The Definitive Guide toAI Data Centers
Ask the GuideAboutAccount
Guide › Numbers register › Claim

Illustrative all-in capacity cost for one fleet-hour: Chapter 1.8's three-year owned cost for the same 64-GPU fleet

$156/fleet-hourmodeled

Value kindmodeled — Interpret this value according to its displayed kind, scope, source, and as-of date.
Scope1.3 ledger only, 64 installed GPUs including idle and reserve. Equals Chapter 1.8's owned case, (K/A + O)/8,760 hours with K = $2.4M, O = $0.40M/year and A = the three-year annuity factor at 10% (about 2.487): about $2.4 per installed GPU-hour. Common downstream application cost excluded from the output ratios. Assumed sensitivity: 126–206 $/fleet-hour, from Chapter 1.8's K ($2–3M) and O ($0.30–0.60M/year) bounds.
As of2026-10-19
SourceGuide assumption, 2026-10-19, derived from Chapter 1.8's owned-fleet inputs. Epoch supplies the cost method, not these prices.
Reviewchecking…review by 2027-04-19 · standard cadence
Recorded changeslast 2026-10-19 · 2 revisions tracked
Claim idguide2-serving-hour-cost

Where the guide uses it

← Full numbers register — every date-stamped figure in the guide, with revision history.