as much as they can get
Co-authored with MB.
Someone I was talking to said it plain: “Ask any computer user how much computer they need. They will tell you — as much as they can get.”
That line has been rattling around in my head because it explains something I’ve watched for years in managed IT and never quite named.
Nobody upgrades to use less. A company replaces a ten-year-old server and the old workloads move to the new hardware — but they don’t shrink. They get re-implemented. New OS, new hypervisor, new monitoring stack wrapping the same application that ran fine before. The application didn’t get more complex. The definition of “running it properly” did.
Then the headroom on that new box fills up. Not with the old workload growing. With new things that only exist because the capacity is there. A 2015 server that ran your whole production environment now runs the observability layer for your production environment. The machine didn’t get weaker. You just found more to do with it.
This is what datacenters look like from the outside: a business that should plateau and never does. More floor space, more power, more cooling — not because old workloads need more room, but because new workloads show up faster than anyone retires the old ones. Every efficiency gain gets reinvested immediately. The compute gets cheaper per unit and the bill stays the same, because the unit count went up.
Silicon will continue to be expensive. Not because it’s scarce. Because it’s useful. Ask anyone with a computer whether they’d take a faster one. The answer is yes. It has always been yes.