B300 CLOUD PRICING

B300 rental pricing.
Compare the actual offer.

Review B300 rates with their configuration, billing unit and purchase terms. Use the provider directory for company profiles and the buying guide to evaluate the proposal.

Compare B300 offers ↓Explore B300 providers ↗

PUBLISHED OFFER SNAPSHOTS

B300 rates.
The terms travel with them.

Alphabetical by provider, not a ranking. Dates apply to individual records. Compare currencies, minimum allocations and operating models before comparing the rate.

Dated B300 offers. Billing units and commitments differ.
Provider	Published price and billing unit	Configuration and terms	Evidence
CoreWeaveGPU cloud	Request on-demand quoteproject quote	On-demand: contact sales HGX B300, eight GPUs. The provider separately publishes spot prices; those are not on-demand quotes. Confirm region, delivery date and cluster network.	Published · checked 2026-09-24Official source Capacity needs confirmation
HyperstackGPU cloud	$7.40USD · GPU-hour8-GPU allocation: $59.20/hour (calculated)	On-demand; billed per minute Published B300 flavor n3-B300-SXM6x8 has eight GPUs in CANADA-1. Confirm stock and the required software image before setting a launch date.	Published · checked 2026-09-24Official source Configuration source Capacity needs confirmation
NebiusGPU cloud	$7.85USD · GPU-hour	On-demand; rate changes October 1 HGX B300. Published rate rises to $9.50 per GPU-hour on October 1, 2026. Minimum allocation and region need confirmation; the per-GPU rate is not proof of a single-GPU purchase.	Published · checked 2026-09-24Official source Capacity needs confirmation
RunpodGPU cloud	$6.94USD · GPU-hour	On-demand Community Cloud B300 Pods, 1 to 8 GPUs; billed by the second. Confirm the host, location and persistent storage. Secure Cloud and multi-node Clusters are separate offers.	Published · checked 2026-09-24Official source Configuration source Capacity needs confirmation
RunpodGPU cloud	$7.89USD · GPU-hour	On-demand Secure Cloud B300 Pods, 1 to 8 GPUs; billed by the second. Confirm selected data center, support and storage scope. This Pod rate does not establish reserved multi-node capacity.	Published · checked 2026-09-24Official source Configuration source Capacity needs confirmation
ScalewayGPU cloud	From €7.50EUR · GPU-hour8-GPU allocation: from €60.00/hour (calculated)	Starting on-demand rate; billed per minute B300-SXM-8-288G: eight GPUs, documented in FR-PAR-2. EUR pricing, not USD. Confirm block storage, support and applicable tax.	Published · checked 2026-09-24Official source Configuration source Capacity needs confirmation
STNGPU cloud	$5.85USD · GPU-hour8-GPU allocation: $46.80/hour (calculated)	Reserved: three-year minimum HGX B300, eight GPUs per node. Published four-year rate $5.70 and five-year rate $5.45 per GPU-hour. Networking and egress included; storage separate. Confirm allocation and contract total.	Published · checked 2026-09-24Official source Capacity needs confirmation
Verda (formerly DataCrunch)GPU cloud	$8.20USD · GPU-hour	On-demand GPU instance 1x B300 SXM6 lists 268 GB VRAM, 30 CPUs and 255 GB RAM. Larger instances are offered; confirm the full configuration and location. Spot and serverless prices are separate.	Published · checked 2026-09-24Official source Configuration source Capacity needs confirmation

These are published snapshots, not live inventory or guaranteed quotes. Serverless GPU-only rates, marketplace listings and whole-instance prices cover different purchases. Confirm storage, transfer, support, tax and start-date terms.

WORK THROUGH IT WITH AN ADVISOR

Make the quote fit your workload.

Bring your workload, launch date and what could change. We can help you evaluate capacity, compare proposals and identify the deployment support you need.

Discuss my project ↗

START WITH THE RIGHT CONFIGURATION

B300 and GB300.
Specify what you need.

A similar name can describe a very different deployment. Put the GPU count, CPU architecture, memory and network in every request for pricing.

B300 GPU instances and nodes

B300 is a Blackwell Ultra GPU. NVIDIA's DGX B300 system combines eight GPUs with Intel Xeon CPUs; its specifications list 288 GB of GPU memory per GPU. A cloud provider's instance can expose a different usable memory figure and bundle of resources.

Our offers retain each provider's stated configuration. Verda lists 268 GB for its B300 instance, while other listings use 288 GB. Confirm usable memory in the proposed instance instead of treating those labels as interchangeable.

NVIDIA DGX B300 system specifications ↗

GB300 NVL72 systems

NVIDIA's GB300 NVL72 combines 72 Blackwell Ultra GPUs and 36 Arm-based Grace CPUs in a rack-scale system. That is a different architecture and network domain from an eight-GPU B300 node.

A provider may sell access to part of the system. Ask what you are renting, what is dedicated to you and whether your CPU software and containers support Arm. Use a separate quote for the requested allocation.

NVIDIA GB300 NVL72 specifications ↗ · Explore researched GB300 offers →

FROM GPU-HOUR TO PROJECT BUDGET

Price the test.
Then price the commitment.

These calculations use the dated offers above. They cover compute only; storage, support, transfer, taxes and failed or repeated jobs may add to the bill.

A single-GPU evaluation

At Runpod Secure Cloud's $7.89 per GPU-hour, a 100-hour B300 test is $789. Keeping one GPU running for an illustrative 730-hour month is $5,759.70. That calculation is not a monthly plan or a capacity reservation.

An eight-GPU allocation

Hyperstack publishes $7.40 per GPU-hour, but its documented B300 flavor contains eight GPUs. The allocation therefore costs $59.20 per hour, or $5,920.00 for 100 hours, before other charges. Code using only one GPU would not reduce the purchased VM size.

A multi-year reservation

STN's three-year rate of $5.85 per GPU-hour corresponds to $46.80 per hour for eight GPUs. Obtain the full contracted payment schedule, ramp dates and exit terms. Compare that obligation with your expected utilization and how much the project might change.

A quote spanning a price change

Nebius publishes $7.85 per GPU-hour now and $9.50 effective October 1, 2026. Ask for the rate that applies over your actual launch and operating window. A September snapshot cannot establish the cost of a later deployment.

CAPACITY THAT CAN GROW

One successful test
does not secure the next node.

Businesses rarely know their final GPU requirement at the start. Give each provider the same three scenarios and ask which capacity they can commit to in writing.

Launch

Specify the exact B300 allocation, region, start date and workload. Request proof that the proposed hardware is ready and available to your account, plus the acceptance test and remedies if delivery slips.

Growth

Ask how much capacity can be added, with how much notice and under what reservation. Confirm whether expansion preserves the network, storage access and data location, or requires a migration.

A demand surge

Separate committed capacity from best-effort bursts. Name an acceptable fallback region or GPU, checkpoint plan and recovery target. Include time to move data and validate performance on the fallback.

Deployment ownership

Agree who configures the cluster, moves data, deploys containers, monitors jobs and responds to failures. Managed infrastructure does not automatically include application deployment and operations.

Use the capacity and delivery checklist → · Download the provider quote brief ↓

TEST THE WORKLOAD, NOT JUST THE GPU

When does B300
earn its price?

Our evaluation approach: compare a representative job at the same output quality on the actual configuration you would rent.

Memory and serving behavior

Measure model memory, context length, batch size and peak cache use. For inference, compare latency and throughput at realistic concurrency, including quiet periods. More memory is useful when it resolves a measured constraint.

Software and startup time

Validate your container, framework, kernels and CPU dependencies before committing. Hyperstack currently specifies a particular Ubuntu, driver and CUDA image for B300 and warns that full boot can take more than an hour. Include provisioning time in your launch test.

Hyperstack B300 compatibility and boot guidance ↗

Communication and recovery

Test the interconnect and storage path with the intended node count. Then test checkpoint restore and a failed-node replacement. NVLink inside a node does not establish the network between nodes or your recovery time.

Cost per useful result

Compare the entire allocation cost, time to completion, retries and support effort against B200 and H200. Choose B300 when the measured benefit supports its full project cost.

B300 RENTAL QUESTIONS

Before you request
the final proposal.

Can I rent one B300?

Some offers support a single GPU, including the Runpod Pods and Verda instance examples above. Other documented configurations require eight GPUs. A published per-GPU rate alone does not establish the minimum purchase.

Is spot pricing a fair comparison?

Keep interruptible offers separate from on-demand and reserved capacity. Include restart time and lost work in the budget. The main table uses on-demand rates or clearly labeled reserved terms; CoreWeave's row requests an on-demand quote.

Does a published rate mean capacity is available?

No. Every offer here still needs confirmation for your region, GPU count and start date. Confirm whether a reservation guarantees capacity or only changes the price, and what happens when you need to expand.

Can GPU Cloud Advisors help with deployment?

We help you compare providers and proposals, clarify capacity and growth requirements, and identify the technical support your project needs. Bring your workload and launch date so we can work through the next steps together.

Discuss my B300 project →

Compare the provider around the hardware.

H100 pricing →H200 pricing →B200 pricing →B300 pricing →All providers →

WORK THROUGH IT WITH AN ADVISOR

Plan your B300 deployment with an advisor.

Bring your workload, launch date and what could change. We can help you evaluate capacity, compare proposals and identify the deployment support you need.

Discuss my project ↗

HOW TO READ THE EVIDENCE

Published facts.
Project-specific questions.

Sources and check dates travel with each record. This directory covers our researched examples, not every provider or GPU they sell.

Published

A linked provider source supports the stated offer or profile fact. It is a dated snapshot. It does not confirm stock, your start date or your access to capacity.

Provider-confirmed

This label is reserved for a dated, direct confirmation covering a specific allocation and scope. No offer in this edition has that status.

Needs confirmation

Current capacity, growth rights and the final commercial and support terms must be resolved for your project. An unverified item is not a negative provider rating.

Fit and buying questions are editorial assessments, not firsthand benchmarks or financial risk ratings. Customer examples are provider-published relationships, not endorsements or proof of a particular GPU deployment unless stated. Our advisory services are free to you; we are compensated by the provider you choose through us. Listing a provider does not establish a reseller relationship.

LET’S FIND YOUR WAY FORWARD

Build your shortlist.
Validate your next move.

Bring your workload, launch window and likely growth range. We can help you compare providers, clarify the evidence you need and plan the support required to move forward.

Our advisory services are free to you. We’re compensated by whichever provider you choose through us.

30-minute consultationNo obligation

YOUR GPU CLOUD CONSULTATION

Bring your questions.
We’ll bring perspective.

A few details help us make the most of your time:

What you’re building or running
Your GPU or memory needs, if known
Target location and launch timeline
Budget range or an existing proposal

Meet with an advisor

Book securely on Calendly. Times appear in your time zone.

Prefer to call? 844-506-2299