The on-demand public rental rate of AI compute, captured in a fixed window across a twenty-seven provider panel and reported in full. The anchor configuration is the H100 SXM 80GB. United States dollars, per GPU-hour.
| Index | Value | Basis |
|---|---|---|
| SVCI Public GPU Rate Index | 100.0 | Headline composite, weights 0.45 / 0.45 / 0.10 |
| Compute Opacity Index | 26.0 | 100 times (1 minus CompositeCoverage) |
| Hyperscaler Premium | 2.9497x | Tier 1 median over the pooled non-hyperscaler median |
| Neocloud Discount | 0.6629 | Tier 2 median relative to the Tier 1 median |
| Price Floor | 2.5450 | Lower-quartile headline-rate floor marker |
| Coverage Score | 0.7400 | Weighted panel coverage |
The Confidence Score is the seventh index in the locked methodology. It is not published in Issue 001 pending the per-row confidence audit, and it returns in a later issue once that audit is complete.
Receive the next SVCI issue, the market reading that explains it, and new buyer tools as they launch. The public issue stays open; the subscription delivers the signal without making you check the site.
| Tier | Weight | Panel slots | N in window | Median USD |
|---|---|---|---|---|
| Tier 1, Hyperscalers | 0.45 | 4 | 4 | 10.5306 |
| Tier 2, GPU clouds and neoclouds | 0.45 | 18 | 10 | 3.5500 |
| Tier 3, Marketplaces and aggregators | 0.10 | 5 | 2 | 4.6200 |
Panel slots are the locked twenty-seven provider composition across three tiers. N in window is the count of in-window observations that resolve to a comparable public rate. Medians are computed from the official capture.
SVCI Issue 001 reviews a twenty-seven provider headline panel. Sixteen providers published comparable public H100 SXM 80GB rates during the capture window, and those sixteen rates enter the headline medians.
The remaining eleven panel slots do not enter the medians. Seven are evidenced non-inclusions, with a documented public pricing posture that is sales-gated, contract-only, marketplace or dynamic, from-price, or non-comparable on product or unit basis. Four are panel-review, marketplace, or decided-out cases, and are disclosed separately in Appendix A.
| Status | Count |
|---|---|
| Headline-eligible public rates | 16 |
| Evidenced non-inclusions | 7 |
| Panel-review, decided-out, scout | 4 |
| Total headline panel | 27 |
| Tier | Resolved | Coverage |
|---|---|---|
| Tier 1, Hyperscalers | 4 / 4 | 100.0% |
| Tier 2, GPU clouds | 10 / 18 | 55.6% |
| Tier 3, Marketplaces | 2 / 5 | 40.0% |
| Weighted Coverage Score | 0.7400 | |
| Compute Opacity Index | 26.0 |
Confidence Score is not published in Issue 001 pending the per-row confidence audit. The Rule I-16 publication gate is met: coverage is non-zero in all three tiers.
Tier 3 is reported as thin coverage: two of the five Tier 3 panel members resolve to a comparable public fixed rate. The denominator is the full five-member tier, not the count of survivors; the fraction is itself a finding, consistent with the index measuring pricing opacity. Tier 3 keeps its locked 0.10 weight regardless of coverage. If the Tier 3 median sits above the Tier 2 median, that inversion is reported as a market-structure finding and not an error, because auction and aggregator rates are excluded as non-comparable.
Every published rate traces to a public source with a recorded URL and a UTC timestamp, and to a stored evidence artifact. Measured rates and any derived figures are labeled separately, so readers always know which they are reading. Corrections are made on the record: each correction is documented in the issue that carries it, with a plain-language statement of what was wrong, what is right, and how the error occurred. The corrections record is permanent and public at sorsoview.com/corrections.
The headline panel contains twenty-seven provider slots. The medians use only comparable public H100 SXM 80GB rates captured inside the official window. The remaining slots are disclosed below and do not enter the medians.
| ID | Provider | Status | Median | Reason | Note |
|---|---|---|---|---|---|
| E001 | AWS | Priced | Yes | Public on-demand p5.48xlarge, normalized per GPU by 8 | |
| E002 | Azure | Priced | Yes | Retail Prices API, ND96isr H100 v5, per GPU by 8 | |
| E003 | GCP | Priced | Yes | a3-highgpu-8g on-demand list, per GPU by 8 | |
| E004 | Oracle | Priced | Yes | BM.GPU.H100.8 public per GPU-hour rate |
| ID | Provider | Status | Median | Reason | Note |
|---|---|---|---|---|---|
| E005 | CoreWeave | Priced | Yes | HGX H100 on-demand, normalized per GPU by 8 | |
| E006 | Crusoe | Priced | Yes | H100 80GB HGX on-demand per GPU | |
| E007 | Lambda | Priced | Yes | H100 SXM 8x on-demand per GPU rate | |
| E008 | Nebius | Priced | Yes | H100 on-demand GPU-hour rate | Repriced effective June 1 2026 |
| E009 | Together AI | Priced | Yes | GPU Clusters on-demand H100 rate | Large move from prior anchor |
| E010 | Hyperstack | Priced | Yes | H100 SXM on-demand public rate | |
| E011 | Verda | Priced | Yes | H100 SXM5 80GB PAYG | Successor to DataCrunch |
| E012 | Voltage Park | Priced | Yes | HGX H100 SXM fixed On-Demand Ethernet tier | |
| E013 | Fluidstack | Evidenced non-inclusion | No | No comparable public fixed rate, contact sales | |
| E014 | Nscale | Evidenced non-inclusion | No | No comparable fixed public per-GPU rate | |
| E015 | IREN | Evidenced non-inclusion | No | Contract-led, no public fixed on-demand rate | |
| E016 | Applied Digital | Panel-review, decided-out | No | Panel-change governance item | Pending formal re-issue |
| E017 | GMI Cloud | Evidenced non-inclusion | No | From-price floor, not a fixed comparable rate | |
| E018 | Cudo Compute | Evidenced non-inclusion | No | Contact-us pricing, no comparable public rate | |
| E019 | Spheron | Evidenced non-inclusion | No | Marketplace from-price, spot posture | |
| E020 | Jarvislabs | Priced | Yes | H100 SXM on-demand per GPU rate | |
| E021 | Hyperbolic | Evidenced non-inclusion | No | Marketplace from-price, supplier-set dynamic rate | |
| E022 | Atlas Cloud | Priced | Yes | H100 SXM fixed on-demand per GPU rate |
| ID | Provider | Status | Median | Reason | Note |
|---|---|---|---|---|---|
| E023 | RunPod | Priced | Yes | Secure Cloud fixed H100 SXM on-demand rate | |
| E024 | Vast.ai | Panel-review, marketplace | No | Dynamic marketplace, non-headline treatment | |
| E025 | Paperspace | Priced | Yes | H100 on-demand promo rate | Commitment rate excluded, promo rate used |
| E026 | TensorDock | Panel-review, marketplace | No | Dynamic from-price marketplace treatment | |
| E027 | Shadeform | Panel-review, decided-out | No | Aggregator, panel-fit governance item | Pending formal re-issue |
| Category | Count |
|---|---|
| Priced, enter medians | 16 |
| Evidenced non-inclusions | 7 |
| Panel-review, decided-out, scout | 4 |
| Total headline panel | 27 |
By the Sorso View Editorial. Full construction at sorsoview.com/methodology.
Newer issues are available in the archive. Issue 001 remains the base issue of the series.