A data centre is a building made for one job: housing thousands of servers and keeping them running. The servers are stacked one above another in racks, tall metal frames that each hold dozens of servers. Many data centres rent out space by the rack, by a locked cage of several racks, or by a whole room. The companies renting it bring their own servers, so the servers of many different apps sit in the same building. The building is connected to many telecom operators by fibre links of high bandwidth, meaning a lot of data can flow at once, and low latency, meaning very little delay. There are several links, so if one is cut, the others keep the traffic flowing. This is called redundancy. The building runs round the clock. It draws a lot of electricity from the grid, and keeps batteries and diesel generators ready to take over the moment the grid fails. All those servers turn electricity into heat, so large air-conditioning systems, and now liquid cooling, stop them from overheating. Guards, cameras and locked cages let many companies share one building without anyone reaching another company's machines. This sharing is called multi-tenancy.

Owning and running a data centre is a business of its own. It is separate from making the servers and equipment inside, which is covered in Market Segment 3.2, Server and storage hardware. It is also separate from renting out computing power, which is the cloud. That is covered in Market Segment 5.1, Cloud infrastructure and web hosting providers
Keep reading, free
Three more sections are on this page: India's data-centre boom, Renting space in a shared building, and A data centre that travels. Sign in to read them here, in full.
Continue with Google- India's data-centre boom
- Renting space in a shared building
- A data centre that travels
Data centres in India. India is one of the fastest-growing data-centre markets in the world. Three things drive it:
- more Indian businesses and apps moving to the cloud
- rules that require some data, like payment data, to be stored inside India
- the rush to build computing for AI
Mumbai is the largest hub, partly because many undersea internet cables land there. Chennai, where many cables also land, is another major hub. Hyderabad, Delhi NCR, Pune and Bengaluru are growing fast.
Two kinds of companies build these data centres. Data-centre operators, such as Sify, CtrlS, Yotta, AdaniConneX, STT GDC and NTT's India business, build buildings that other companies rent space in. These operators are not covered in this Market Segment, since most of their engineering jobs are in building and running the data centres, not in software. The biggest cloud companies, called hyperscalers, such as Amazon, Microsoft and Google, also build their own data centres. They are covered elsewhere.
This Market Segment covers two companies, each with a different angle: a global operator with an engineering team in India, and a start-up that puts a whole data centre in a shipping container.
A rack of AI accelerators, the graphics chips (GPUs) that train and run AI models, uses many times the power of an ordinary rack. It also gives off so much heat that air cannot cool it. It needs liquid cooling, with cold liquid piped right up to the chips. So data centres are being redesigned from the floor up. Operators used to look first for sites near fibre. Now they look first for sites with plenty of electricity.

Who these companies hire, and for what, is on What data centres and servers hire for.