GPU-hour pricing — B300 & RTX PRO 6000

Market rental rates by deployment, contract type and provider tier. Market comps only — no cost, margin or internal price.

Deployment
Contract
Tier

Headline rates

By contract type

Median $/GPU-hour across the provider set for the selected tier. Reserved and spot samples are thin — provider counts are on every bar's tooltip and in the table below.

Median $/GPU-hour by contract type

By provider tier

On-demand only. The tier a megawatt is sold into moves the price about as much as the chip does.

Median on-demand $/GPU-hour by provider tier

Price over time

Median on-demand $/GPU-hour, all tiers

B300 and RTX PRO 6000 in colour; other NVIDIA datacenter SKUs in grey for context. Aggregator listing medians and index composites — a consistent method within each line, not across lines.

Full grid

Every deployment × contract × tier combination in the current dataset. This is the table view for the charts above — every plotted value is here. Change columns show movement against the same measurement method, so a blank means no comparable earlier datapoint exists yet, not that the price was flat.

Provider detail

Method and limits

Pull fresh data

This page updates itself every Friday, 7:00am Pacific. Just reload it — the date badge at the top right tells you how current it is.

The page can't re-scrape on its own: the pricing aggregators block cross-origin requests and none of them publish an open API. A scheduled job re-reads the sources and republishes this file.

Need it sooner? If you own this dashboard, paste the line below into Cowork. Otherwise ask whoever does.

Refresh the GPU pricing dashboard now — re-read the providers and aggregators, append today's datapoint, re-render and republish.