Spot, one snapshot
B300 spot prices per GPU-hour
6 spot rows from 5 providers in our data. Spot capacity can be reclaimed and its price can move, so these rows are never mixed into the on-demand ranking. Read 12 November 2026.
- Spot rows
- 6
- Lowest, provider-set
- $4.48
- Median, provider-set
- $4.99
- Lowest on-demand
- $6.60
5 providers
per GPU-hour
5 rows
Massed Compute
What counts as spot here
Definitions- Spot
- Capacity sold at a price that can change, which the provider may reclaim. The provider’s own word is preemptible or interruptible. Terms differ and are recorded per row below.
- Per GPU-hour
- Every figure is per GPU per hour. For an 8-GPU row the page prices the node and we divide by 8, or the page already gives a per-GPU rate.
- Provider-set
- The provider sets the price. A figure flagged as marketplace is a host-set or "from" price: shown, labelled, and left out of the spot median.
- On-demand reference
- The lowest provider-set on-demand B300 price is $6.60 at Massed Compute; the median is $9.50. They are drawn as dashed lines on the chart.
Spot against on-demand
Bars are spot rows, lines are on-demandEvery spot row
With the comparison arithmetic| Provider | Config | Region | Spot USD per GPU-hr | Same provider, on-demand | Spot vs it | Spot vs lowest on-demand | Read | Source |
|---|---|---|---|---|---|---|---|---|
| Nebiusfrom price | 8x HGX | not stated | $0.99 | $9.50 | -90% | -85% | 2026-11-12 | source |
| CoreWeave | 8x HGX | North America | $4.48 | no row at this GPU count | not compared | -32% | 2026-11-12 | source |
| CoreWeave | 8x HGX | Europe | $4.59 | no row at this GPU count | not compared | -30% | 2026-11-12 | source |
| Together AI | 8x HGX | not stated | $4.99 | $9.99 | -50% | -24% | 2026-11-12 | source |
| Verda | 1x SXM | not stated | $6.03 | $12.05 | -50% | -8.7% | 2026-11-12 | source |
| DigitalOcean (Paperspace) | 8x HGX | not stated | $6.17 | no row at this GPU count | not compared | -6.5% | 2026-11-12 | source |
What each page says
Notes as recorded in the data- Nebius, 8x HGX: Page column 'Preemptible, GPU-hour' shown as a 'from' price; page says spot prices are dynamic and can change as often as every 15 minutes.
- CoreWeave, 8x HGX, North America: Derived: $35.84 / 8 GPUs = $4.48 per GPU per hour. Page states one hourly price for the whole 8-GPU instance.
- CoreWeave, 8x HGX, Europe: Derived: $36.70 / 8 GPUs = $4.5875 per GPU per hour. Page states one hourly price for the whole 8-GPU instance.
- Together AI, 8x HGX: Together GPU Clusters; all prices per GPU per hour (page). 8-GPU count inferred from the HGX baseboard. Column 'Preemptible Compute, Pay as you go'.
- Verda, 1x SXM: Spot price column for the same 1x GPU instance row.
- DigitalOcean (Paperspace), 8x HGX: Current spot price per GPU per hour (interruptible, reclaimable with about 2 hours' notice; price locked at creation). Spot size is an 8-GPU droplet per the page's link slug. Page states new pricing is effective as of August 1, 2026.
6 rows, 5 providers, 1 marketplace or "from" figure. With this few rows, one provider changing its price changes every spot statistic on this page.
Questions
Computed from the rowsQ1How many B300 spot prices are there?
6 rows from 5 providers: Nebius, CoreWeave, Together AI, Verda, and DigitalOcean (Paperspace). 5 are 8-GPU and 1 is 1-GPU. That is few, so no spot statistic here is robust.
Q2What is the lowest B300 spot price?
Among provider-set spot rows, $4.48 per GPU-hour at CoreWeave. The median of those 5 rows is $4.99 and the highest is $6.17. Nebius at $0.99 is shown as a "from" figure and left out of those numbers.
Q3How does spot compare with on-demand at the same provider?
Nebius spot $0.99 against on-demand $9.50 (-90%), Together AI spot $4.99 against on-demand $9.99 (-50%), and Verda spot $6.03 against on-demand $12.05 (-50%). Only providers that list both at the same GPU count can be compared this way.
Q4How does spot compare with the on-demand market?
The median provider-set spot row is $4.99, $4.51 (47%) below the on-demand median of $9.50. The lowest on-demand price is $6.60 at Massed Compute. Spot prices can change after the day we read them, so the figures describe this snapshot.
Q5Does Runcrate list B300 spot?
Not in our data: Runcrate has no B300 spot row, so it does not appear on this page.
Also on this desk
- Who lists a B300 pricePublic price, marketplace, announced, spot only, or B200 only
- B300 against B200Premium per provider, same GPU count
- Blackwell price ladderB300, B200, H200 and H100 on one axis
- How thin the market isSpread, and the median with providers removed
- B300 reserved rowThe reserved price and what it sits against
- 8-GPU node costPer hour, day and 720-hour month
Disclosure
B300 Watch is published by the Runcrate team, who also run Runcrate, an AI cloud. Runcrate has no B300 spot row in this data, so it does not appear on this page. Other providers' prices are read from their own public pricing pages, read 12 November 2026. We have no paid placement.