Here a professional question calls for a computed figure, not an opinion. Each domain is backed by the engine that answers it, with its method, version and date. The examples below are computed as you read this page, not copied in.
The question : When will I have to place the order?
The question asked everywhere is “when will I be full?”. It is the wrong one: what decides is the order-by date — saturation minus procurement lead time. The bottleneck is therefore not the resource that saturates first, but the one whose order comes due first.
Computed example — A room using 45 of its 100 kVA of electrical power and growing 20 % a year, while keeping enough spare to absorb one failure: it will be full on 2 May 2027. But the order must be placed by 12 January 2027, i.e. 4 months earlier. The delivery lead time sets the date, not how fast you fill up.
The question : Is this site worth its price?
A rack’s listed price is not what you pay: energy, cross-connects, PDUs, IP addresses, remote hands and setup fees add up. Every listing that publishes its price list now carries its computed true cost, line by line, with unpublished items named rather than guessed.
The question : Who can deliver 10G here, and where does my traffic go?
Carrier presence per site, reachable exchange points, direct cloud on-ramps, observed and dated AS relationships, three-tier anti-DDoS — including the upstream transit provider, because volumetric protection is inherited and does not live on the customer AS.
Computed example — 14,535 networks classified by nature, and cross-referenced with what we observe site by site. Starting from a single IP address, we trace back to the company holding it, then to the network carrying it, then to the data centers in our catalogue where that network is present. No public registry makes that chain on its own.
The question : Does this site deliver what it claims?
A “Tier III” badge is shown solid only when it carries a number and an expiry; otherwise it is marked “declared”. Certifications expire. Sovereignty outside the US CLOUD Act is stated when observed. Nothing that is not verifiable or dated.
The question : How many nodes, which storage, how many kVA?
Full sizing of a Proxmox VE 9 private cloud (quorum, N+1, Ceph replica 3, 3-2-1 backup, subscriptions at the vendor’s public list price) and high-density GPU hosting, where the trade-off is the number of racks and never the space.
Computed example — For 40 virtual machines of 4 processors and 16 GB of memory, with a high availability requirement: 4 servers of 8 cores and 64 GB each. Enough is left over that one failing does not stop you. Memory is never promised twice; computing power is, because it can be shared.
Computed example — For ten RTX PRO 4500 graphics cards used for artificial intelligence: 2 servers, fitting in 8 rack slots. Space is not the problem. The problem is subscribing 4.1 kVA of power, where a standard colocation rack supplies 3.
The question : What does a kW cost, and what moves it?
A rack’s cost is computed from the local industrial tariff, the site’s PUE, retail colocation margin and local market tightness. The formula is published and versioned, and the gap between our calculation and the recorded price is published rather than smoothed.
The question : Who should run it?
SIP trunking, PBX, managed services and on-site work: matching runs on measured criteria — dated and perishable availability, observed response time, published ranking scale. It is the thinnest of the seven domains today, and we say so.
What this page refuses to be: a catalogue of generic articles. An engine that computes can be checked and challenged; an article can only be copied. If one of these calculations looks wrong to you, the method is published — tell us where.