Start a request

COMPARE GPU PROVIDERS

Find capacity.
Plan for what comes next.

Compare GPU clouds, marketplaces and an ownership alternative. Explore published offers and provider fit, then validate the capacity and terms your project will need.

Compare providers ↓What to check before committing ↗

01 / EXPLORE YOUR OPTIONS

Start with your
project requirements.

Filters cover the offers we have researched, not each provider's full catalog. GPU, service and allocation filters match the same offer. Footprint and deployment help describe the provider; confirm them for the selected GPU.

Showing 38 providers · 77 researched offers

How we label evidence ↓

Published does not mean available now. Region filters describe the sourced footprint, not a confirmed GPU allocation. Colocation appears only without a GPU or rental-allocation filter.

GPU cloudProfile checked 2026-09-24

AceCloud

A cloud infrastructure provider with H200 NVL plans and support for migration and workload configuration.

WHO SHOULD CONSIDER IT · OUR ASSESSMENT

India-based teams seeking H200 NVL and help with deployment.

Published footprint
Noida, India for the listed rate. Mumbai and Atlanta require a separate quote and hardware confirmation. Location source
Customer examples
Bluechip Infocorp: published H200 case study. Customer source
Deployment help
Migration and workload configuration assistance described in the linked offering. Confirm the scope for your project.

Our evaluation: the decision point

Evaluate the location and the scope of migration or configuration help together. Confirm that the proposed GPU configuration and expansion path fit the workload before choosing the plan.

What to ask for

Request the initial Noida allocation and a path to the next GPU count, including whether expansion changes the server, location, connectivity or scope of deployment help.

Researched offers

Needs confirmation: launch capacity, expansion and final terms.

Provider information

Discuss AceCloud for my project ↗
GPU cloudProfile checked 2026-09-24

Beam

Offers on-demand GPU machines alongside serverless tasks, sandboxes and cluster services. Its listed H200 machine includes CPU, RAM and NVMe.

WHO SHOULD CONSIDER IT · OUR ASSESSMENT

Engineering teams combining rented machines with programmatic AI workloads.

Published footprint
Region depends on the machine or connected cloud. Request the location of the quoted H200 before committing. Location source
Customer examples
Magellan AI and Hooktheory: published customer stories. Customer source
Deployment help
Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

Our evaluation: the decision point

Choose between a rented machine, serverless tasks and a connected cluster before comparing costs. Bundled CPU, RAM and local storage can matter as much as the accelerator price.

What to ask for

Ask how to grow from the bundled machine to the next allocation or a cluster, including lead time, region, storage migration and any new reservation requirement.

Researched offers

Needs confirmation: launch capacity, expansion and final terms.

Provider information

Discuss Beam for my project ↗
GPU cloudProfile checked 2026-09-24

CoreWeave

An AI-focused cloud combining GPU instances with Kubernetes, networking and storage services. The published example rents a full eight-GPU HGX H200 node.

WHO SHOULD CONSIDER IT · OUR ASSESSMENT

AI teams evaluating connected GPU infrastructure for substantial training or inference deployments.

Published footprint
North America and Europe pricing regions. Confirm the specific data center and H200 allocation. Location source
Public records
6 US sites checked. Highest stage: Operating (4 of 6). See the records →
Customer examples
Perplexity: featured in CoreWeave customer stories. Customer source
Deployment help
Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

Our evaluation: the decision point

Evaluate the full allocation, network and storage design. A per-GPU calculation does not reduce a whole-node minimum purchase; size the pilot and production budget accordingly.

What to ask for

Request the initial node count and an expansion plan that keeps the required network and storage design intact. Confirm notice periods and what capacity is committed at each stage.

Researched offers

Needs confirmation: launch capacity, expansion and final terms.

Provider information

Discuss CoreWeave for my project ↗
GPU cloudProfile checked 2026-09-24

Crusoe

Provides GPU compute plus managed training and inference services. Those delivery models have separate prices, so choose the operating model before comparing costs.

WHO SHOULD CONSIDER IT · OUR ASSESSMENT

Teams evaluating a path from GPU instances to managed AI services.

Published footprint
Cloud locations include Iceland, Houston, Virginia and Sparks, Nevada. H200 stock varies by location. Location source
Public records
4 US sites checked. Highest stage: Operating (1 of 4). See the records →
Customer examples
Jua: climate-model training on Crusoe Cloud. Customer source
Deployment help
Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

Our evaluation: the decision point

Decide which infrastructure and application tasks your team will operate. Compare raw compute and managed AI proposals with separate scopes and acceptance criteria.

What to ask for

Ask how the chosen compute or inference service handles a larger workload, which capacity requires advance notice and who owns changes to the application or cluster software.

Researched offers

Needs confirmation: launch capacity, expansion and final terms.

Provider information

Discuss Crusoe for my project ↗
GPU cloudProfile checked 2026-09-24

DigitalOcean

GPU Droplets provide GPU instances within DigitalOcean’s broader cloud. Buyers can evaluate GPU compute together with Kubernetes and storage requirements.

WHO SHOULD CONSIDER IT · OUR ASSESSMENT

Product companies wanting GPU capacity alongside familiar application infrastructure.

Published footprint
Region-specific GPU Droplets. Use the regional availability matrix to confirm H200, rather than assuming every DigitalOcean region supports it. Location source
Public records
4 US sites checked. Highest stage: Permitted (1 of 4). See the records →
Customer examples
Workato: published H200 GPU Droplets and Kubernetes deployment. Customer source
Deployment help
Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

Our evaluation: the decision point

Size the initial workload and its next step before choosing a GPU Droplet. Compare the complete bill and the changes required when moving to a larger allocation.

What to ask for

Ask what changes when moving from one GPU to eight or to several machines: region, data migration, downtime, quota approval and reservation terms.

Researched offers

Needs confirmation: launch capacity, expansion and final terms.

Provider information

Discuss DigitalOcean for my project ↗
GPU cloudProfile checked 2026-09-24

E2E Networks

An India-focused GPU cloud offering individual GPU instances and larger training configurations. Its platform also offers model deployment services.

WHO SHOULD CONSIDER IT · OUR ASSESSMENT

Teams evaluating India-based infrastructure for training and production inference.

Published footprint
India-focused cloud; confirm the precise H200 data center and capacity in the order. Location source
Customer examples
CamCom AI and Crownit: named GPU customer testimonials. Customer source
Deployment help
Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

Our evaluation: the decision point

Check whether an India-based deployment fits the users, data and connectivity requirements. Include the complete machine bundle, transfer path and applicable taxes in the proposal.

What to ask for

Ask for the initial allocation and the additional GPUs available in the same site, with notice periods, connectivity requirements and complete costs including applicable taxes.

Researched offers

Needs confirmation: launch capacity, expansion and final terms.

Provider information

Discuss E2E Networks for my project ↗
MarketplaceProfile checked 2026-09-24

Edgevana

Lists dedicated GPU servers through a marketplace. Configurations and prices differ across inventory, making the exact server listing central to the comparison.

WHO SHOULD CONSIDER IT · OUR ASSESSMENT

Buyers comparing bare-metal configurations across multiple locations.

Published footprint
Historical H200 listings included North America, Europe and the Middle East. The marketplace is currently unavailable; confirm the exact deployment location and current inventory with Edgevana. Location source
Customer examples
No named GPU customer verified in the sources reviewed.
Deployment help
Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

Our evaluation: the decision point

Make the exact server listing the unit of comparison. Match configuration, location, term and support responsibility, then establish how additional matching capacity could be secured.

What to ask for

Ask whether additional matching machines can be secured from the same listing or operator, how long the offer is held and who handles support if expansion requires a different host.

Researched offers

Needs confirmation: launch capacity, expansion and final terms.

Provider information

Discuss Edgevana for my project ↗
GPU cloudProfile checked 2026-09-24

Gcore

Combines GPU infrastructure with a broader cloud and edge services portfolio. Its H200 calculator example describes an eight-GPU cluster.

WHO SHOULD CONSIDER IT · OUR ASSESSMENT

Teams evaluating GPU clusters alongside networking and managed inference services.

Published footprint
H200 introduction documented in Luxembourg. The broader edge footprint does not establish H200 availability at every location. Location source
Customer examples
Riga Technical University: genomic research on GPU Cloud. Leonardo.Ai: AI services customer story. Customer source
Deployment help
Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

Our evaluation: the decision point

Resolve the calculator settings, currency, minimum allocation and commitment before using a headline rate. Compare infrastructure and managed inference as separate services.

What to ask for

Request the terms behind the eight-GPU calculator example and the next capacity increment, including region, lead time, contract duration, currency and VAT.

Researched offers

Needs confirmation: launch capacity, expansion and final terms.

Provider information

Discuss Gcore for my project ↗
GPU cloudProfile checked 2026-09-24

GMI Cloud

Offers GPU infrastructure and inference services on one platform. Its published H200 entry rate needs a configuration and commercial-term check.

WHO SHOULD CONSIDER IT · OUR ASSESSMENT

AI companies comparing dedicated training capacity and managed inference, including APAC options.

Published footprint
Provider describes US facilities in Silicon Valley and Colorado, plus Taiwan, Thailand and Malaysia. Confirm H200 at the specific site. Location source
Public records
1 US site checked. Highest stage: Announced. See the records →
Customer examples
Mirelo AI and Higgsfield: named infrastructure examples. Customer source
Deployment help
Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

Our evaluation: the decision point

Qualify the starting price against the proposed facility, GPU count and term. Separate dedicated infrastructure from managed inference and ask how each service grows.

What to ask for

Ask which initial and expanded GPU allocations qualify for the starting rate, in which facilities, on what dates and with what minimum commitment.

Researched offers

Needs confirmation: launch capacity, expansion and final terms.

Provider information

Discuss GMI Cloud for my project ↗
GPU cloudProfile checked 2026-09-24

Hyperstack

NexGen Cloud’s GPU platform offers on-demand, reserved and spot VMs. Region and flavor determine the networking options available to a workload.

WHO SHOULD CONSIDER IT · OUR ASSESSMENT

Teams seeking GPU VMs with published rates and regional deployment choices.

Published footprint
Documented regions: Quebec, Canada; Vestland, Norway; and Texas, US. Confirm which supports your H200 flavor. Location source
Public records
1 US site checked. Highest stage: Announced. See the records →
Customer examples
No named GPU customer verified in the sources reviewed.
Deployment help
Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

Our evaluation: the decision point

Match the VM flavor, network and region to the workload. Confirm the minimum purchase and whether additional VMs can meet the same requirements.

What to ask for

Request the GPU flavor and region for launch and growth. Ask whether added VMs can use the required network and what notice or reservation is needed to secure them.

Researched offers

Needs confirmation: launch capacity, expansion and final terms.

Provider information

Discuss Hyperstack for my project ↗
GPU cloudProfile checked 2026-09-24

Jarvislabs

Offers root-access GPU VMs, preconfigured development templates and GPU clusters. The platform supports notebook, SSH and API-based workflows.

WHO SHOULD CONSIDER IT · OUR ASSESSMENT

Small AI teams wanting ready-to-run environments, with a route to larger GPU jobs.

Published footprint
H200 is documented in IN2 and EU1. EU1 supports one or eight GPUs per instance; confirm the physical site for residency needs. Location source
Customer examples
Hugging Face, Zoho and Smallest AI: named in the provider’s “trusted by” list; no specific H200 deployment established. Customer source
Deployment help
Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

Our evaluation: the decision point

Use the development environment for a representative pilot, then verify the larger production configuration and storage plan. A working experiment does not reserve future capacity.

What to ask for

Ask for a pilot-to-production plan showing the region, next GPU counts, persistent storage and how capacity would be secured if demand rises faster than forecast.

Researched offers

Needs confirmation: launch capacity, expansion and final terms.

Provider information

Discuss Jarvislabs for my project ↗
GPU cloudProfile checked 2026-09-18

Lambda

Offers on-demand GPU instances and reserved multi-node 1-Click Clusters connected with InfiniBand.

WHO SHOULD CONSIDER IT · OUR ASSESSMENT

Teams comparing individual development instances with reserved multi-node training.

Published footprint
Confirm the physical region and GPU configuration for the selected instance or cluster. Location source
Public records
4 US sites checked. Highest stage: Operating (1 of 4). See the records →
Customer examples
No named GPU customer verified for this profile.
Deployment help
Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

Our evaluation: the decision point

Choose between an individual instance and a connected cluster before comparing rates. The larger allocation needs its own workload test and capacity plan.

What to ask for

Request launch and growth allocations, networking, storage and reservation terms for the actual cluster.

Researched offers

Needs confirmation: launch capacity, expansion and final terms.

Provider information

Discuss Lambda for my project ↗
ServerlessProfile checked 2026-09-24

Modal

Runs GPU-backed functions and containers with usage-based billing. The H200 rate covers the accelerator; CPU and memory add to the bill.

WHO SHOULD CONSIDER IT · OUR ASSESSMENT

Python engineering teams serving variable inference or batch workloads without managing a persistent VM.

Published footprint
Multi-cloud region selection includes US, Europe and Asia-Pacific. GPU availability and region-selection surcharges must be checked. Location source
Customer examples
Ramp: published background coding-agent story, not evidence of an H200 rental. Customer source
Deployment help
Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

Our evaluation: the decision point

For variable workloads, compare response-time behavior and the full application bill. Include startup behavior, warm capacity, CPU, memory and concurrency limits.

What to ask for

Test a realistic demand surge. Confirm account and concurrency limits, expected time to add GPU containers, regional requirements and a bill including warm containers, CPU and memory.

Container startup guidance

Researched offers

Needs confirmation: launch capacity, expansion and final terms.

Provider information

Discuss Modal for my project ↗
GPU cloudProfile checked 2026-09-24

Nebius

Provides GPU compute and services for AI development, training and inference. H200 is one of several GPU generations in its published catalog.

WHO SHOULD CONSIDER IT · OUR ASSESSMENT

AI teams comparing GPU cloud infrastructure with managed orchestration and technical support.

Published footprint
Public regions include Finland, France, Israel, Kansas City and the UK. Confirm the H200 configuration within the selected region. Location source
Public records
4 US sites checked. Highest stage: Operating (2 of 4). See the records →
Customer examples
Shopify and Photoroom: documented AI workloads. Customer source
Deployment help
Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

Our evaluation: the decision point

Use rates that apply during the actual project window. Evaluate the orchestration service and distinguish a price discount from a commitment to supply the required capacity.

What to ask for

Request a dated launch and expansion quote, including the applicable October rate, GPU counts, region, required notice and explicit capacity commitments.

Researched offers

Needs confirmation: launch capacity, expansion and final terms.

Provider information

Discuss Nebius for my project ↗
GPU cloudProfile checked 2026-09-24

Oblivus

Provides GPU compute with published per-GPU pricing and a separate regional availability page. A listed SKU does not establish deployable stock.

WHO SHOULD CONSIDER IT · OUR ASSESSMENT

Teams evaluating GPU instances with bundled CPU, RAM and base storage.

Published footprint
Documented multi-tenant sites include Montreal and Norway. Check the availability page for the exact H200 region. Location source
Customer examples
No named GPU customer verified in the sources reviewed.
Deployment help
Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

Our evaluation: the decision point

Check deployable inventory before optimizing the price comparison. Match the selected region and tenancy model to the project, and confirm storage behavior when compute stops.

What to ask for

Confirm the actual launch allocation first, then request a realistic expansion plan and notice period. Document what happens to data and charges if compute must be stopped or moved.

Check current inventory

Researched offers

Needs confirmation: launch capacity, expansion and final terms.

Provider information

Discuss Oblivus for my project ↗
GPU cloudProfile checked 2026-09-24

OVHcloud

Offers GPU instances alongside public cloud infrastructure and separate managed AI products. The listed H200 example is an eight-GPU instance.

WHO SHOULD CONSIDER IT · OUR ASSESSMENT

Teams evaluating H200 within a broader cloud environment, especially European deployment options.

Published footprint
Paris is listed for public-cloud H200; managed AI products also document H200 in Gravelines. Service and region must be matched. Location source
Public records
2 US sites checked. Highest stage: Operating (2 of 2). See the records →
Customer examples
Picsell.ia: documented use of GPU instances for its computer-vision platform. Customer source
Deployment help
Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

Our evaluation: the decision point

Choose the specific public-cloud instance or managed AI service first. Services under the same provider can have different regions, minimum allocations and operating responsibilities.

What to ask for

Ask how to add capacity beyond the first eight-GPU instance in the selected region, including account limits, availability, networking and changes to the full project cost.

Researched offers

Needs confirmation: launch capacity, expansion and final terms.

Provider information

Discuss OVHcloud for my project ↗
GPU cloudProfile checked 2026-09-24

Runpod

Offers GPU Pods, serverless workers and clusters. Each is a different operating model, so rates should be compared within the product you intend to use.

WHO SHOULD CONSIDER IT · OUR ASSESSMENT

Developer-led companies wanting flexible GPU Pods and a separate serverless deployment option.

Published footprint
Location is selected with the available Pod or cluster. Confirm data center and Secure versus Community Cloud before deployment. Location source
Public records
4 US sites checked. Highest stage: Permitted (1 of 4). See the records →
Customer examples
TOOL, Aneta and Gendo: published customer stories across media and inference workloads. Customer source
Deployment help
Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

Our evaluation: the decision point

Choose between Pods, Serverless and Clusters before comparing rates. Establish where checkpoints and model files live and how they survive a change of machine.

What to ask for

Ask whether extra Pods can access the required persistent data in the selected location, how capacity is secured and how checkpoints survive a change of machine.

Storage persistence rules

Researched offers

Needs confirmation: launch capacity, expansion and final terms.

Provider information

Discuss Runpod for my project ↗
GPU cloudProfile checked 2026-09-24

Scaleway

European cloud provider with an eight-GPU B300 instance and separate storage and cloud services.

WHO SHOULD CONSIDER IT · OUR ASSESSMENT

Teams that need a documented Paris deployment and can use an entire eight-GPU node.

Published footprint
B300-SXM-8-288G is documented in FR-PAR-2, Paris, France. Listing a region does not confirm available capacity. Location source
Customer examples
No named GPU customer verified for this profile.
Deployment help
Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

Our evaluation: the decision point

Budget for eight GPUs and test the exact network topology. A published location or headline bandwidth does not establish the inter-node design your training job requires.

What to ask for

Request the node count available in Paris for launch, how expansion is reserved, the inter-node network and storage quote, and who handles workload deployment.

Researched offers

Needs confirmation: launch capacity, expansion and final terms.

Provider information

Discuss Scaleway for my project ↗
MarketplaceProfile checked 2026-09-24

Spheron

Sources GPU rental capacity from data center partners. Its H200 page includes dedicated and spot options, with different interruption terms.

WHO SHOULD CONSIDER IT · OUR ASSESSMENT

Buyers comparing partner-supplied capacity across regions through one marketplace.

Published footprint
H200 capacity advertised across North America, Europe and Asia. The dashboard identifies live partner availability. Location source
Customer examples
Compute Desk: published GPU capacity-sourcing case study. Customer source
Deployment help
Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

Our evaluation: the decision point

Evaluate the underlying partner offer as well as the marketplace. Match interruption terms, region and support ownership to the workload and its tolerance for disruption.

What to ask for

Ask whether growth can stay with the same partner and region, who owns incidents across partners and what capacity is committed versus dependent on future marketplace offers.

Researched offers

Needs confirmation: launch capacity, expansion and final terms.

Provider information

Discuss Spheron for my project ↗
GPU cloudProfile checked 2026-09-21

STN

GPU One offers reserved, single-tenant NVIDIA capacity and managed Kubernetes.

WHO SHOULD CONSIDER IT · OUR ASSESSMENT

Teams evaluating a longer-term dedicated allocation with a managed orchestration layer.

Published footprint
GPU One lists Silicon Valley, Los Angeles and Chicago. Confirm GPU model and allocation at the selected site. Location source
Public records
3 US sites checked. Highest stage: Permitted (3 of 3). See the records →
Customer examples
No named GPU customer verified for this profile.
Deployment help
Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

Our evaluation: the decision point

Compare the published three-year minimum with the uncertainty in your workload. Managed Kubernetes does not by itself define responsibility for deploying and operating your application.

What to ask for

Request a staged allocation plan, growth rights, deployment responsibilities and the full commitment cost before choosing a term.

Researched offers

Needs confirmation: launch capacity, expansion and final terms.

Provider information

Discuss STN for my project ↗
ColocationProfile checked 2026-09-21

TierPoint

Provides high-density colocation for customer-owned AI hardware, including cooling options for dense GPU deployments.

WHO SHOULD CONSIDER IT · OUR ASSESSMENT

Companies evaluating ownership and hosting of GPUs alongside cloud rental.

Published footprint
Confirm the proposed facility, available power and cooling design for your deployment. Location source
Public records
4 US sites checked. Highest stage: Operating (1 of 4). See the records →
Customer examples
No named GPU customer verified for this profile.
Deployment help
Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

Our evaluation: the decision point

This is an ownership and hosting option. Include hardware purchase, operations, power and connectivity when comparing it with rental.

What to ask for

Request energized capacity, installation milestones, an expansion path and responsibility for replacing failed equipment.

Ownership and hosting option

    This profile covers colocation. No GPU rental offer is listed. Compare GPU ownership and hosting →

    Needs confirmation: launch capacity, expansion and final terms.

    Provider information

    Discuss TierPoint for my project ↗
    GPU cloudProfile checked 2026-09-24

    Together AI

    Offers GPU clusters as well as managed inference and model services. H200 cluster rental is priced separately from model API usage.

    WHO SHOULD CONSIDER IT · OUR ASSESSMENT

    AI companies comparing dedicated GPU training with a platform that also supplies inference services.

    Published footprint
    Confirm the H200 cluster’s physical region in the quote. The public GPU rate does not specify the location of your allocation. Location source
    Public records
    3 US sites checked. Highest stage: Operating (1 of 3). See the records →
    Customer examples
    Cursor: published story about low-latency inference at scale. Customer source
    Deployment help
    Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

    Our evaluation: the decision point

    Separate GPU cluster rental from model APIs and endpoint usage. Match the reservation to a tested schedule, including setup, validation and expected expansion.

    What to ask for

    Request a plan for increasing the cluster during the project: GPU increments, network design, lead time and whether the expansion changes the reservation term or rate.

    Researched offers

    Needs confirmation: launch capacity, expansion and final terms.

    Provider information

    Discuss Together AI for my project ↗
    MarketplaceProfile checked 2026-09-24

    Vast.ai

    A GPU host marketplace with dynamic offers. Price and suitability depend on the selected host, machine and rental terms.

    WHO SHOULD CONSIDER IT · OUR ASSESSMENT

    Technically capable teams willing to compare individual hosts and machine specifications.

    Published footprint
    Host-specific locations in live marketplace listings. Confirm the physical location and data-handling requirements with the chosen host. Location source
    Customer examples
    No named GPU customer verified in the sources reviewed.
    Deployment help
    Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

    Our evaluation: the decision point

    Evaluate the selected host, storage, network and reliability characteristics. Resolve contractual and support requirements before depending on a particular marketplace offer.

    What to ask for

    Ask whether the exact host can supply your next allocation. If not, establish how to move data, reproduce the environment and arrange support on a replacement host.

    Researched offers

    Needs confirmation: launch capacity, expansion and final terms.

    Provider information

    Discuss Vast.ai for my project ↗
    GPU cloudProfile checked 2026-09-24

    Verda (formerly DataCrunch)

    Combines GPU instances, instant clusters, serverless containers and storage. Formerly DataCrunch, it operates a full-stack AI cloud.

    WHO SHOULD CONSIDER IT · OUR ASSESSMENT

    AI teams evaluating a Europe-based platform from experimentation through training and serving.

    Published footprint
    Published Nordic footprint includes Finland and Iceland. Confirm the H200 region and residency scope for the selected service. Location source
    Customer examples
    Magnific, ExpressVPN and 1X: published customer stories across media, confidential inference and robotics. Customer source
    Deployment help
    Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

    Our evaluation: the decision point

    Choose the instance, cluster or serverless service before comparing rates. Confirm that the intended compute, storage and networking are available together in the required region.

    What to ask for

    Request a growth plan within the selected service and Nordic region. Clarify which changes require a new cluster, data movement, reservation or revised support scope.

    Researched offers

    Needs confirmation: launch capacity, expansion and final terms.

    Provider information

    Discuss Verda (formerly DataCrunch) for my project ↗
    GPU cloudProfile checked 2026-09-24

    Massed Compute

    Offers hourly GPU virtual machines. Its published B300 configuration bundles eight GPUs with CPU and memory.

    WHO SHOULD CONSIDER IT · OUR ASSESSMENT

    Teams evaluating an eight-GPU VM for training or inference with hourly billing.

    Published footprint
    B300 deployment location needs confirmation; the reviewed pricing page does not establish a specific available region. Location source
    Customer examples
    No named GPU customer verified in the sources reviewed.
    Deployment help
    Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

    Our evaluation: the decision point

    Budget for the entire eight-GPU VM, then validate the network and storage requirements of your workload.

    What to ask for

    Request the deployment location, initial node count and lead time for additional matching nodes. Confirm interconnect and recovery support.

    Researched offers

    Needs confirmation: launch capacity, expansion and final terms.

    Provider information

    Discuss Massed Compute for my project ↗
    GPU cloudProfile checked 2026-09-24

    Latitude.sh

    Provides dedicated bare-metal GPU servers and clusters. Its g4.b300.large configuration contains eight HGX B300 GPUs with advertised dual-plane RoCE networking.

    WHO SHOULD CONSIDER IT · OUR ASSESSMENT

    Teams that want dedicated hardware and can operate their training or inference software stack.

    Published footprint
    B300 deployment location needs confirmation; the reviewed pricing page does not establish a specific available region. Location source
    Customer examples
    No named GPU customer verified in the sources reviewed.
    Deployment help
    Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

    Our evaluation: the decision point

    Check the exact server configuration and network topology against a workload trial before reserving a larger cluster.

    What to ask for

    Request the available B300 region, expansion increments, network design and notice required to add matching nodes.

    Researched offers

    Needs confirmation: launch capacity, expansion and final terms.

    Provider information

    Discuss Latitude.sh for my project ↗
    GPU cloudProfile checked 2026-09-24

    Cirrascale

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

    WHO SHOULD CONSIDER IT · OUR ASSESSMENT

    Teams evaluating a dedicated eight-GPU configuration for model training or fine-tuning.

    Published footprint
    B300 deployment region needs confirmation; the reviewed source does not establish a specific available location. Location source
    Public records
    2 US sites checked. Highest stage: Permitted (1 of 2). See the records →
    Customer examples
    No named GPU customer verified in the sources reviewed.
    Deployment help
    Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

    Our evaluation: the decision point

    Match the proposed allocation, access requirements and contract term to your workload before comparing the full project cost.

    What to ask for

    Request a written launch allocation and expansion schedule, including location, notice periods, network design and responsibility for deployment.

    Researched offers

    Needs confirmation: launch capacity, expansion and final terms.

    Provider information

    Discuss Cirrascale for my project ↗
    GPU cloudProfile checked 2026-09-24

    CUDO Compute

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

    WHO SHOULD CONSIDER IT · OUR ASSESSMENT

    Enterprise teams seeking a tailored GPU cluster proposal with an explicit capacity growth plan.

    Published footprint
    B300 deployment region needs confirmation; the reviewed source does not establish a specific available location. Location source
    Customer examples
    No named GPU customer verified in the sources reviewed.
    Deployment help
    Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

    Our evaluation: the decision point

    Match the proposed allocation, access requirements and contract term to your workload before comparing the full project cost.

    What to ask for

    Request a written launch allocation and expansion schedule, including location, notice periods, network design and responsibility for deployment.

    Researched offers

    Needs confirmation: launch capacity, expansion and final terms.

    Provider information

    Discuss CUDO Compute for my project ↗
    GPU cloudProfile checked 2026-09-24

    IBM Cloud

    HGX B300 virtual server profiles have Select Availability. Documentation lists Dallas and Washington DC zones.

    WHO SHOULD CONSIDER IT · OUR ASSESSMENT

    Organizations evaluating GPU virtual servers within IBM Cloud and able to complete its access process.

    Published footprint
    B300 documentation lists Dallas and Washington DC zones, subject to Select Availability and account eligibility. Location source
    Public records
    4 US sites checked. Highest stage: Operating (2 of 4). See the records →
    Customer examples
    No named GPU customer verified in the sources reviewed.
    Deployment help
    Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

    Our evaluation: the decision point

    Match the proposed allocation, access requirements and contract term to your workload before comparing the full project cost.

    What to ask for

    Request a written launch allocation and expansion schedule, including location, notice periods, network design and responsibility for deployment.

    Researched offers

    Needs confirmation: launch capacity, expansion and final terms.

    Provider information

    Discuss IBM Cloud for my project ↗
    GPU cloudProfile checked 2026-09-24

    Oracle Cloud Infrastructure

    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.

    WHO SHOULD CONSIDER IT · OUR ASSESSMENT

    Teams evaluating bare-metal B300 nodes within OCI for single-node or distributed AI workloads.

    Published footprint
    B300 deployment region needs confirmation; the reviewed source does not establish a specific available location. Location source
    Public records
    4 US sites checked. Highest stage: Power secured (1 of 4). See the records →
    Customer examples
    No named GPU customer verified in the sources reviewed.
    Deployment help
    Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

    Our evaluation: the decision point

    Match the proposed allocation, access requirements and contract term to your workload before comparing the full project cost.

    What to ask for

    Request a written launch allocation and expansion schedule, including location, notice periods, network design and responsibility for deployment.

    Researched offers

    Needs confirmation: launch capacity, expansion and final terms.

    Provider information

    Discuss Oracle Cloud Infrastructure for my project ↗
    GPU cloudProfile checked 2026-09-29

    Voltage Park

    Voltage Park sells bare-metal NVIDIA HGX H100 capacity on demand and on long-term reserve. Its site lists B200, B300, GB200 and GB300 systems on long-term contracts, and says the company has merged with Lightning AI.

    WHO SHOULD CONSIDER IT · OUR ASSESSMENT

    Teams that want self-serve H100 clusters now, or a long-term contract for Blackwell systems in US data centers.

    Published footprint
    Voltage Park says it runs six Tier 3+ designed data centers in four US states. The site does not name the sites, so confirm where your GPU would run. Location source
    Public records
    2 US sites checked. Highest stage: Site identified (1 of 2). See the records →
    Customer examples
    Phind. Its CEO is quoted as a Voltage Park customer on the infrastructure page. Customer source
    Deployment help
    Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

    Our evaluation: the decision point

    Use the self-serve H100 tiers for bursts and testing. Treat Blackwell as a contract negotiation, since B200 and B300 are sold only on long-term terms.

    What to ask for

    Request a written H100 rate for your cluster size and network type, and a B200 or B300 quote with term, delivery date and site.

    Researched offers

    Needs confirmation: launch capacity, expansion and final terms.

    Provider information

    Discuss Voltage Park for my project ↗
    GPU cloudProfile checked 2026-09-29

    Denvr Dataworks

    Denvr Dataworks is a Calgary-based AI cloud that sells GPU VMs, bare metal and model endpoints. It publishes hourly rates for H100 and lists H200 and B200 as reserved only.

    WHO SHOULD CONSIDER IT · OUR ASSESSMENT

    Teams that want published H100 rates, or data residency in Canada or the US.

    Published footprint
    Documented regions: MSC1 Canada West (Calgary) and HOU1 US South (Houston). The docs list H100 nodes in HOU1 only, so confirm the region for your GPU. Location source
    Public records
    1 US site checked. Highest stage: Operating. See the records →
    Customer examples
    No named GPU customer verified in the sources reviewed.
    Deployment help
    Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

    Our evaluation: the decision point

    H100 has a published rate and on-demand VMs. H200 and B200 need a reserved contract, so plan for a term commitment before relying on them.

    What to ask for

    Request H200 or B200 reserved pricing by term and cluster size, the region, and the date capacity can be delivered.

    Researched offers

    Needs confirmation: launch capacity, expansion and final terms.

    Provider information

    Discuss Denvr Dataworks for my project ↗
    GPU cloudProfile checked 2026-09-29

    Vultr

    Vultr is a global cloud that sells NVIDIA and AMD GPUs as VMs, bare metal and self-service clusters. It publishes starting per-GPU rates for B200 and H100 and sells B300 through sales.

    WHO SHOULD CONSIDER IT · OUR ASSESSMENT

    Teams that want GPU capacity alongside a full cloud stack across many global regions.

    Published footprint
    Vultr lists 33 cloud data center regions across North America, South America, Europe, Asia, Australia and Africa. The pricing page does not tie GPUs to regions, so confirm which region offers your GPU. Location source
    Public records
    3 US sites checked. Highest stage: Power secured (1 of 3). See the records →
    Customer examples
    Athos Therapeutics. Vultr’s customer page quotes its CEO on using Vultr Cloud GPU. Customer source
    Deployment help
    Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

    Our evaluation: the decision point

    Treat the published rates as starting points and get the exact rate for your term in writing. Confirm which region has stock before you set a launch date.

    What to ask for

    Request the hourly rate, term and region for an eight-GPU B200 or H100 system. Ask for a B300 quote with delivery date.

    Researched offers

    Needs confirmation: launch capacity, expansion and final terms.

    Provider information

    Discuss Vultr for my project ↗
    GPU cloudProfile checked 2026-09-29

    IREN

    IREN, formerly Iris Energy, sells bare-metal NVIDIA GPU clusters through IREN Cloud from data centers it owns and operates. Clusters use InfiniBand networking and are sold through sales, with no published rates.

    WHO SHOULD CONSIDER IT · OUR ASSESSMENT

    Teams that want a contracted bare-metal GPU cluster from an operator that owns its data centers and power.

    Published footprint
    IREN lists sites in British Columbia, Canada (Mackenzie, Prince George, Canal Flats) and Texas, US (Childress, with Sweetwater under construction). Oklahoma is in development and South Australia is planned. Confirm which site would host your cluster. Location source
    Public records
    4 US sites checked. Highest stage: Operating (1 of 4). See the records →
    Customer examples
    Hume AI, which IREN says uses IREN Cloud to train and run its voice AI models. Customer source
    Deployment help
    Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

    Our evaluation: the decision point

    Use IREN for a contracted cluster, not for on-demand hours. Every GPU is contact sales, so compare quotes on term, node count and site.

    What to ask for

    Request a quote with the GPU model, node count, site, term, start date and storage. Ask whether the listed node specs match what you would receive.

    Researched offers

    Needs confirmation: launch capacity, expansion and final terms.

    Provider information

    Discuss IREN for my project ↗
    GPU cloudProfile checked 2026-09-29

    Nscale

    Nscale is a UK-based AI cloud that sells bare-metal NVIDIA GPU nodes and reserved clusters, with managed Kubernetes, Slurm and inference services on top. GPU capacity is reserved through sales.

    WHO SHOULD CONSIDER IT · OUR ASSESSMENT

    Teams that want reserved Blackwell or Hopper capacity in Europe or the US, with optional managed orchestration.

    Published footprint
    Nscale lists its own data centers in Narvik and Glomfjord, Norway; Loughton, UK; and Texas, US. Partner-run sites include Sines, Portugal; Keflavik and Blönduós, Iceland; and Oslo and Stavanger, Norway. Confirm which site supports your GPU. Location source
    Public records
    3 US sites checked. Highest stage: Operating (2 of 3). See the records →
    Customer examples
    Microsoft, which contracted about 200,000 NVIDIA GB300 GPUs across Europe and the US. Customer source
    Deployment help
    Needs confirmation for your project. Infrastructure or managed orchestration does not establish application deployment support.

    Our evaluation: the decision point

    Use Nscale for reserved cluster capacity. Its docs say GB300 NVL72 and B200 are sold as reservations of whole NVLink domains, so size the order up front.

    What to ask for

    Request the GPU, region, reservation size in units, term and delivery date. Ask whether a reservation can be expanded later, since the docs say it is fixed in size.

    Researched offers

    Needs confirmation: launch capacity, expansion and final terms.

    Provider information

    Discuss Nscale for my project ↗
    GPU cloudProfile checked 2026-09-29

    TensorWave

    An AMD-only AI cloud that sells bare-metal AMD Instinct nodes with optional managed Kubernetes and Slurm. Published product pages list MI300X, MI325X and MI355X nodes with eight GPUs each.

    WHO SHOULD CONSIDER IT · OUR ASSESSMENT

    Teams ready to run training or inference on eight-GPU AMD nodes with vendor engineering support.

    Published footprint
    Company posts describe its largest AMD cluster in Tucson, Arizona and a 10MW site near Pittsburgh, Pennsylvania targeted for the first half of 2026. Confirm which site hosts your GPU. Location source
    Public records
    3 US sites checked. Highest stage: Operating (1 of 3). See the records →
    Customer examples
    Fireworks AI and Luma AI: named in TensorWave’s Series B release. Customer source
    Deployment help
    Migration and workload configuration assistance described in the linked offering. Confirm the scope for your project.

    Our evaluation: the decision point

    Treat the published rates as starting points for full eight-GPU nodes. Confirm the term behind the rate and test your models on ROCm before you commit.

    What to ask for

    Request the rate, term and site for the node count you need, plus a proof of concept on your own models.

    Researched offers

    Needs confirmation: launch capacity, expansion and final terms.

    Provider information

    Discuss TensorWave for my project ↗
    GPU cloudProfile checked 2026-09-29

    Hot Aisle

    An AMD-only cloud that sells self-service MI300X virtual machines and eight-GPU bare-metal servers. Customers add credits and launch compute from a terminal UI, API or CLI without a sales call.

    WHO SHOULD CONSIDER IT · OUR ASSESSMENT

    Developers and small teams that want to test or serve models on one to eight MI300X GPUs with no minimum commitment.

    Published footprint
    Documented site: Switch Pyramid data center in Grand Rapids, Michigan. Confirm capacity before planning a multi-node cluster. Location source
    Public records
    1 US site checked. Highest stage: Operating. See the records →
    Customer examples
    No named GPU customer verified in the sources reviewed.
    Deployment help
    Migration and workload configuration assistance described in the linked offering. Confirm the scope for your project.

    Our evaluation: the decision point

    A low-friction way to trial MI300X with a single GPU. Confirm how many eight-GPU servers are free before you plan a larger launch.

    What to ask for

    Ask how many bare-metal MI300X nodes are open today and what notice is needed to add more.

    Researched offers

    Needs confirmation: launch capacity, expansion and final terms.

    Provider information

    Discuss Hot Aisle for my project ↗
    MarketplaceProfile checked 2026-09-29

    Hyperbolic

    Hyperbolic sells on-demand and reserved GPU instances drawn from an aggregated supplier network. It also arranges single-tenant Private Cloud clusters under custom contracts.

    WHO SHOULD CONSIDER IT · OUR ASSESSMENT

    Teams that want self-serve H100 or H200 capacity by the hour, with a path to reserved or private clusters later.

    Published footprint
    Documented regions include US Central, US East, US Southeast, Canada East, Northern Europe, Western Europe, the UK and Japan. H200 is listed only in some US East, US Southeast and Northern Europe regions, so confirm which supports your GPU. Location source
    Customer examples
    No named GPU customer verified in the sources reviewed.
    Deployment help
    Migration and workload configuration assistance described in the linked offering. Confirm the scope for your project.

    Our evaluation: the decision point

    Treat the published rates as starting prices from a supplier network. Rates vary by region and stock, so confirm the actual GPU, region and interconnect in the console before you plan a launch.

    What to ask for

    Ask which supplier site and region will host the GPUs, whether added nodes can join the same interconnect, and when B200 or B300 capacity is live.

    Researched offers

    Needs confirmation: launch capacity, expansion and final terms.

    Provider information

    Discuss Hyperbolic for my project ↗

    WORK THROUGH IT WITH AN ADVISOR

    Turn your shortlist into a project plan.

    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 ↗

    02 / VALIDATE THE SHORTLIST

    Five questions.
    A more useful comparison.

    Use the same evidence standard with every provider. Our buying guide turns nine deal risks into practical checks.

    01 / Capacity and growth

    Can they deliver and expand?

    Initial allocation, launch date, expansion and priority.

    Evidence to request →    02 / Workload performance

    Will the workload perform?

    Test the proposed cluster, software and recovery path.

    Evidence to request →    03 / Cost and flexibility

    Will the economics still work?

    Model the full bill, commitments, resizing and renewal.

    Evidence to request →    04 / Operational continuity

    Can the service keep running?

    Evaluate provider continuity, site resilience and fallback capacity.

    Evidence to request →    05 / Data and responsibilities

    Are data and responsibilities covered?

    Define security, location, portability and operating ownership.

    Evidence to request →
    H100 pricing →H200 pricing →B200 pricing →B300 pricing →All providers →

    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.