B300 CLOUD PRICING

B300 rental pricing.
Compare the actual offer.

On demand, 6 providers we track list B300 from $6.60 to $9.99 per GPU-hour, checked September 24 to September 29, 2026. Compare each rate with its configuration, billing unit and purchase terms.

PUBLISHED OFFER SNAPSHOTS

B300 price per GPU-hour.
By provider, with terms.

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.
ProviderPublished price and billing unitConfiguration and termsEvidence
DigitalOceanGPU cloud$7.94USD · GPU-hourReserved: 12-month contractual commitment

HGX B300 GPU Droplets. Confirm minimum GPU allocation, region and total contract cost with sales; per-GPU pricing does not establish a single-GPU purchase.

Published · checked 2026-09-24Official source Capacity needs confirmation
DigitalOceanGPU cloud$8.00USD · GPU-hourSpot: interruptible capacity

HGX B300 GPU Droplets can be reclaimed. The provider aims for two hours of notice but may give less or none. Confirm allocation and region; plan checkpointing and recovery.

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
Latitude.shGPU cloudFrom $64.00USD · 8-GPU server-hourFrom $8.00 / GPU-hour · calculatedStarting hourly bare-metal rate

g4.b300.large: eight HGX B300 GPUs, 288 GB GPU memory each, advertised 800 Gbps dual-plane RoCE. From $64 per server-hour, equivalent to a calculated starting rate of $8 per GPU-hour. Confirm location, stock and final configuration.

Published · checked 2026-09-24Official source Capacity needs confirmation
Massed ComputeGPU cloud$52.80USD · 8-GPU instance-hour$6.60 / GPU-hour · calculatedOn-demand eight-GPU VM

B300 SXM6 x8, 280 vCPUs and 2,800 GB RAM. The $52.80 hourly price buys the full eight-GPU allocation; $6.60 per GPU-hour is calculated, not a single-GPU offer. Confirm region, storage and interconnect.

Published · checked 2026-09-24Official source Capacity needs confirmation
NebiusGPU cloud$7.85USD · GPU-hour$9.50 from October 1, 2026On-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 2026-09-28Official source Capacity needs confirmation
RunpodGPU cloud$6.94USD · GPU-hourOn-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-hourOn-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 cloudFrom €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
Together AIGPU cloud$9.99USD · GPU-hourOn-demand GPU clusters

HGX B300 cluster compute. Confirm minimum cluster size, region, storage and networking. Dedicated inference is a separate contact-sales product; reserved cluster pricing also requires a quote.

Published · checked 2026-09-24Official source Capacity needs confirmation
Together AIGPU cloud$4.99USD · GPU-hourPreemptible GPU clusters: interruptible

HGX B300 cluster compute. Confirm minimum allocation, region and interruption terms. Budget for checkpointing and restarted work; this is not a guaranteed continuous allocation.

Published · checked 2026-09-24Official source Capacity needs confirmation
Verda (formerly DataCrunch)GPU cloud$8.54USD · GPU-hourOn-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 2026-09-29Official 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 ↗

MORE B300 OPTIONS

B300 providers.
Request a project quote.

These providers publish B300 offerings or deployment documentation. A numeric B300 rate is not verified here. Access, allocation and contract terms differ; confirm your eligibility and launch capacity.

Dated B300 offers. Billing units and commitments differ.
ProviderPublished price and billing unitConfiguration and termsEvidence
CirrascaleGPU cloudRequest a quoteproject quoteEight-GPU cloud instance; monthly or longer terms

Lists eight NVIDIA B300 GPUs with dual 64-core processors and 3 TB RAM. Monthly, six-month and annual prices require a quote. Confirm region, networking and available node count.

Published · checked 2026-09-24Official source Capacity needs confirmation
CoreWeaveGPU cloudRequest on-demand quoteproject quoteOn-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
CrusoeGPU cloudRequest a quoteproject quoteB300 cloud deployment documented; request project pricing

The September 21 announcement documents HGX B300 training in Iceland and advertises trial runs. No numeric B300 rental rate was verified in the reviewed source. Deployment evidence does not establish spare inventory or a guaranteed start date.

Published · checked 2026-09-24Official source Capacity needs confirmation
CUDO ComputeGPU cloudRequest a quoteproject quoteDedicated enterprise infrastructure; quote on request

The provider lists B300 separately from GB300 with pricing on request. Confirm GPU count, deployment location, network topology and commitment before comparing proposals.

Published · checked 2026-09-24Official source Capacity needs confirmation
HyperbolicMarketplaceContact salesoffer dependentContact sales

B300 with 288 GB is listed for On-Demand, Reserved and Private Cloud. No public B300 rate. The docs say B300 capacity is rolling out and list no B300 region yet.

Published · checked 2026-09-29Official source Capacity needs confirmation
IBM CloudGPU cloudRequest a quoteproject quoteSelect customers only; support case required

HGX B300 virtual server profiles have Select Availability. Documentation lists Dallas and Washington DC zones. Open a support case to establish eligibility, configuration and pricing; listed regions do not confirm available capacity.

Published · checked 2026-09-24Official source Capacity needs confirmation
IRENGPU cloudContact salesoffer dependentContact sales

B300 with 288GB. Listed node has 256 vCPUs, 3,096 GB RAM and 30TB NVMe. GPU count per node is not stated. Bare metal on InfiniBand. Confirm site and term.

Published · checked 2026-09-29Official source Capacity needs confirmation
LambdaGPU cloudRequest a quoteproject quoteSuperclusters; large multi-year deployments

HGX B300 is listed under Superclusters, whose published product range is 4,000–165,000+ GPUs on 3+ year contracts. Confirm B300-specific sizing, location and price. This is not evidence of hourly single-B300 rental.

Published · checked 2026-09-24Official source Capacity needs confirmation
Oracle Cloud InfrastructureGPU cloudConfirm project pricingproject quoteDocumented eight-GPU bare-metal shape; pricing to confirm

BM.GPU.B300.8 contains eight B300 SXM6 GPUs for single-node and scale-out workloads. No numeric rate was verified in the reviewed documentation. Confirm region, tenancy quota and project pricing; this is separate from GB300 shapes.

Published · checked 2026-09-24Official source Capacity needs confirmation
Voltage ParkGPU cloudContact salesoffer dependentContact sales; long-term contract

HGX B300 listed with up to 2.3TB total GPU memory, 2x Intel Xeon CPUs and 8x 3.84TB NVMe drives. The site says B300 is available only on a long-term contract.

Published · checked 2026-09-29Official source Capacity needs confirmation
VultrGPU cloudContact salesoffer dependentContact sales

HGX B300 listing with 8 GPUs, 2232 GB total GPU memory, 256 vCPUs, 2048 GB RAM and 61 TB storage. The page says to contact sales to reserve capacity.

Published · checked 2026-09-29Official source Capacity needs confirmation

WORK THROUGH IT WITH AN ADVISOR

Compare the full proposal, not just an hourly rate.

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 table labels on-demand, reserved, spot and preemptible offers separately. DigitalOcean spot and Together AI preemptible rates buy interruptible capacity; 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.

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.

WHY WORK WITH US

Powered by
Bridgepointe Technologies.

GPU Cloud Advisors is powered by Bridgepointe Technologies, a technology advisory firm since 2002. You get one advisor, the research on this site, and the provider relationships behind it.

Bridgepointe clients include

  • NVIDIA
  • Salesforce
  • Adobe
  • Uber
  • Netflix
  • Airbnb
  • Roblox
  • LinkedIn
  • CrowdStrike
  • Sony
  • PayPal
  • Mastercard

75+ Fortune 500 clients

Including more than 20 of the Fortune 100, with 97% client satisfaction.

1,200+ data center and cloud projects

Requirements assessed, capacity sourced and providers selected.

Terms, not just the rate

We negotiate ramp schedules, early billing and expansion terms, so you aren’t paying for capacity before your workload can use it.

One requirement, every fit

Bridgepointe partners with 380+ technology providers. We take one requirement to the providers that fit it and bring back a side-by-side comparison you can actually decide from.

Case study: Bridgepointe helped Hive, an AI company, choose data center and connectivity providers across 15 projects, saving its team at least 180 hours. Ask us for the case study

LET’S PLAN YOUR GPU PROJECT

Tell us what
you need.

Two short steps. Tell us what you’re building and an advisor will help you evaluate providers, confirm capacity requirements, and plan the support you need to get into production.

Not sure which GPU or how many you need? Start with your workload and what could change.

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

No obligationSizing help welcome
What we’ll work through with you
  • A provider shortlist suited to your workload and location.
  • Initial capacity, expansion needs and delivery timing to confirm.
  • Total cost, commitment terms and deployment support to compare.
Step 1 of 2What you need

Include GPU type, quantity and likely growth if known. A rough outline is enough.

Step 2 of 2Who you are

We’ll use these details to respond to your project request. Please leave out sensitive data and credentials. Privacy notice.

Prefer to talk it through? Book a free 30-minute consultation or call 844-506-2299.