The market intelligence layer for AI compute
Flopdesk exists because the fastest-growing cost in technology has no transparent market. We are building the reference price for compute.
The problem
GPU pricing is opaque, fragmented, and changes daily. The same H100 can vary by more than 2x across providers and regions, reserved discounts swing with supply, and capacity appears and disappears without notice. Teams spending millions on compute are flying blind.
What we do
Flopdesk aggregates and normalizes on-demand and reserved GPU pricing, capacity, and availability across every major AI cloud, then tracks it over time. We turn a scattered set of pricing pages into a single, comparable, historical market feed.
Who it is for
AI startups managing burn, infrastructure investors and hedge funds modeling the compute market, data center operators benchmarking their rates, and enterprise AI teams negotiating capacity. If the price of compute matters to your decisions, it belongs on your desk.
Our principles
Normalize everything to a comparable unit. Show history, not just spot. Be explicit about sources and freshness. Stay vendor-neutral. Prices are indicative and aggregated from public sources, and are not investment advice.