Candidate benchmark for global NVIDIA H100 SXM on-demand compute pricing, normalized into USD per GPU-hour.
candidateSXMGlobalOn-DemandMethodology 1.0
Candidate Benchmark. This series is methodology-defined but is not represented as an official or regulated reference rate. A live candidate is shown only when current public evidence clears the registered thresholds.
Current level
Live candidate assessment
No published institutional fixing exists yet; the value below is recomputed from current eligible public evidence.
Statistics are calculated across provider-level contributions, not raw listing count.
Low$1.99
Q1$2.99
Median$3.63
Q3$4.41
High$12.29
Publication gate
Current evidence clears thresholds
Providers17 / 3
Regions12 / 2
Comparable observations970 / 5
Freshness window30 days
Availability & capacity depth
Current market structure
Tracked availability and explicit GPU counts are kept separate from the price fixing. Unknown capacity is not converted to zero and current supply evidence is not a reservation.
Tracked offerings49Known available GPUs0Status coverage2.0%Available / limited100.0%Capacity providers0Priced capacity0
Availability evidence
Known status mix
Available1
Limited0
Unavailable0
Unknown48
Capacity evidence
Explicit vs unknown counts
Offerings with explicit capacity0
Available / limited with unknown capacity1
Providers with explicit capacity0
Capacity depth curve
Known GPUs at or below tracked price ceilings
No current price-linked explicit capacity is available for this series.
Constituent panel
Current provider contributions
One contribution per independent provider. When multiple comparable observations share that provider’s latest eligible timestamp, Kovara uses their median before the cross-provider fixing.
No institutional fixing has been published for this series yet.
Methodology
Kovara Compute Index Price Methodology 1.0
Produce transparent candidate reference levels for exact-GPU public on-demand compute pricing using one provider contribution per independent provider.
