Quiz: The Power-Bound Era & AI Infrastructure Roadmaps (Part 16)

10 questions from Part 16 of the guide. Score instantly; every explanation links to the chapter that derives the answer.

The central claim of the power-bound era is that the binding constraint moved from…
software to hardwarecapital to laborland to waterthe accelerator you can buy to the megawatt you can energize (and its date)
A market brief compares the generation-and-storage queue with AI data-center capacity under construction. What can the developer infer directly from that ratio?
the share of queued generation available to the data centersthe data centers' energization datesnothing without reconciling population, unit, geography, readiness stage, withdrawal risk, and the utility's large-load processthe projects' network-upgrade costs
The guide's advice on the facility 'substrate' is to…
pre-invest in floor loading, water headroom, electrical risers, and pipe-rack space for racks that don't exist yetlet tenants fund all upgradesbuild exactly to current needsavoid any speculative capacity
Budgeting inference revenue against which curve is the guide's rule?
the ~10×/yr algorithmic-and-price deflation curveGPU rental prices5% annual price growthflat $/token
The AI-optimized facility build-cost band the guide uses for underwriting is…
~$100M/MW~$15–25M/MW (shell + fit-out, excluding IT)~$5M/MW all-in~$1–2M/MW
The three demand scenarios the guide says you must consciously underwrite against are…
supercycle, digestion, and overbuildbull, bear, basetraining, inference, edge2026, 2028, 2030
Which subsystem trends does Part 16 consolidate into one roadmap view?
none; roadmaps are per-vendor onlyonly GPUsdensity, power (DC architectures), cooling, network, and silicon — because they co-movemarketing and sales
A 1 GW AI campus's total program cost runs approximately…
~$3B~$10B~$100B~$38B (~$8.5B/yr)
Announced vs real: of the ~12–16 GW of US capacity targeted for 2026, how much was actively under construction?
All of it~4–5 GW — roughly one-thirdNone~8–10 GW
The organizational re-orientation the power-bound era demands is…
hiring more ML engineersre-organizing the entire program around time-to-megawatt rather than chip allocationcentralizing in one regionoutsourcing all construction

Check your understanding: Trends, Roadmaps & the Future