GPU.ai

GPU.ai — First-party hourly GPU supply-and-price panel generated by operating a live compute marketplace: the true floor price per GPU-hour per model across 12+ aggregated clouds, split by supply tier (Secure = vetted data-centre capacity vs Community = independent marketplace suppliers), and paired with actual launchable inventory counts, largest real multi-GPU node in stock, and per-GPU rate on that node — plus an external 'market range' band set against the platform's own floor. This is transaction-adjacent ground truth (real deployable nodes), not scraped price-sheet data.. Small operator…

GPU.ai offers (Alternative, Price, Supply Chain) — 744 GPUs launchable at the 2026-10-02 snapshot across 28 models and 12+ clouds, with 7 published fields per model row (VRAM, Secure rate, Community rate, largest node, node per-GPU rate, stock count, market range); refreshed hourly, so ~24 snapshots/day and roughly 480K model-observations per year if the full hourly series were retained.

Ground-truth GPU supply monitoring — launchable unit counts per model is a direct read on accelerator availability that scraped price sheets cannot provide; early detection of supply tightness or glut by model generation as an AI-capex leading indicator; measuring marketplace disintermediation (the floor-vs-market-range gap quantifies how much legacy clouds overcharge vs the spot floor); cross-tier arbitrage between vetted and community supply; inference cost-floor modelling; and calibrating GPU residual-value or utilisation models against real deployable inventory.

Coverage spans US; IT Consulting & Other Services, Semiconductors; alternative, price, supply_chain; equities.

The supplier's own access route is a contact form. Request a sample through Brickroad to start.