Published cloud GPU rates across 5 tracked providers with valid current listings, normalized to USD per GPU per hour. Index-eligible on-demand rows are separated from reservation and marketplace rows.
As of 2026-08-10, the B200 index median is $6.35/GPU/hr across 2 index-eligible providers, 5% below the highest monthly index point in the tracked history. The current page tracks 25 valid listings across 5 providers.
B200 pricing is still early and capacity can be uneven, so provider coverage matters more than a single headline quote.
Shadeform · on demand
Provider sources do not publish availability
current maximum ÷ minimum; billing modes remain labeled
latest two recorded monthly median points
| Architecture | NVIDIA Blackwell |
| Memory | up to 192GB HBM3e |
| Form factor tracked | SXM |
| Common workloads | frontier training, large-scale inference, new Blackwell deployments |
| Provider | Region | Billing | Total $/GPU/hr | Quality | Updated | |
|---|---|---|---|---|---|---|
S Shadeform marketplace | US, Dallas, TX | on-demand | $3.74/hr | Community | ● …Not published | ▶ |
S Shadeform marketplace | US, Kansas City, MO | on-demand | $4.51/hr | Community | ● …Not published | ▶ |
R RunPod marketplace | — | on-demand | $5.98/hr | Community | ● …Not published | ▶ |
H Hyperstack neocloud | — | on-demand | $6.00/hr | Reliable | ● …Not published | ▶ |
S Shadeform marketplace | FI, Central | on-demand | $6.31/hr | Community | ● …Not published | ▶ |
S Shadeform marketplace | FI, Central | on-demand | $6.42/hr | Community | ● …Not published | ▶ |
S Shadeform marketplace | FI, Central | on-demand | $6.52/hr | Community | ● …Not published | ▶ |
S Shadeform marketplace | FI, Central | on-demand | $6.52/hr | Community | ● …Not published | ▶ |
L Lambda neocloud | US | on-demand | $6.69/hr | Reliable | ● …Not published | ▶ |
S Shadeform marketplace | AU, Sydney | on-demand | $6.69/hr | Community | ● …Not published | ▶ |
L Lambda neocloud | US | on-demand | $6.79/hr | Reliable | ● …Not published | ▶ |
R RunPod marketplace | — | on-demand | $6.79/hr | Community | ● …Not published | ▶ |
S Shadeform marketplace | AU, Sydney | on-demand | $6.79/hr | Community | ● …Not published | ▶ |
L Lambda neocloud | US | on-demand | $6.89/hr | Reliable | ● …Not published | ▶ |
S Shadeform marketplace | AU, Sydney | on-demand | $6.89/hr | Community | ● …Not published | ▶ |
L Lambda neocloud | US | on-demand | $6.99/hr | Reliable | ● …Not published | ▶ |
S Shadeform marketplace | AU, Sydney | on-demand | $6.99/hr | Community | ● …Not published | ▶ |
AC AWS Capacity Blocks hyperscaler | US East (Dallas) Local Zone | reservation / capacity block | $10.58/hr | Experimental | ● …Not published | ▶ |
AC AWS Capacity Blocks hyperscaler | US East (Dallas) Local Zone | reservation / capacity block | $10.58/hr | Experimental | ● …Not published | ▶ |
AC AWS Capacity Blocks hyperscaler | — | reservation / capacity block | $12.36/hr4 listings | Experimental | ● …Not published | ▶ |
AC AWS Capacity Blocks hyperscaler | — | reservation / capacity block | $12.87/hr2 listings | Experimental | ● …Not published | ▶ |
These providers publish the GPU resource charge, while CPU and memory are billed separately. No total is estimated, and these rows never enter the on-demand index median.
| Provider | Billing | Published GPU component | Total $/GPU/hr | Source |
|---|---|---|---|---|
Modal serverless | serverless | $6.25/GPU/hrUSD 0.001736/second × 3600 | not published | modal.com ↗ |
These are published rates, not transaction prices. Large buyers routinely pay materially less through committed contracts, volume discounts, and reserved capacity. gpuprice.fyi is built for developers comparing rental options, not for settlement-grade transaction indices. Read full methodology →
As of 2026-08-10, the median on-demand B200 price is $6.35/GPU/hr across 2 index-eligible providers, per the gpuprice.fyi index.
As of 2026-08-10, published B200 cloud GPU listings range from $3.74-$12.87/GPU/hr across 25 valid listings.
The B200 index normalizes each listing to USD per GPU per hour, selects each provider's cheapest current eligible on-demand row, and calculates the median across providers.
The B200 page reports published list prices, not negotiated transaction prices; reservation, marketplace, and non-comparable billing modes are labeled separately from the core on-demand median.
B200 is tracked as SXM in gpuprice.fyi; listings with unclear form factor are not forced into SXM or PCIe medians.
Cite the B200 page using the citation block near the bottom of the page, including the access date 2026-08-10, methodology version, and snapshot link.
Source: gpuprice.fyi B200 Cloud GPU Pricing Index, latest price snapshot, data as of 2026-08-10, accessed 2026-08-10.↓ Download CSV