CCIR publishes a reference rate at each Intelligence Factory tier — Tier 1, Tier 2, Tier 3 — because the deployment grade attached to a SKU (fabric, host class, operational bundle) determines what a dollar of GPU-hour actually buys.
The same H100 silicon is deployed across all three tiers. A buyer choosing among a Tier 1 hyperscale deployment, a Tier 2 dedicated cluster, and a Tier 3 marketplace listing isn't choosing between three prices of the same product — they're choosing between three products that happen to share a chip.
| Silicon | T1IF | T2IF | T3IF | ||||||
|---|---|---|---|---|---|---|---|---|---|
| Median $/hr | Trend | n | Median $/hr | Trend | n | Median $/hr | Trend | n | |
| GB200 CRI-T1IF-GB200-ALL-GTD-OD-ALL | $13.52 | 3 | — | — | — | — | |||
| B200 CRI-T1IF-B200-ALL-GTD-OD-ALL | $8.60 | 3 | $6.59 | 5 | $3.77 | 4 | |||
| H200 CRI-T1IF-H200-ALL-GTD-OD-ALL | $7.11 | 4 | $4.79 | 5 | $3.44 | 7 | |||
| H100 CRI-T1IF-H100-ALL-GTD-OD-ALL | $6.52 | 4 | $3.59 | 7 | $2.69 | 9 | |||
| A100 CRI-T2IF-A100-40GB-ALL-GTD-OD-ALL | — | — | $1.99 | 3 | $1.07 | 6 | |||
Each tier is a defined capability bundle — fabric, host, support, compliance. CCIR publishes one reference rate per tier so users can match the index cited to the deployment grade they pay for, narrowing basis risk relative to a single-rate alternative.
Top-grade fabric, dedicated single-tenant bare-metal, full enterprise bundle — managed operations, tiered support, observability, egress allowance, regulated-workload compliance. The rate buys capability that translates directly into useful compute per dollar at production scale.
Dedicated single-tenant bare-metal with disclosed fabric, basic support, no managed-ops wrapper. Where most production training spend lands. The middle grade — the rate is lower, but each $/GPU-hr buys less of the bundle that turns silicon into output.
Co-located or marketplace clearing on variable Ethernet fabric. Self-serve, no support tier, interruptibility common. The lowest rate band — and the lowest effective throughput per nominal $/GPU-hr once interrupt risk is priced in.
| Attribute | T1IF | T2IF | T3IF |
|---|---|---|---|
| Tenancy | Dedicated bare-metal, single-tenant | Dedicated bare-metal, single-tenant | Shared-pool or marketplace |
| Fabric | IB-NDR or eq. (3.2 Tbps RDMA, 8×400 Gbps) | IB-HDR / RoCE-v2 / EFA / mixed disclosed | Ethernet / variable / undisclosed |
| Bundle depth | Managed ops, tiered support, observability, egress allowance, compliance | Basic support, no managed ops | Self-serve, no support tier |
| Typical operators | AWS · Azure · OCI · GCP · CoreWeave · Nebius | Lambda OD · RunPod Secure · Crusoe · Hyperstack | RunPod Community · Vast.ai · Together MoE |
| Use case | Production training, regulated workloads | Dedicated training, less ceremony | Cost-sensitive batch, dev/test, spot |
Each tier is a capability bundle. The matrix describes what the rate at each tier actually packages — fabric, host, bundle depth, typical operators, use case.
Provider-posted rate cards and on-demand pricing pages, scraped daily. No solicited quotes, no broker feeds.
Operator-equal median: each operator counts as one vote regardless of SKU count. Median of per-source medians at each tier.
3 sources / 5 observations to publish. Below-gate cells emit as Shadow series and are flagged in the explorer.
Per-SKU algorithmic classifier on a capability bundle (fabric, host topology, support, compliance). Premium silicons only.
See /tiers for the editorial framing, /spreads for the T1IF-anchored spread matrix, and /applications for worked credit-document examples.