Resolves as YES if there is strong evidence that a blockchain‑ (or similar‑technology‑) based decentralized marketplace for buying and selling compute resources achieves “scaled usage” in any 30‑consecutive‑day window ending before January 1, 2031, by meeting at least 3 of the following 5 criteria:

1. Money (GMV):
   Buyer‑to‑provider payments settled by the marketplace’s smart contracts total ≥ $25,000,000 in that 30‑day window (USD‑equivalent at time of settlement). Only payments tied to completed/active compute jobs count; pure token swaps, grants, staking rewards, or emissions do not count.
   
   

2. Delivered compute (FLOP·s, pick one precision class):
   Aggregate delivered compute (summed across all jobs in the window) is ≥ 1,000 ZFLOP·s at FP16/BF16, or ≥ 2,000 ZFLOP·s at FP8.
   How to measure: For each job, use the vendor‑published peak TFLOP/s for the GPU model at the relevant precision class × job duration (seconds) × the job’s reported average utilization (if available; otherwise assume 100% only if the network provides no lower utilization metric). Sum across jobs. (If a network doesn’t expose per‑job precision, classify the job under FP16/BF16 for this calculation.)
   
   

3. Delivered compute (GPU‑hours alternative):
   Aggregate delivered H100‑equivalent FP16 GPU‑hours ≥ 500,000 in the window or A100‑80GB‑equivalent FP16 GPU‑hours ≥ 750,000.
   How to measure: Multiply job duration (hours) × number/type of GPUs × (device FP16 tensor‑throughput / reference device FP16 tensor‑throughput). The reference devices are NVIDIA H100 (FP16 tensor) and NVIDIA A100‑80GB (FP16/BF16 tensor).
   
   

4. Market breadth (participants):
   At least 2,500 distinct provider addresses each complete and settle ≥ 1 job, and at least 500 distinct buyer addresses each consume ≥ $100 of compute via the marketplace in the same window. (Addresses controlled by the same entity may be consolidated if the marketplace or credible analytics identifies them as one actor; exchanges and custodians don’t count as “buyers” unless they are the end consumer of compute.)
   
   

5. Energy delivered (optional physical corroboration):
   Compute delivered under marketplace jobs consumes ≥ 150 MWh of electrical energy in the window.
   How to measure: For each job, estimate energy as Σ(GPU TDP in kW × GPU‑hours) + documented CPU/overhead if available; otherwise GPU TDP‑only is acceptable.
   
   

Gating conditions (must all be true for any of the above to count):

• The system is permissionless or broadly open to independent providers and buyers (no single company’s captive fleet only).

• Payments settle on a public ledger (L1/L2/DAG or equivalent) using programmatic contracts/escrow tied to job IDs; off‑chain fiat invoices don’t count.

• Evidence for GMV/compute/breadth/energy is publicly auditable (e.g., marketplace explorer, contract logs, reputable analytics/dashboards, or provider job receipts that can be matched to on‑chain settlement).

• One marketplace/protocol must meet the thresholds on its own (it may span multiple chains under the same canonical protocol); you cannot sum activity across unrelated protocols.
  
  

Exclusions / clarifications:

• A centralized cloud that merely accepts crypto is not sufficient.

• “Available capacity,” “devices onboarded,” or idle GPU counts do not count toward compute thresholds without delivered jobs.

• Payments to intermediaries count only if they are the marketplace’s official settlement contracts distributing funds to providers (i.e., buyer → protocol escrow/settlement → provider).

• For currency conversion, use the settlement timestamp price or a volume‑weighted on‑chain oracle if the marketplace publishes one.

• If a job’s precision class is unknown, it may only be counted under the FP16/BF16 class in Criterion (2).
  
  

Resolution notes:
The first verified 30‑day window meeting the rule resolves YES. If no window before Jan 1, 2031 satisfies it, resolves NO. If evidence is conflicting, the burden is on YES to show auditable links between jobs and on‑chain settlements that satisfy the criteria above.