The wording is certainly interesting. This article is written as if cloud was a holy grail, with companies still being stuck in the stone-age era of.... owning servers. They also pretend that the cloud is the future for everything, and that on-premise servers are legacy technology.

The following paragraph is the most illuminating: "The decision to build or rent may depend on a user’s data intensity. Consider a firm in America with a medium-sized data centre which thinks its computing load will rise roughly four-fold over a decade. In that case, building its own breaks even in seven years or so, and ends up being 5% cheaper overall than leasing, according to data from Schneider Electric. If the load remains stable, renting is the less expensive option, by a similar amount (assuming a flat cost of capital)."

This assumes that cloud costs are stagnant and reliable. This is a very naive assumption - it can be seen that if the cloud providers think they can get away with cost increases, they will. On-premises works out cheaper in the majority of cases, even including the cost of a sysadmin team.