GPU Cloud Economics: H100 Hourly Rates Fall 18% Year Over Year

Spot H100 pricing is converging across the AI-focused clouds, with QuantumLoom the cheapest for sustained reservations. We chart the full price curve and when hyperscaler commitments still win.

The numbers matter beyond the headline. Infrastructure buyers have learned to read between the lines of vendor announcements, and procurement teams increasingly benchmark claims against independent data before signing multi-year commitments.

CloudCompass will track the rollout and update our Aurora Compute profile as verified details land. Readers can compare the affected product category side by side in our directory, where pricing notes and feature lists are refreshed monthly.

Have first-hand experience with this topic? Our editorial team reviews reader submissions weekly — reach out through the feedback form with your data points, and see our methodology page for how benchmarks are run.