GPUBank vs CUDO Compute
CUDO Compute, run by Cudo Ventures Ltd under English law, now markets quote-only enterprise clusters but keeps a self-serve console and public API with live prices in two Nordic data centers. Its H100 SXM VM costs $1.79 per GPU-hour for the GPU alone, with vCPU, RAM, disk and IPv4 billed on top; GPUBank's H100 SXM is $1.85 with all of these included. CUDO's 8x H100 bare metal and its InfiniBand clusters suit larger jobs; GPUBank includes compute in the price, stops billing compute on a stopped instance and takes crypto.
GPUBank
Tallinn, Estonia
H100 SXM
$1.85 / h
- vCPU, RAM, NVMe, IPv4 and traffic included
- Per-second billing after the health check
- Top up in Bitcoin, USDT and 12+ other coins
vCPU, RAM, NVMe, IPv4 and traffic included
CUDO Compute
United Kingdom
H100 SXM
$1.79 / h
- H100 SXM GPU rate of $1.79 per GPU-hour, before vCPU and RAM
- 8x H100 SXM or H100 NVL bare metal at $14.32 per hour
- Self-serve clusters of 8-GPU nodes with InfiniBand
CUDO Compute
GPU cloud
26 points compared37 of their pages readprices last checked on 29 September 2026page updated 2 November 2026
The verdict
A lower H100 GPU rate and 8-GPU bare metal at CUDO. All-in prices and a cheap stop state at GPUBank.
CUDO suits teams that want H100 or L40S capacity in Norway or Sweden, full 8-GPU bare-metal nodes, and API or Terraform automation. Its H100 SXM GPU rate of $1.79 is below GPUBank's $1.85, but on CUDO's VMs that rate is the GPU only, and vCPU, RAM, disk and IPv4 are metered separately; its 8x H100 bare metal at $14.32 per hour includes CPU, RAM and NVMe. A stopped CUDO VM keeps billing for all its resources, and resources are deleted when the balance reaches zero. GPUBank is cheaper on L40S, runs four regions, and keeps only the disk on the bill when you stop.
CUDO ComputeGPUBank 01The real cost
The same GPUs, side by side, per hour.
Every GPU model both of us rent, at each provider's own published price for one card, in US dollars, excluding VAT. Prices last checked on 29 September 2026.
| GPU | CUDO Compute per hour | GPUBank per hour | Difference | 100 h at CUDO Compute | 100 h at GPUBank |
|---|---|---|---|---|---|
 H100 SXM80 GB |
$1.79 | $1.85 | +3 % | $179.00 | $185.00 |
 L40S48 GB |
$0.75 | $0.55 | −27 % | $75.00 | $55.00 |
A fine-tune, 40 h on an H100 SXM
CUDO Compute
$71.60
GPUBank
$74.00
Costs $2.40 more at GPUBank (+3 %)
An inference box, a month, 720 h, on an L40S
CUDO Compute
$540.00
GPUBank
$396.00
Saves $144.00 at GPUBank (−27 %)
02Their catalogue
Every GPU CUDO Compute lists.
CUDO Compute lists 24 GPU models. The 11 below are the ones recorded in our check, each at its lowest public price per GPU-hour. 4 of them are also rented at GPUBank (in blue).
GB300reservedNo price- GB200reservedNo price
B300reservedNo price- B200180 GB · reservedNo price
H200141 GB · reservedNo price- H100 SXM80 GB · on demand$1.79
H100 NVL96 GB · on demand$1.79- B100192 GB · reservedNo price
- VR200288 GB · reservedNo price
A100 PCIe 80GB80 GB · on demand$1.50- L40S48 GB · on demand$0.75
03What the price includes
What appears on the bill besides the GPU.
The same H100 SXM session of 2 h 14 m 03 s, line by line, as each provider describes its billing.
CUDO Compute
2 h 14 m 03 s on H100 SXM
- GPU · $1.79 / h
- $4.00
- vCPU and RAM
- Billed separately
- Storage
- $0.00014 per GiB-hour, $0.10 per GiB-month
- Egress
- Not published
- Public IPv4
- $0.0035 per hour in Norway, free in Sweden
- Billing granularity
- Per second
- Minimum charge
- Not published
GPUBank
2 h 14 m 03 s on H100 SXM
- GPU · $1.85 / h
- $4.13
- vCPU and RAM
- Included
- Storage
- Local NVMe included; persistent disk $0.03 per GB-month
- Egress
- Included
- Public IPv4
- Included
- Billing granularity
- Per second, from the health check
- Minimum charge
- None
While stopped
What keeps running on the bill.
CUDO Compute
All resources billed until deleted.
GPUBank
Only the persistent disk, $0.03 per GB-month. Compute, IPv4 and scratch cost nothing.
04Where the GPUs run
2 countries against 4, and where they are.
Countries where CUDO Compute says its GPUs run, from its own region list or documentation, against GPUBank's four regions. A pin marks a country, not the exact site.
Norway
Sweden
Paris
Frankfurt
Ashburn
Singapore
CUDO Compute2
Norway
Sweden
GPUBank4
Paris
Frankfurt
Ashburn
Singapore
CUDO Compute · 2 countriesGPUBank · 4 regions
05Paying and signing up
How each one takes your money.
CUDO Compute
Payments and sign-up
Payment methods
Invoice, enterprise only
No cryptoMinimum top-up $10
To start
- 1Email (verified)
- 2Billing account: full name, country, address
- 3Card or PayPal to add credit
Billing modelPrepaid credits
GPUBank
Payments and sign-up
Payment methods
A prepaid USD balance, topped up in crypto. Every GPU-second is drawn from it.
Minimum top-up $25
To start
- 1Email and password
- 2Top up in crypto
- 3Launch a GPU
Billing modelPrepaid balance
06Trust and support
Certificates, uptime and who answers.
| CUDO Compute | GPUBank | |
|---|---|---|
| SOC 2 | Listed, report not public | Not published |
| ISO 27001 | Data centers only | Not published |
| HIPAA | Not published | No |
| GDPR DPA | Listed, report not public | Published DPA, applies automatically |
| Availability target | 99.5 % on committed resources only | 99.9 % monthly, per region |
| Status page | status.cudocompute.com | Public status page |
| Support | Console tickets; 24/7 on enterprise clusters | Tickets from a $25 balance |
07Their site, our site
What you see before you pay.
Both sites as a visitor sees them, without an account: the CUDO Compute pricing page captured on 29 September 2026, and GPUBank's live board. Cookie banners and chat bubbles are closed; nothing else is edited.
- 1The pricing page lists 8 GPUs, all marked Quote on request.
- 2Each GPU product page shows Starting from: Contact us for pricing.
- 3The marketing site does not link to the self-serve console or its prices; the docs still describe the on-demand platform.
- 1Live stock per region and every price, without an account.
- 2Per-hour and per-second prices, with vCPU, RAM, NVMe, IPv4 and traffic in them.
- 3A machine-readable price feed at /prices.json.
08The spec sheet
26 points, read off both sites.
GPUBank aheadCUDO Compute aheadEven
| Criterion | GPUBank | CUDO Compute |
|---|---|---|
| Price and catalogue | ||
| H100 SXM, per GPU-hourvCPU, RAM, disk and IPv4 are billed on top. The 8x bare-metal node is $14.32 per hour. | $1.85, all-in | $1.79, GPU only on VMsAhead |
| Cheapest GPU-hour | $0.23, RTX 4090 | $0.75, L40S on an 8-GPU node |
| GPU models listedMost models are quote-only or had no price or capacity on the check date. | 10 | 24 listed, 4 with a self-serve priceAhead |
| What the GPU price includesBare-metal nodes include CPU, RAM and local NVMe. | vCPU, RAM, NVMe, IPv4, traffic | VMs: GPU only; vCPU, RAM, disk, IPv4 extra |
| Billing | ||
| Billing granularity | Per second | Per second |
| Minimum charge | None | Not published |
| Stopped instance | Disk only | All resources billed until deleted |
| Persistent storage | $0.03 per GB-month | $0.00014 per GiB-hour, $0.10 per GiB-month |
| Egress traffic | Included | Not published |
| Public IPv4 | Included | $0.0035 per hour in Norway, free in Sweden |
| Billing model | Prepaid balance | Prepaid credits |
| Payment and sign-up | ||
| Crypto payments | Bitcoin, USDT, Ethereum and 12+ more | Not published |
| Card payments | No | Visa, Mastercard, PayPalAhead |
| Minimum deposit | $25 | $10Ahead |
| Platform | ||
| API and CLI | REST API, CLI, Python SDK | REST API, cudoctl CLI, Python client |
| Terraform | Official provider | Official provider |
| Serverless GPUs | No | No |
| Managed Kubernetes | No | No |
| Ready-made templates | 25 | 7 quick-deploy apps |
| Interruptible or spot | Yes, 40 % off | No |
| GPUs per instance | Up to 8 | 8 per machine; self-serve clustersAhead |
| Multi-node interconnect | Not offered | InfiniBand on cluster machine typesAhead |
| Network and regions | ||
| RegionsMarketing pages cite 16 data centers for enterprise clusters without naming them. | 4 | 2 self-serve |
| Countries | France, Germany, US, Singapore | Norway, Sweden |
| EU regions | Paris, Frankfurt | Norway and Sweden |
| Trust and support | ||
| Availability targetNo SLA is published for on-demand use. | 99.9 % monthly | 99.5 % on committed resources only |
| Compliance reportsThe About page says SOC 2-aligned controls. No report was seen. | None published | SOC 2 Type II and ISO 27001 claimedAhead |
| Public status page | Yes | status.cudocompute.com |
| Support | Tickets from a $25 balance | Console tickets; 24/7 on enterprise clustersAhead |
09The right fit
Different needs. A clear choice.
What CUDO Compute does better
- H100 SXM GPU rate of $1.79 per GPU-hour, before vCPU and RAM
- 8x H100 SXM or H100 NVL bare metal at $14.32 per hour
- Self-serve clusters of 8-GPU nodes with InfiniBand
- Cards and PayPal, from a $10 top-up
- Enterprise access to B300, GB300 NVL72 and Vera Rubin by quote
Choose GPUBank if
- vCPU, RAM, NVMe, IPv4 and traffic included in the GPU price
- Stop an instance and pay only for its disk, at $0.03 per GB-month
- L40S at $0.55 per GPU-hour, against $0.75 on an 8-GPU node
- Four regions in France, Germany, the US and Singapore, against two Nordic sites
- Interruptible at -40 %, crypto top-ups and a published uptime target
Moving from CUDO Compute takes an afternoon.
CUDO boot disks can be saved as private images and its object storage is S3-compatible, so data moves out with standard S3 tools. A stopped CUDO VM keeps billing, so delete it once the copy is done.
- 1Launch the same GPU on GPUBank with a persistent disk in Frankfurt or Paris, the regions closest to the Nordics.
- 2Copy data with rsync over SSH, or from CUDO object storage with s3cmd or s5cmd.
- 3Run your Docker image or a matching template on GPUBank, then delete the CUDO VM and disks.
# 1. a GPU and a disk on GPUBank gpubank launch --gpu h100-sxm --region eu-paris --disk data # 2. copy your data from the old machine rsync -azP old-host:/workspace/ root@<new-ip>:/workspace/ # 3. run the same image or template ssh root@<new-ip> 'cd /workspace && ./start.sh'
A few things you might be wondering.
On the GPU rate alone, yes. CUDO's H100 SXM VM costs $1.79 per GPU-hour against GPUBank's $1.85, but CUDO adds vCPU at $0.0059 per vCPU-hour, RAM at $0.0049 per GiB-hour, disk and IPv4. GPUBank's $1.85 includes vCPU, RAM, NVMe, IPv4 and traffic.
Yes. CUDO's docs say all billable resources keep accruing charges while a VM is stopped or powered off, and resources must be deleted to stop billing. A stopped GPUBank instance costs only its persistent disk, at $0.03 per GB-month.
On the check date, CUDO's public API returned priced capacity in two data centers, Kristiansand in Norway and near Stockholm in Sweden. Marketing pages cite 16 data centers for enterprise clusters without naming them. GPUBank runs Paris, Frankfurt, Ashburn and Singapore.
CUDO does not list crypto. Its payment methods page lists Visa and Mastercard through Stripe and PayPal, with a $10 minimum top-up. GPUBank is paid from a USD balance topped up in Bitcoin, USDT, Ethereum and more than a dozen other coins, from $25.
CUDO. It sells 8-GPU bare-metal nodes and self-serve clusters with InfiniBand, and enterprise clusters by quote. GPUBank rents up to 8 GPUs per instance and does not sell InfiniBand clusters.
Without. Every price on this page excludes VAT and sales tax, for CUDO Compute and for GPUBank, so the two columns compare like for like. Tax is added at checkout according to where the buyer is based.
Sources
Every figure about CUDO Compute was read on its own pages or official documentation. Prices last checked on 29 September 2026.
- 01cudocompute.com/pricingEnterprise GPU list, quote on request
- 02cudocompute.com/Positioning, regions, SOC 2 Type II and ISO 27001 claims, capacity pipeline
- 03cudocompute.com/about-usLeadership, compliance wording, operating regions
- 04cudocompute.com/products/nvidia-h100H100 node specs, 16 ISO-certified data centres, 8 to 1, 000+ GPUs
- 05cudocompute.com/products/nvidia-h200H200 specs
- 06cudocompute.com/products/nvidia-b200B200 specs
- 07cudocompute.com/products/nvidia-b300B300 specs
- 08cudocompute.com/products/nvidia-gb200-nvl72GB200 specs
- 09cudocompute.com/products/nvidia-gb300GB300 specs
- 10cudocompute.com/products/nvidia-vera-rubinVera Rubin specs
- 11cudocompute.com/products/nvidia-b100B100 specs
- 12cudocompute.com/products/clustersCluster scale 128 to 40, 000 GPUs
- 13cudocompute.com/service-level-agreement99.5% uptime for committed resources, service credits
- 14cudocompute.com/terms-of-serviceContract documents, DPA, English law
- 15rest.compute.cudo.org/v1/vms/machine-types-2VM GPU, vCPU, RAM, disk and IPv4 prices
- 16rest.compute.cudo.org/v1/machines-typesBare-metal node prices and specs
- 17rest.compute.cudo.org/v1/clusters/machine-typesCluster node price, InfiniBand
- 18rest.compute.cudo.org/v1/vms/gpu-modelsGPU catalogue and VRAM
What we could not verify (9)
- Commitment (reserved) prices: the API supports purchasing VM commitments but the public machine-types endpoints only return COMMITMENT_TERM_NONE prices
- Enterprise cluster prices for all 8 pricing-page GPUs and B100: quote only
- 13 GPU models listed by the public gpu-models API (MI300X, A40, A800 PCIe, H200 NVL, H200 SXM, RTX 3080, RTX 4000 SFF Ada, RTX A4000, RTX A4000 Ada, RTX A4500, RTX A5000, RTX A6000, V100) had no machine type or price on 2026-09-29; they are counted in catalogue_count
- NFS volume prices: authenticated endpoint only
- Egress price: not published
- Crypto payment: not mentioned in the payment methods page
- HQ city and country, founding year, funding: not stated on the pages read (terms under English law)
- Full list of the '16 ISO-certified data centres' used for enterprise clusters: not published
- SOC 2: home page lists 'SOC 2 Type II' while the About page says 'SOC 2-aligned'; no report seen
In short
GPUBank vs CUDO Compute, in facts.
The essentials of this comparison in one table, for quick readers and AI assistants. Prices last checked on 29 September 2026.
Read as Markdown- H100 SXM, per GPU-hour
- CUDO Compute: $1.79, GPU only on VMs
GPUBank: $1.85, all-in - Cheapest GPU-hour
- CUDO Compute: $0.75, L40S on an 8-GPU node
GPUBank: $0.23, RTX 4090 - Billing granularity
- CUDO Compute: Per second
GPUBank: Per second - Stopped instance
- CUDO Compute: All resources billed until deleted
GPUBank: Disk only - Persistent storage
- CUDO Compute: $0.00014 per GiB-hour, $0.10 per GiB-month
GPUBank: $0.03 per GB-month - Egress traffic
- CUDO Compute: Not published
GPUBank: Included - Crypto payments
- CUDO Compute: Not published
GPUBank: Bitcoin, USDT, Ethereum and 12+ more - Card payments
- CUDO Compute: Visa, Mastercard, PayPal
GPUBank: No - Minimum deposit
- CUDO Compute: $10
GPUBank: $25 - GPU models listed
- CUDO Compute: 24 listed, 4 with a self-serve price
GPUBank: 10 - Regions
- CUDO Compute: 2 self-serve
GPUBank: 4 - Serverless GPUs
- CUDO Compute: No
GPUBank: No - Availability target
- CUDO Compute: 99.5 % on committed resources only
GPUBank: 99.9 % monthly - Compliance reports
- CUDO Compute: SOC 2 Type II and ISO 27001 claimed
GPUBank: None published - Support
- CUDO Compute: Console tickets; 24/7 on enterprise clusters
GPUBank: Tickets from a $25 balance - Headquarters
- CUDO Compute: United Kingdom
GPUBank: Tallinn, Estonia
How this page was made
Every price and policy of CUDO Compute was read on its own website, documentation or public API on 29 September 2026, with the address of each page listed above. A value we could not read on a primary source is marked "Not published" rather than estimated. GPUBank figures come from this site and its terms. Prices change: check both providers before you buy. CUDO Compute, its logo and its website belong to their owner and are shown only to identify the service compared. GPUBank is not affiliated with CUDO Compute.
