All GPU pricing
Back to the hub for per-chip bands.
GPU Pricing
Current NVIDIA B200 price per hour, sourced Blackwell list prices, H100-equivalent rates, and dated public price-page links.
B200 on-demand capacity is sourced from 3 provider price pages, each linked and dated below. Much of the B200 market still sells through reserved and contact-sales terms that never publish a price — this table is the public on-demand slice, not the whole market.
Lowest to highest sourced public list price, as of Sep 17, 2026
Public price pages with a B200 on-demand rate that cleared editorial review
The Hopper-generation read the Blackwell premium is usually priced against
Sourced rates
3 sourced provider rows, lowest list price first. Every row links the public price page it was observed on.
| Provider | Region | List $/GPU-hr | H100e $/hr | Availability | No change in | Source |
|---|---|---|---|---|---|---|
| Daytona | Multi-region | $6.25 | $3.13 | Available | 0 days | Open price page |
| Lambda | Multi-region | $6.99 | $3.50 | Available | 97 days | Open price page |
| CoreWeave | us | $8.60 | $4.30 | Available | 99 days | Open price page |
Public on-demand list prices from provider price pages — not negotiated quotes, reserved terms, or spot rates. Each row carries its own source link and price date. How we source, label, and date evidence: methodology.
See every reviewed provider rowHow to read it
B200 is the Blackwell generation — 180 GB of HBM3e and a large step up from Hopper. But new-generation capacity is allocated before it is listed: several providers sell B200 only through reserved or contact-sales terms, so a missing provider here usually means an unpublished price, not missing capacity.
Next
Back to the hub for per-chip bands.
The Hopper memory-upgrade read: sourced H200 on-demand rates.
How the generations compare on performance per dollar, memory, and availability.
Turn any rate in the table into a monthly estimate with utilization and overhead.