A clear way to evaluate AI compute infrastructure.

FlopSource is an independent directory and comparison site for AI compute infrastructure. It tracks GPU clouds, neoclouds, and bare-metal providers, verifies their specs and pricing, and lets buyers compare them side by side instead of piecing it together from thirty different marketing pages.

The directory covers 30 providers today, with data refreshed regularly: GPU types and counts, per-GPU-hour pricing, interconnect, cooling, jurisdiction, certifications, and availability. Comparison tools weight providers against your actual workload, because the right pick for training is rarely the right pick for inference.

Why it exists

Evaluating AI infrastructure is broken. Sticker prices hide interconnect quality, egress fees, and lead times. Real spec sheets sit behind sales calls. Procurement teams end up making six-figure decisions from vendor framing. FlopSource exists to make this category transparent and comparable, without another sales deck.

Who runs it

FlopSource is built and run by Tyler Babin, an infrastructure-minded builder who got tired of watching good teams pick the wrong provider for the wrong reasons. Provider data is gathered from public sources and verified on a rolling basis. Providers can request listings and corrections, and featured placements are always labeled as such. Rankings are never for sale.

Questions, corrections, or provider inquiries: hello@flopsource.com.