Measured Rates · U.S. & Europe 2026-11-17 print · published 2026-11-18 01:55 CT
Daily Market Rates
GPU instances you rent and run yourself (virtual machines,
bare-metal servers and marketplace instances), priced per
GPU-hour. Guaranteed and interruptible on-demand rates per chip,
across the three tiers. Every cell is a documented
series with a public method.
- → /term — committed tenors
- → /explorer — regional breakouts, every grain
- → Containers — managed platforms, billed by the second
On-demand, guaranteed · U.S. & Europe USD/GPU-hr · daily · 2026-11-17 print
| Silicon | Hyperscaler | Neocloud | Marketplace |
B300 | — | 8.97 | 10.63 |
| 14.24 | 8.50 | 7.87 |
| 10.30 | 4.85 | 4.46 |
| 11.25 | 3.37 | 2.34 |
| 4.45 | 2.04 | 1.21 |
Per-chip pages → On-demand, interruptible · U.S. & Europe USD/GPU-hr · daily · 2026-11-17 print
| Silicon | Hyperscaler | Neocloud | Marketplace |
B300 | — | 6.17 | — |
| — | 4.05 | — |
| 6.68 | 2.60 | — |
| 2.79 | 2.15 | — |
| 2.71 | 1.20 | — |
The global hyperscale clouds — AWS, Azure, Google Cloud, Oracle. Deep managed capacity, enterprise contracts, and the widest service surface. Guaranteed lists sit at the top of the market; their interruptible grade prices the same silicon’s spare capacity down into the neocloud range, repricing against available capacity while the lists hold still.
The dedicated GPU specialists — CoreWeave, Lambda, Nebius and peers. They sell capacity under their own service commitment. Single-tenant clusters built for training and inference, priced below the hyperscaler lists. Where these venues post an interruptible tier, it is typically a fixed discount off their own card.
Marketplaces such as Clore.ai and Lium, where independent hosts list capacity with no service commitment from the venue. A marketplace company’s own service-backed capacity sits with the neoclouds. Self-serve capacity that reprices continuously, where posted cards elsewhere move rarely. Both grades trade here and the interruptible legs float. This page’s ladder compares guaranteed rates; the interruptible grade is published as its own ladder on the home page and in the explorer.
| Attribute | Hyperscaler | Neocloud | Marketplace |
| Footprint | Managed multi-service hyperscale cloud | Dedicated single-tenant GPU clusters | Independent hosts on a marketplace |
| Networking | Hyperscale backbone, managed networking | Dedicated cluster fabric, often InfiniBand | Variable — Ethernet or undisclosed |
| Services | Full enterprise services, support, compliance | Focused GPU stack, lighter wrap | Self-serve, minimal support |
| Examples in this tier | e.g. AWS · Azure · Google Cloud · Oracle | e.g. CoreWeave · Lambda · Nebius | e.g. Clore.ai · Lium |
| Typical use | Enterprise & regulated workloads | Dedicated training & inference | Cost-sensitive batch, dev/test, spot |
The tier follows the capacity, not the chip or the
company. The same chip is priced in all three tiers, so the ladder
compares one chip's hyperscaler, neocloud and marketplace rates side
by side.
| Silicon | USD/GPU-hr |
| | 5.49 |
| | 5.49 |
| | 4.09 |
| L40S | 2.76 |
Suggested citation Compute Credit Index Research, Daily GPU Rental Rates, https://ccir.io/rates, retrieved 2026-11-17.