Why GPU watts are not a data center power budget
From chip power to IT equipment, facility demand and annual energy.
Explore resource →The publication
Good questions make better power decisions.
Technical explainers, worked examples and source-led research. Commercial contributions are labeled and held to the same evidence standard.
Browse the guide library
Transformer sizing: inputs before answers
A practical framework for understanding load and rating.
Explore resource →Researching 34.5 kV switchgear
What to document before a technical specification is written.
Explore resource →Liquid-filled vs dry-type transformers
Compare project constraints rather than universal winners.
Explore resource →How to research transformer lead times
Separate dated market context from a project-specific delivery commitment.
Explore resource →No matches. Try another keyword or select all categories.
From the editorial desk
400 V vs 480 V: Three-Phase Current at the Same kVA
At the same balanced three-phase kVA, 400 V requires 20% more current than 480 V. That is a conversion result, not an equipment interchangeability rule.
Read article →DC Voltage Drop: One-Way Cable Length vs Loop Length
The current travels through both outgoing and return conductors. Whether you enter one-way or total length depends on what the calculator already counts.
Read article →UPS Runtime at Half Load: Why Doubling Is Only an Estimate
Halving a UPS load doubles runtime only when usable energy and efficiency stay constant. A worked example shows what the shortcut can and cannot tell you.
Read article →Data Center Generator Sizing: Running Load vs Load Steps
A 500 kW running-load estimate does not prove that a generator can accept the load sequence. Learn what a steady-state envelope leaves out.
Read article →5% vs 6% Transformer Impedance: Fault Current Compared
Changing transformer impedance from 5% to 6% reduces this idealized terminal fault-current estimate by 16.7%, not 1%. Follow the calculation and its limits.
Read article →Transformer Spare Capacity vs Redundancy: Avoid Double-Counting Reserve
A transformer can have spare capacity and still be a single point of failure. Use a simple reserve ledger to separate growth allowance from outage capability.
Read article →PUE 1.2 vs 1.4: Annual Energy for a 10 MW IT Load
At a constant 10 MW IT load for 8,760 hours, PUE 1.2 versus 1.4 changes modeled annual facility energy by 17.52 GWh. See the assumptions and lower-load case.
Read article →Transformer Losses at 50% vs 80% Load: A Worked Comparison
A lower no-load loss does not guarantee lower total losses at every operating point. Compare two hypothetical transformers at 50% and 80% loading.
Read article →AI Rack Power Budget: A Whole-Server Worked Example
Turn a whole-server power schedule into a rack and cluster budget, without mixing facility overhead, shared equipment or redundant feed capacity.
Read article →N+1 vs 2N: A 1 MW Data Center Capacity Example
One megawatt of IT load becomes four duty blocks in this stated scenario. Compare five-block N+1 with eight-block 2N, and see what the arithmetic cannot prove.
Read article →Why GPU Watts Are Not a Data Center Power Budget
Learn why GPU wattage is not a complete data center power budget. Separate chip power, server input, facility demand and annual energy assumptions.
Read article →How to Research Transformer Lead Times
Research transformer lead times with a checklist for configuration, production milestones, drawing approval, testing and logistics. Record dated evidence.
Read article →Have something useful to add?
We welcome original explanations, documented case studies and corrections. We do not accept unverified certification claims, disguised advertising or articles written only to manipulate search rankings.
Read contribution guidelines →