A two-layer asset

COMPUTE-PULSEbuyer room · privateManage buyer room

A two-layer asset. (1) Observation-level GPU price and capacity catalog: listings in USD per GPU-hour each retaining collection method, fetch time, optional upstream update time and source URL, filterable by GPU, provider, region, billing type, observed availability, price, minimum capacity and interconnect — spanning providers such as GMI Cloud, fal.ai (serverless GPUs at $4.00-$12.99/hr across 4 models) and Beam. (2) Specification-fixed benchmark indices for H100, H200, B200 and A100, each pinned to memory size, 8-GPU offer size, provider segment, geography and on-demand term, with the explicit rule that different model configurations are never normalised together; thin contracts return a 'building' state with a null value rather than an imputed number, and a dated SemiAnalysis public composite is returned only as an attributed external cross-check that never rescales the house value. Surrounding both is an active two-sided capacity marketplace: supply listings with selling and buying intent, RFQ endpoints, provider capacity submissions held unpublished until moderation, and double-opt-in subscriptions.