—
Bubbles mark headquarters countries; switch to DC cities to see where providers actually run data centers. Click any country or city to filter every tab.
Every published offer, converted to US dollars per GPU-hour at the lowest rate it lists (spot, reserved or on-demand). "Contact sales" quotes are left out.
Each dot is one offer. The bar spans the middle half, the dark tick is the median. Log scale. Click a GPU to focus on it.
Lowest published rate per provider for the focus GPU, with its pricing terms (PAYG on-demand, commitment, spot or serverless). Click a row for the profile.
| Provider | Pricing | $/GPU-hr |
|---|
Median of each provider's lowest rate; n = providers.
Capacity in megawatts as disclosed in public sources. Fewer than a third of providers publish a figure, so totals are a floor, not a market size.
Live capacity today and where each provider says it is heading. Log scale. Click a provider for its profile.
Sum of disclosed MW. Planned figures can overlap with live capacity.
Providers per business model, split by whether they own the facilities they run in.
GPU families listed by each provider, grouped by vendor and generation. Click a GPU to make it the focus GPU across the dashboard.
Each provider counted once, by the newest architecture it lists.
Share of providers headquartered in each region that list the GPU.
Every provider that matches the filters. Sort by any column; click a row for the full profile.