NVIDIA B300 / BLACKWELL ULTRA / 288 GB
NVIDIA B300 cluster rental price: on-demand and reserved
Published on-demand rates for a B300 run from $6.60 to $15.00 per GPU-hour. Reserved terms cut that. The lowest published 36-month rate is $4.70. Most providers quote reserved B300 capacity privately.
Updated October 17, 2026. By GPU Rentals, powered by Bridgepointe Technologies. Every figure comes from the provider's own pricing page. Sources are listed at the end.
Per GPU-hour, 9 providers
Per GPU-hour, posted by 2 providers
Per GPU-hour, posted by 2 providers
Our B300 reserved cluster rates
We compare each term against the lowest AWS, Microsoft or Google rate. AWS is the only one of the three that publishes a B300 rate. Rates are per GPU-hour, billed monthly, with no upfront payment.
| Term | GPU Rentals rate per GPU-hour | Lowest AWS, Microsoft or Google rate |
|---|---|---|
| 12 months | $8.77 | $14.02, AWS 1-year Savings Plan |
| 24 months | $6.79 | Not offered |
| 36 months | $4.81 | $7.69, AWS 3-year Savings Plan |
| 60 months | $4.33 | Not offered |
AWS, Microsoft and Google sell only 1-year and 3-year terms, so 24 and 60 months have no direct comparison. Every rate depends on confirmed capacity for your GPU count, start date and region.
B300 on-demand price per GPU-hour
On-demand means you pay by the hour and can stop anytime. It is the most expensive way to run a B300 for months. Most providers rent it in whole 8-GPU servers.
| Provider | On-demand per GPU-hour | How it is sold |
|---|---|---|
| Massed Compute | $6.60 | 8-GPU instance at $52.80 per hour |
| Hyperstack | $7.40 | Per GPU |
| Civo | $7.49 | 8-GPU node |
| Latitude.sh | $8.00 | 8-GPU bare metal server at $64 per hour |
| Runpod | $8.99 | Secure Cloud |
| Nebius | $9.50 | Per GPU |
| Together AI | $9.99 | GPU cluster |
| Verda | $10.08 | Self-service instance |
| Oracle Cloud | $15.00 | BM.GPU.B300.8 bare metal |
Spot and preemptible rates are left out. They are cheaper, but the provider can take the GPUs back. AWS sells B300 as short-term Capacity Blocks at $16.15 per GPU-hour, prepaid for 1 to 182 days.
B300 reserved cluster pricing by term
A reserved term commits you to a set number of GPUs for months or years. In return the hourly rate drops. Longer terms drop it further.
| Provider | 12 months | 24 months | 36 months | Notes |
|---|---|---|---|---|
| Civo | $6.49 | $5.99 | $5.49 | 8-GPU node. On-demand is $7.49. |
| TheAI Cloud | Not offered | $5.00 | $4.70 | Whole 8-GPU nodes. 30% due at signing. |
| DigitalOcean | $7.94 | Not published | Not published | 12-month contract through sales. |
| Verda | $9.27 | $7.56 | Not published | Calculated, not posted: Verda publishes term discounts, not dollar rates. Its 8% and 25% discounts applied to its $10.08 on-demand rate. |
| Flexicloud | Not published | Not published | Not published | Posts only "from $4.50" on its longest term, with a 1-year minimum. |
Nebius, CoreWeave, Runpod, Together AI, Hyperstack, Latitude.sh, Massed Compute, Crusoe, Oracle and Vultr all rent B300 on reserved terms but quote the price privately. Nebius says reserved runs "up to 35% less" than on-demand. CoreWeave says committed use runs "up to 60%" less.
What a published rate does not tell you
- Whether the provider has free B300 capacity on your start date.
- The price for a large cluster. Big commitments are negotiated one deal at a time.
- Prepayment, cancellation and renewal terms.
- Networking between servers, storage and data transfer, which are often billed separately.
That is the part we handle. Tell us the GPU count, term and start date, and we check which providers can actually deliver it.
How B300 clusters are sized: 8, 64, 512 GPUs
A B300 is sold inside an 8-GPU server, so cluster sizes move in steps of 8. A request for 500 GPUs becomes 496 or 504 in a real quote.
Large training clusters grow in bigger blocks. NVIDIA's own B300 reference design is built from units of 64 servers, which is 512 GPUs. A buyer asking for about 500 GPUs is usually best served by one such unit.
Each B300 server draws about 14 kW. That matters if you lease servers into your own data center rather than renting them in the cloud.
The GB300 NVL72 is a different product. It is a liquid-cooled rack of 72 GPUs that act as one unit, and it draws up to about 140 kW. If you need it, ask for it by name.
Large clusters rarely come through self-serve checkout. New cloud accounts usually start with a quota of zero high-end GPUs, and big orders go through a sales contract. Lambda, for example, offers B300 only in clusters of 4,000 or more GPUs on contracts of three years or more.
GET A B300 CLUSTER QUOTE
Reserved B300 clusters.
Checked for you.
Tell us how many B300s you need, for how long, and when. We check providers and reply by email with options and pricing. Nothing is booked or charged.
REQUEST SENT
Thanks. We have your B300 request.
We will reply by email with provider options and pricing. Nothing has been booked or charged.
B300 QUESTIONS
Before you rent a B300.
What are GPU Rentals' B300 cluster rates?
Our reserved B300 rates are $8.77 per GPU-hour for 12 months, $6.79 for 24 months, $4.81 for 36 months and $4.33 for 60 months, billed monthly. Rates are subject to confirmed capacity.
How much does it cost to rent a B300?
Published on-demand rates run from $6.60 to $15.00 per GPU-hour. A full 8-GPU server at $8.00 per GPU-hour costs $64 per hour, or about $46,000 for a 30-day month.
Is a reserved B300 cluster cheaper than on-demand?
Yes. The lowest published rates are $6.49 for 12 months, $5.00 for 24 months and $4.70 for 36 months. Most providers quote reserved prices privately, so the best rate for your size and term comes from a quote.
How many B300s come in one server?
Eight. Providers rent B300 in whole 8-GPU servers, so cluster sizes go up in steps of 8.
Can I rent a 500-GPU B300 cluster?
Yes, through a reserved contract rather than self-serve checkout. A real quote rounds to 496 or 504 GPUs, or to 512, which is one unit of NVIDIA's reference cluster design.
How much does a 64-GPU B300 cluster cost per month?
At published on-demand rates of $6.60 to $15.00 per GPU-hour, 64 GPUs cost about $304,000 to $691,000 for a 30-day month. At the lowest published 36-month reserved rate of $4.70, the same cluster is about $217,000 a month.
How much does a 512-GPU B300 cluster cost per month?
At $6.60 to $15.00 per GPU-hour on demand, 512 GPUs cost about $2.43 million to $5.53 million for a 30-day month. At the lowest published 36-month reserved rate of $4.70, it is about $1.73 million a month. Clusters this size are negotiated, so a quote can come in lower.
What is the difference between B300 and GB300 NVL72?
B300 servers hold 8 GPUs each and connect to each other over the network. A GB300 NVL72 is a liquid-cooled rack of 72 GPUs that work as one unit. It suits the largest models and needs much more power per rack.
What should a reserved B300 quote include?
The GPU count, term, start date and region. It should also state networking between servers, storage, prepayment, cancellation and renewal terms. A low hourly rate can hide costs in those lines.
Sources, checked October 17, 2026
Civo: civo.com/pricing. DigitalOcean: digitalocean.com/pricing/gpu-droplets, effective August 1, 2026. Verda: verda.com/pricing. TheAI Cloud: cloud.theai.com. Flexicloud: flexicloud.co/gpu/nvidia-b300. AWS Capacity Blocks: aws.amazon.com/ec2/capacityblocks/pricing. Massed Compute: vm.massedcompute.com/pricing. Hyperstack: hyperstack.cloud/gpu-pricing. Latitude.sh: latitude.sh/pricing. Runpod: runpod.io/pricing, updated October 8, 2026. Nebius: nebius.com/prices. Together AI: together.ai/pricing. Oracle: oracle.com/cloud/price-list. CoreWeave: coreweave.com/pricing. Lambda: lambda.ai. NVIDIA DGX B300 SuperPOD reference design: docs.nvidia.com. DGX B300 power: nvidia.com.
Prices change often and are shown as published. A published price is not a promise that capacity is available.