CCIR publishes a graded reference rate at each Intelligence Factory tier. Throughput is hard to standardize — usable compute per dollar varies with workload, framework, and scale efficiency — so CCIR does what commodity markets do when output is hard to measure: grade the inputs. Crude oil prices light sweet, medium, and heavy sour as distinct feedstock grades because refinery yield is hard to standardize but feedstock is observable. Compute is the same: fabric, host topology, and operational bundle are observable. MLPerf submissions on identical silicon show realized throughput varies materially with system configuration — fabric, host topology, and operational scaffolding — which is what the tier construct grades.
Top-grade fabric (≥ 3 Tbps/node disclosed RDMA — IB-NDR, EFAv2, or RoCE-v2 in non-blocking 8-rail builds), dedicated single-tenant bare-metal, full enterprise bundle — managed operations, tiered support, observability, egress allowance, regulated-workload compliance. The deepest capability bundle in the CCIR panel.
Dedicated single-tenant bare-metal with disclosed real fabric (IB-HDR, RoCE-v2 mid, or sub-3 Tbps IB-NDR variants), no requirement for full enterprise bundle. The middle grade — real fabric, lighter wrap.
Co-located, shared, or marketplace clearing — plus dedicated hosts that don't disclose a real fabric. Variable Ethernet underlay, self-serve, no support tier, interruptibility common. The lowest rate band — and the most constrained bundle behind it.
| Attribute | T1IF | T2IF | T3IF |
|---|---|---|---|
| Tenancy | Dedicated bare-metal, single-tenant | Dedicated bare-metal, single-tenant | Shared-pool or marketplace |
| Fabric | ≥ 3 Tbps/node disclosed (8-rail IB-NDR / EFAv2 / RoCE-v2 non-blocking) | ≥ 400 Gbps/node disclosed (IB-HDR / RoCE / sub-3 Tbps IB-NDR) | Ethernet, undisclosed, or sub-400 Gbps trivial RDMA |
| Bundle depth | Managed ops, tiered support, observability, egress allowance, compliance | Bundle varies — not gated; observed mix of basic to richer disclosures | Self-serve, no support tier |
| Typical operators | AWS · Azure · OCI · CoreWeave · Nebius | Crusoe · Denvr · Lambda Cluster · RunPod Cluster · RunPod Secure | Vast.ai · DigitalOcean · Hyperstack · Scaleway · DataCrunch · RunPod Community · Genesis Cloud · Massed Compute · TensorDock · Lambda OD |
| 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.
Median rate per chip across each Intelligence Factory tier. n is the number of independent operators contributing to that median. An em-dash means the cell has no Published row in today's snapshot.
| Silicon | T1IF Tier 1 IF | T2IF Tier 2 IF | T3IF Tier 3 IF | ||||||
|---|---|---|---|---|---|---|---|---|---|
| Median $/hr | Trend · 7d | n | Median $/hr | Trend · 7d | n | Median $/hr | Trend · 7d | n | |
| B200 | $11.30 | 4 | $6.04 | 4 | $4.89 | 5 | |||
| H200 | $7.91 | 5 | $4.54 | 4 | $3.44 | 7 | |||
| H100 | $6.88 | 5 | $3.29 | 4 | $2.72 | 10 | |||
| A100 | — | — | $2.33 | 8 | $1.26 | 3 | |||
The same chip can trade across all three tiers. An H100 SXM in a Tier 1 hyperscale deployment, a Tier 2 dedicated cluster, and a Tier 3 marketplace listing all carry identical silicon at fundamentally different rate bands — because the deployment is the product. A lower nominal rate at a downstream tier reflects a thinner capability bundle behind it.
Tier assignment is per-SKU and grade-based — driven by the capability bundle each SKU exposes, not by which operator hosts it. Fabric grade is scored bandwidth-first: a disclosed per-node bandwidth maps to top grade at ≥ 3 Tbps, mid grade at ≥ 400 Gbps, with the protocol name (IB-NDR / EFA / RoCE-v2 / Ethernet) used as fallback when bandwidth isn't disclosed. An interruptible SKU from any operator routes to T3IF; a dedicated-host SKU with top-grade fabric and a full enterprise bundle routes to T1IF; a dedicated-host SKU with mid-grade disclosed fabric routes to T2IF; everything else routes to T3IF. The same operator can therefore appear in two or three tiers across its SKU catalog, depending on what each SKU actually packages.