In the days of on-premises data centers, when procurement cycles were measured in months, overprovisioning was a logical risk management strategy — so commonplace that no one gave it a second thought.
In theory, the rise of elastic utility compute offered by public cloud providers should have gotten us away from it, but it simply hasn’t. Companies I’ve worked with are typically overprovisioned by at least 50%, even in the cloud. Zombie servers, which were once doing useful work but are no longer, abound.
