'd like an early read, the feedback form is open.