B300 on demand, one snapshot read 12 November 2026
Only 11 providers list a public on-demand B300 price
The B300 is new hardware and the market for it is early and thin. 19 providers list an H100 in the same data. With 11 prices a median is fragile: removing one provider moves it by up to $0.25. Everything below is one reading of the public price pages, with the source and the date on each row.
- Providers
- 11
- Offers
- 21
- Lowest
- $6.60
- Median
- $9.50
- Highest
- $17.80
with a public on-demand price
14 on-demand, 6 spot, 1 reserved
Massed Compute, per GPU-hour
of each provider’s lowest price
Amazon Web Services
Read this first
Only 11 providers publish a B300 on-demand price we can read and rank. There are 21 B300 rows in all: 11 provider-set on-demand rows in effect, 2 marketplace figures, 1 announced for later, 6 spot and 1 reserved. Spot and reserved rows never enter the on-demand numbers.
The market is early and thin. The lowest price is $6.60 and the highest $17.80, a ratio of 2.70 to 1. Drop the lowest-priced provider and the median becomes $9.75; drop the highest and it becomes $9.25. The numbers behind that.
Runcrate's live B300 price is $17.02 per GPU-hour for a 1-GPU listing: rank 10 of 11 from the lowest. We run Runcrate, and it is sorted by price like every other row.
Who lists what
Coverage matrixOne row per provider with any B300 row in our data. A cell holds a price from the data or says plainly that no public price is held. The last column places each on-demand price on the B300 range. Who lists a B300 price adds the providers that list a B200 but no B300.
| Provider | B300 on-demand | B300 spot | B300 reserved | B200 on-demand | On the B300 range |
|---|---|---|---|---|---|
| Massed Compute | $6.608x SXM | no public price | no public price | $5.438x SXM | |
| Hyperstack | $7.401x | no public price | no public price | $6.001x | |
| Civo | $7.498x | no public price | no public price | no public price | |
| Crusoe | $8.968x HGX | no public price | no public price | $8.508x HGXannounced, from 2026-12-01 | |
| RunPod | $8.991x | no public price | no public price | $7.991x | |
| Nebius | $9.508x HGX | $0.998x HGXfrom price | no public price | $8.508x HGX | |
| Together AI | $9.998x HGX | $4.998x HGX | no public price | $8.198x HGX | |
| Verda | $12.051x SXM | $6.031x SXM | no public price | $9.231x SXM | |
| Oracle Cloud | $15.001x | no public price | no public price | $14.001x | |
| Runcrateours, live | $17.021x | no public price | no public price | no public price | |
| Amazon Web Services | $17.808x | no public price | no public price | $14.248x | |
| CoreWeave | no public price | $4.488x HGX | no public price | $8.608x HGX | |
| DigitalOcean (Paperspace) | no public price | $6.178x HGX | $7.941x HGX | no public price | |
| Vast.ai | $9.941xmarketplace figure | no public price | no public price | $7.501xmarketplace figure |
provider-set pricemarketplace figure or from priceRange runs from $6.60 to $17.80 per GPU-hour. Spot and reserved columns show each provider's lowest B300 row of that type; they are never mixed into on-demand statistics.
B300 against B200, provider by provider
Same provider, same GPU count8 providers list both GPUs on demand at the same GPU count. The median B300 premium over the same provider's B200 is +22%; the largest is +31% at Verda. Percent figures are B300 over B200, on each provider's own prices.
Every B300 price in effect
13 rows, lowest firstRank counts provider-set rows only. A provider with two sizes appears twice. Marketplace figures are set by hosts, not by the provider, so they are shown with a label and left out of the rank and the median. Of the provider-set rows, 5 are 1-GPU and 6 are 8-GPU; the median 1-GPU row is $12.05 and the median 8-GPU row is $9.23.
| # | Provider | Config | USD per GPU-hr | Vs median | Billing | Region | Read | Source |
|---|---|---|---|---|---|---|---|---|
| 1 | Massed Compute | 8x SXMB300 SXM6 | $6.60 | -31% | not stated | not stated | 2026-11-12 | source for Massed Compute |
| 2 | Hyperstack | 1xNVIDIA B300 | $7.40 | -22% | per minute | not stated | 2026-11-12 | source for Hyperstack |
| 3 | Civo | 8xExtra Large 8 x NVIDIA B300 | $7.49 | -21% | per hour | not stated | 2026-11-12 | source for Civo |
| 4 | Crusoe | 8x HGXNVIDIA B300 HGX | $8.96 | -5.7% | per hour | not stated | 2026-11-12 | source for Crusoe |
| 5 | RunPod | 1xSecure Cloud B300 | $8.99 | -5.4% | per second | not stated | 2026-11-12 | source for RunPod |
| 6 | Nebius | 8x HGXNVIDIA HGX B300 | $9.50 | 0% | not stated | not stated | 2026-11-12 | source for Nebius |
| Vast.aimarketplace figure | 1xMarketplace lowest listed (B300) | $9.94 | not compared | not stated | not stated | 2026-11-12 | source for Vast.ai | |
| 7 | Together AI | 8x HGXNVIDIA HGX B300 cluster on-demand | $9.99 | +5.2% | not stated | not stated | 2026-11-12 | source for Together AI |
| Vast.aimarketplace figure | 1xMarketplace median (B300) | $10.00 | not compared | not stated | not stated | 2026-11-12 | source for Vast.ai | |
| 8 | Verda | 1x SXM1x B300 SXM6 268GB (30 vCPU, 255 GB RAM) | $12.05 | +27% | not stated | not stated | 2026-11-12 | source for Verda |
| 9 | Oracle Cloud | 1xOracle Cloud Infrastructure - Compute - GPU - B300 (B112237) | $15.00 | +58% | not stated | not stated | 2026-11-12 | source for Oracle Cloud |
| 10 | Runcrateours, live | 1x | $17.02 | +79% | per second | Boston, MA, United States | 2026-11-12 | source for Runcrate |
| 11 | Amazon Web Services | 8xp6-b300.48xlarge | $17.80 | +87% | per second | us-east-1 | 2026-11-12 | source for Amazon Web Services |
Announced, not yet in effect
| Provider | Config | USD per GPU-hr | Billing | Region | Read | Source |
|---|---|---|---|---|---|---|
| Crusoeannounced, from 2026-12-01 | 8x HGXNVIDIA B300 HGX (new on-demand from 2026-12-01) | $9.90 | per hour | not stated | 2026-11-12 | source for Crusoe |
These carry a future start date on the provider's page. They are left out of every statistic until they take effect.
Median of 11 provider prices: $9.50. Highest to lowest, $17.80 against $6.60. The lowest-priced provider here is Massed Compute; the lowest 8-GPU row is Massed Compute.
Questions about a thin market
Computed from the rows aboveQ1How many providers list a B300 price?
Q2What is the lowest listed B300 price?
Q3Why is the median fragile?
Q4Is Runcrate the highest-priced listing?
Q5How much more is a B300 than a B200?
Q6Why do 1-GPU and 8-GPU rows both appear?
Q7How current is this?
Provider desks
- Massed Compute$6.60 per GPU-hour, 8-GPU
- Hyperstack$7.40 per GPU-hour, 1-GPU
- Civo$7.49 per GPU-hour, 8-GPU
- Crusoe$8.96 per GPU-hour, 8-GPU
- RunPod$8.99 per GPU-hour, 1-GPU
- Nebius$9.50 per GPU-hour, 8-GPU
- Together AI$9.99 per GPU-hour, 8-GPU
- Verda$12.05 per GPU-hour, 1-GPU
- Oracle Cloud$15.00 per GPU-hour, 1-GPU
- Runcrate$17.02 per GPU-hour, 1-GPU
- Amazon Web Services$17.80 per GPU-hour, 8-GPU
- CoreWeaveno provider-set on-demand price
- DigitalOcean (Paperspace)no provider-set on-demand price
- Vast.aino provider-set on-demand price
Analysis
- 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 spot rowsEvery spot price we hold, labelled
- 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 is listed in the table for B300, marked "ours, live". It is sorted by price like every other row, with no pinned position. Other providers' prices are read from their own public pricing pages. We have no paid placement.
Providers considered: Massed Compute, Hyperstack, Civo, Crusoe, RunPod, Nebius, Together AI, Verda, Oracle Cloud, Runcrate, Amazon Web Services, CoreWeave, DigitalOcean (Paperspace), and Vast.ai. A provider missing from the list has no B300 row in our data, which is not a claim that it offers none.