B300 over B200
What a provider charges for a B300 over its own B200
8 providers list both GPUs on demand at the same GPU count. The B300 costs a median of 22% more than the same provider's B200, ranging from +7.1% to +31%. All prices are provider-set and in effect on the day read, 12 November 2026.
- Providers with both
- 8
- Median premium
- +22%
- B300 lowest
- $6.60
- B200 lowest
- $3.50
same provider, same GPU count
$1.28 per GPU-hour
Massed Compute
Vultr
Same provider, same GPU count
8 pairs, lowest B300 first| Provider | GPUs | B300 | B200 | Premium USD | Premium | Sources |
|---|---|---|---|---|---|---|
| Massed Compute | 8 | $6.60 | $5.43 | +$1.17 | +21% | B300, B200 |
| Hyperstack | 1 | $7.40 | $6.00 | +$1.40 | +23% | B300, B200 |
| RunPod | 1 | $8.99 | $7.99 | +$1.00 | +13% | B300, B200 |
| Nebius | 8 | $9.50 | $8.50 | +$1.00 | +12% | B300, B200 |
| Together AI | 8 | $9.99 | $8.19 | +$1.80 | +22% | B300, B200 |
| Verda | 1 | $12.05 | $9.23 | +$2.82 | +31% | B300, B200 |
| Oracle Cloud | 1 | $15.00 | $14.00 | +$1.00 | +7.1% | B300, B200 |
| Amazon Web Services | 8 | $17.80 | $14.24 | +$3.56 | +25% | B300, B200 |
Premium is the B300 price minus the B200 price, over the B200 price, each the provider's lowest provider-set on-demand row at that GPU count. Bar length is the premium relative to the largest in the table.
The pairs on one axis
B200 square, B300 dotAcross the whole market
Different providers, so not a like-for-like premium- Lowest listed
- B300 $6.60 at Massed Compute; B200 $3.50 at Vultr. The two lowest prices come from different providers.
- Median
- B300 $9.50 across 11 providers; B200 $8.34 across 10. The B300 median is $1.16 (14%) above the B200 median.
- Highest
- B300 $17.80 at Amazon Web Services; B200 $14.24 at Amazon Web Services.
- Per GB, median
- B300 $0.033 over 288 GB; B200 $0.046 over 180 GB, catalog sizes. The B300 is 29% lower than the B200 per GB.
What the maker says
Maker figures, with a caveat| Figure | B300 | B200 |
|---|---|---|
| Memory, maker maximum | 288 GB HBM3E | 192 GB HBM3E |
| Memory bandwidth | 8 TB/s | 8 TB/s |
| FP8 dense | 5,000 TFLOPS | 5,000 TFLOPS |
| BF16 dense | not stated | not stated |
| Maximum power | 1,400 W | 1,200 W |
| Release year | not stated | not stated |
Both columns come from one comparison table on the maker's developer blog, linked here for the B300 and the B200. The memory row is the chip's maximum, so shipping instances may carry less; the specs file records a datasheet figure of 279 GB for the B300. A cell that says not stated was not stated on the page, and we do not fill it. The table lists the same FP8 dense rate for both GPUs, so the price premium here is not explained by that figure.
Questions
Computed from the pairsQ1How much more does a B300 cost than a B200 at the same provider?
Q2Is the premium the same everywhere?
Q3Which providers list both?
Q4Why are some providers left out of the pairs?
Q5What do the maker figures say about memory and bandwidth?
Also on this desk
- Who lists a B300 pricePublic price, marketplace, announced, spot only, or B200 only
- 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.