GPU MODELS / B300

B300 cloud GPU pricing

32 listings across 22 providers. Cheapest published rate: $4.480/GPU/hr at IO.NET. Sorted by per-GPU hourly price. Every row links to the provider's own published price.

WHAT IS THE B300

The B300 (Blackwell Ultra) is NVIDIA's inference-first Blackwell refresh: 288 GB of HBM3e - half again the B200's memory - with FP4 throughput tuned for reasoning models that generate long chains of thought. It targets the largest serving deployments; training fleets stay on B200-class parts.

MAKER
NVIDIA
ARCHITECTURE
Blackwell Ultra
LAUNCHED
2025
MEMORY
288 GB HBM3e
MEMORY BANDWIDTH
8 TB/s
TENSOR COMPUTE
~1.5x B200 tensor throughput (FP4-focused)
POWER DRAW
up to 1,400 W
INTERCONNECT
NVLink 5.0 1.8 TB/s

WHAT IT IS GOOD AT

Reasoning-model serving
Long-generation workloads (o1/R1-class) are what the FP4 and memory bumps are for.
Largest-model inference
288 GB per GPU serves the biggest open models without multi-node splits.

WHO SHOULD NOT RENT THE B300

Training fleets - it is an inference-first part.
Models under ~100B parameters - B200 or H200 serve them cheaper.
ProviderConfigVRAMLocation$/hr$/GPU/hrSource
IO.NETIO.NETB300 SXM × 1288 GBNot published$4.480$4.480receipt
Deep InfraDeep InfraB300 × 1270 GBNot published$4.890$4.890receipt
TensorPoolTensorPoolB300 SXM × 1288 GBNot published$5.490$5.490receipt
CyfutureCyfutureB300 × 1288 GBIndia$5.945$5.945receipt
Massed ComputeMassed ComputeB300 SXM6 × 8288 GBNot published$52.800$6.600receipt
RunPodRunPodB300 × 1288 GBGlobal$6.940$6.940receipt
ModalModalB300 × 1288 GBNot published$7.099$7.099receipt
HyperstackHyperstackB300 × 1288 GBUS / EU / CA$7.400$7.400receipt
UpCloudUpCloudB300 × 1288 GBNot published$7.601$7.601receipt
Yotta LabsYotta LabsB300 × 1262 GBNot published$7.640$7.640receipt
LyceumLyceumB300 × 1262 GBEurope$7.990$7.990receipt
Verda (formerly DataCrunch)Verda (formerly DataCrunch)B300 SXM6 × 4268 GBEurope (Finland / Iceland)$33.470$8.367receipt
Verda (formerly DataCrunch)Verda (formerly DataCrunch)B300 SXM6 × 8268 GBEurope (Finland / Iceland)$66.940$8.367receipt
Verda (formerly DataCrunch)Verda (formerly DataCrunch)B300 SXM6 × 1268 GBEurope (Finland / Iceland)$8.368$8.368receipt
Verda (formerly DataCrunch)Verda (formerly DataCrunch)B300 SXM6 × 2268 GBEurope (Finland / Iceland)$16.740$8.370receipt
falfalB300 × 1288 GBNot published$8.500$8.500receipt
Verda (formerly DataCrunch)Verda (formerly DataCrunch)B300 CC SXM6 × 4268 GBEurope (Finland / Iceland)$34.140$8.535receipt
Verda (formerly DataCrunch)Verda (formerly DataCrunch)B300 CC SXM6 × 8268 GBEurope (Finland / Iceland)$68.280$8.535receipt
Hostnot GPUHostnot GPUB300 × 1288 GBEU West$8.680$8.680receipt
Hostnot GPUHostnot GPUB300 × 1288 GBUS West$8.680$8.680receipt
Thai AI CloudThai AI CloudB300 × 8288 GBThailand$69.900$8.738receipt
Thai AI CloudThai AI CloudB300 × 1288 GBThailand$8.900$8.900receipt
Thai AI CloudThai AI CloudB300 × 2288 GBThailand$17.800$8.900receipt
Thai AI CloudThai AI CloudB300 × 4288 GBThailand$35.600$8.900receipt
NebiusNebiusHGX B300 × 1288 GBEU / US$9.500$9.500receipt
SpheronSpheronB300 × 1288 GBNot published$10.210$10.210receipt
GPUBrazilGPUBrazilB300 × 1288 GBGlobal$10.454$10.454receipt
GPUBrazilGPUBrazilB300 × 1269 GBEurope$10.725$10.725receipt
FireworksFireworksB300 × 1288 GBGlobal$15.000$15.000receipt
Oracle OCIOracle OCIB300 × 1263 GBNot published (global list)$15.000$15.000receipt
Latitude.shLatitude.shB300 × 8288 GBUnited States$128.000$16.000receipt
SelectelSelectelB300 × 1288 GBRussia$18.802$18.802receipt

THE B300 MARKET, BY THE NUMBERS

PROVIDERS
22
CONFIGURATIONS
32
MEDIAN $/GPU/HR
$8.54
SPREAD
4.2x

22 providers publish 32 B300 configurations on the panel today. Published per-GPU hourly rates run from $4.480 (IO.NET) to $18.80 (Selectel), a 4x spread between the cheapest and the most expensive published rate for the same chip. The median listing sits at $8.54/GPU/hr. 10 rows were reported in stock by the provider at last observation; the rest are published catalog rates. All rates are on-demand - committed-use and reserved pricing is excluded.

CHEAPEST B300 BY GEOGRAPHY
indiaCyfuture$5.945/hrunited statesHyperstack$7.400/hrcanadaHyperstack$7.400/hrfinlandVerda (formerly DataCrunch)$8.367/hricelandVerda (formerly DataCrunch)$8.367/hrthailandThai AI Cloud$8.738/hr
B300 PRICE TREND, DAILY PANEL SNAPSHOTS

Across every B300 row on the panel, the daily floor moved from $5.490 to $4.480/GPU/hr over 11 days of tracking (-18.4%). The median continuously-tracked listing moved from $7.65 to $8.37 (+9.4%). Snapshots run daily; the full history is public in the repo.

B300 PROVIDER NOTES
IO.NET · 1 config
Sets the panel floor for this model.
from $4.480/hr
Deep Infra · 1 config · 1 in stock
Within 15% of the panel floor.
from $4.890/hr
TensorPool · 1 config
Published catalog rates.
from $5.490/hr
Cyfuture · 1 config · 1 region · 1 in stock
1 of 1 configs reported in stock at last check.
from $5.945/hr
Massed Compute · 1 config
Published catalog rates.
from $6.600/hr
RunPod · 1 config · 1 region · 1 in stock
1 of 1 configs reported in stock at last check.
from $6.940/hr
Modal · 1 config
Published catalog rates.
from $7.099/hr
Hyperstack · 1 config · 1 region
Published catalog rates.
from $7.400/hr
BUYING B300: WHAT THE PANEL SAYS

At the current floor, one B300 running around the clock costs about $3,270 a month (730 hours of on-demand arithmetic). Node shape matters: the single-GPU floor is $4.480/GPU/hr against $6.600/GPU/hr on an 8-GPU node (+47.3%). Per-GPU floors by config size: 1x $4.48 · 2x $8.37 · 4x $8.37 · 8x $6.60. Compare the alternatives below before committing - the same budget often buys more than one chip class.

B300 VS THE ALTERNATIVES

B300 vs B200
Same bandwidth, 50% more memory, more FP4. Serving fleets with huge models take the B300; everyone else compares live rates.

Compare live rates: B300 vs B200 · B300 vs MI355X

B300 PRICING FAQ

What is the cheapest B300 cloud GPU right now?
The cheapest published B300 rate on the panel is $4.480 per GPU-hour at IO.NET (B300 SXM x 1, Not published). Every price links to the provider's own published rate.
How much does a B300 cost per hour?
Across 22 providers tracked today, published per-GPU hourly rates for B300 run from $4.48 to $18.80, with a median of $8.54. Committed-use and reserved rates are excluded; everything shown is on-demand.
How many providers rent B300 GPUs?
22 providers publish 32 distinct B300 configurations on Compute Cafe. 10 of those rows were reported in stock by the provider's live endpoint at last observation.
Where is B300 cheapest?
By published per-GPU hourly rate, the cheapest geography for B300 today is india at $5.95/hr (Cyfuture). The spread across 6 tracked geographies runs to $8.74/hr in thailand.
Are B300 prices going up or down?
Over the 11 days Compute Cafe has tracked the panel, the B300 floor is falling: $5.490 to $4.480/GPU/hr (-18.4%). The median continuously-tracked listing moved from $7.65 to $8.37 (+9.4%). Daily snapshots are public in the repo, so the trend is auditable.
What does a B300 cost per month?
At the current floor of $4.480/GPU/hr, a B300 running 24/7 costs about $3,270 per month (730 hours of on-demand arithmetic on the cheapest published rate). Committed-use contracts can price lower; the panel tracks on-demand rates only.
Is a full 8-GPU B300 node cheaper than a single card?
On today's panel: the single-GPU floor is $4.480/GPU/hr and the 8-GPU node floor is $6.600/GPU/hr - more expensive per GPU on a full node (+47.3%). Check the table for the exact configs behind each floor.
Who is the B300 for?
Providers serving the largest open models and reasoning workloads. Its 288 GB and FP4 focus pay off only at the top end of model size; smaller models are cheaper on B200 or H200.

See also alternatives to the B300 · cheapest GPU-hour per model · inference chips · training chips · by location