Insights · Editorial
Long enough to argue with.
Short enough to actually read.
We publish when there is something specific to say — not on a content calendar. Each post tries to give you the load-bearing point in the first paragraph and the supporting reasoning underneath, so you can stop reading whenever you have what you need.
-
NIS2 essential vs. important entity — what it means for where you host
Most colocation buyers we talk to know they fall under NIS2 — but cannot say whether they are an essential or important entity. The difference changes the facility shortlist.
Read → -
PUE belongs in the procurement filter, not in the marketing deck
Power Usage Effectiveness has been a sustainability talking point for fifteen years. It is also one of the few facility metrics that translates directly into operating cost. We treat it as a hard input, not a brochure number.
Read → -
Datacenter tier classifications, the version that actually filters facilities
Tier I through IV is a Uptime Institute framework, not a quality grade. We use it as a hard filter on the shortlist — and we tell buyers when their stated tier is over- or under-specified relative to the workload.
Read → -
Direct-to-chip vs. immersion cooling — pick the one your facility can actually deliver
Liquid cooling has stopped being a research topic and started being a procurement question. The right answer depends less on the technology and more on what the facility is willing to commit to in writing.
Read → -
AI training infrastructure — power, cooling, and network, in that order
GPU clusters fail at the facility, not at the chip. Most AI infrastructure procurement gets the order wrong — picking compute first and trying to retrofit power and cooling. We work the other way around.
Read →