GPUBank vs CoreWeave
CoreWeave is a US AI cloud from Livingston, New Jersey, that sells bare-metal NVIDIA nodes through managed Kubernetes and Slurm, mostly as 8-GPU HGX nodes and NVL72 racks. Its HGX H100 node costs $49.24 per hour, or $6.155 per GPU; its classic a la carte H100 is $4.76 per GPU-hour with CPU and RAM on top. GPUBank rents a single H100 SXM at $1.85 per GPU-hour all-in, while CoreWeave offers rack-scale Blackwell, InfiniBand clusters and SOC 2 and ISO certifications.
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
CoreWeave
Livingston, United States
H100 SXM
$4.76 / h
- GB200 NVL72, HGX B200 and B300 nodes, 21 GPU models in total
- Managed Kubernetes (CKS) and Slurm (SUNK), control plane free
- InfiniBand at 3,200 Gbps across multi-node clusters
CoreWeave
Enterprise
24 points compared15 of their pages readprices last checked on 29 September 2026page updated 2 November 2026
The verdict
Clusters, Blackwell racks and certifications at CoreWeave. Single GPUs at a fraction of the price at GPUBank.
CoreWeave is built for AI labs and enterprises that run multi-node HGX or GB200 NVL72 clusters on Kubernetes or Slurm, under a sales contract with SOC 2 and ISO paperwork. On the GPUs both of us rent, GPUBank costs 52 to 76 % less per GPU-hour, and it rents one GPU at a time where most CoreWeave on-demand GPUs come as full 8-GPU nodes. CoreWeave onboards through a sales form and bills monthly by card or wire; GPUBank opens with an email and password and takes crypto. Both charge nothing for egress.
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 | CoreWeave per hour | GPUBank per hour | Difference | 100 h at CoreWeave | 100 h at GPUBank |
|---|---|---|---|---|---|
| $8.60 | $4.15 | −52 % | $860.00 | $415.00 | |
| $6.31 | $2.49 | −61 % | $631.00 | $249.00 | |
| $4.76 | $1.85 | −61 % | $476.00 | $185.00 | |
| $4.25 | $1.35 | −68 % | $425.00 | $135.00 | |
| $2.50 | $1.15 | −54 % | $250.00 | $115.00 | |
| $2.21 | $0.95 | −57 % | $221.00 | $95.00 | |
| $2.25 | $0.55 | −76 % | $225.00 | $55.00 |
CoreWeave's current prices are per 8-GPU node, divided by 8 here. The H100 SXM, H100 PCIe and A100 figures are classic a la carte GPU prices with vCPU and RAM on top; the HGX H100 node is $6.155 per GPU.
A fine-tune, 40 h on an H100 SXM
CoreWeave
$190.40
GPUBank
$74.00
Saves $116.40 at GPUBank (−61 %)
An inference box, a month, 720 h, on an L40S
CoreWeave
$1,620.00
GPUBank
$396.00
Saves $1,224.00 at GPUBank (−76 %)
A long-context job, 40 h on an H200
CoreWeave
$252.00
GPUBank
$99.60
Saves $152.40 at GPUBank (−60 %)
02Their catalogue
Every GPU CoreWeave lists.
CoreWeave lists 21 GPU models. The 20 below are the ones recorded in our check, each at its lowest public price per GPU-hour. 7 of them are also rented at GPUBank (in blue).
- Blackwell and Hopper7
GB200186 GB · on demand$10.50
B300270 GB · spot$4.48
B200180 GB · on demand$8.60
H200141 GB · on demand$6.31
GH20096 GB · on demand$6.50
H100 SXM80 GB · on demand$4.76
H100 PCIe80 GB · on demand$4.25
- Ampere, Ada and older data-centre7
A100 SXM 80GB80 GB · on demand$2.21
A100 PCIe 80GB80 GB · on demand$2.21- A100 40GB40 GB · on demand$2.06
L40S48 GB · on demand$2.25
L4048 GB · on demand$1.25
A4048 GB · on demand$1.28
V10016 GB · on demand$0.80
- Workstation6
RTX PRO 600096 GB · on demand$2.50
RTX A600048 GB · on demand$1.28
RTX A500024 GB · on demand$0.77
RTX A400016 GB · on demand$0.61
Quadro RTX 500016 GB · on demand$0.57
Quadro RTX 40008 GB · on demand$0.24
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.
CoreWeave
2 h 14 m 03 s on H100 SXM
- GPU · $4.76 / h
- $10.63
- vCPU and RAM
- Included
- Storage
- $0.07 per GB-month (file); object $0.06 Hot
- Egress
- Free
- Public IPv4
- $4.00 per public IP per month
- Billing granularity
- Not published
- 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.
CoreWeave
Not published.
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 CoreWeave 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.
CoreWeave2
United States
Canada
GPUBank4
Paris
Frankfurt
Ashburn
Singapore
CoreWeave · 2 countriesGPUBank · 4 regions
05Paying and signing up
How each one takes your money.
CoreWeave
Payments and sign-up
Payment methods
Minimum top-up Not published
To start
- 1Business email
- 2Company
- 3Sales contact form
Billing modelPostpaid monthly, in arrears
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.
| GPUBank | ||
|---|---|---|
| SOC 2 | Type II | Not published |
| ISO 27001 | Listed, report not public | Not published |
| HIPAA | Not published | No |
| GDPR DPA | DPA available | Published DPA, applies automatically |
| Availability target | SLO 99.9 % multi-region, 99 % single region | 99.9 % monthly, per region |
| Status page | status.coreweave.com | Public status page |
| Support | Email and tickets; 24/7 stated | 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 CoreWeave 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 shows hourly prices per 8-GPU node for North America and Europe, without an account.
- 2A per-GPU 'Inference Single GPU Price' column applies only to inference platform customers.
- 3Reserved capacity is described only as 'up to 60% discounts', and docs for regions and billing ask for an access code.

- 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
24 points, read off both sites.
GPUBank aheadCoreWeave aheadEven
| Criterion | GPUBank | |
|---|---|---|
| Price and catalogue | ||
| H100 SXM, per GPU-hour | $1.85, all-in | $4.76 GPU only (classic); $6.155 on an 8-GPU node |
| Cheapest GPU-hourvCPU and RAM are billed on top of classic GPU prices. | $0.23, RTX 4090 | $0.24, Quadro RTX 4000, GPU only (classic) |
| GPU models listed | 10 | 21Ahead |
| What the GPU price includesCurrent node prices include CPU, RAM and local storage; classic a la carte prices do not. | vCPU, RAM, NVMe, IPv4, traffic | Nodes: vCPU, RAM, local NVMe; classic: GPU only |
| Billing | ||
| Billing granularityPrices are shown per hour; the billing docs need an access code. | Per second | Not published |
| Minimum charge | None | Not published |
| Stopped instance | Disk only | Not published |
| Persistent storage | $0.03 per GB-month | $0.07 per GB-month (file); object $0.06 Hot |
| Egress traffic | Included | Free |
| Public IPv4 | Included | $4.00 per public IP per month |
| Billing model | Prepaid balance | Postpaid monthly, in arrears |
| Payment and sign-up | ||
| Crypto paymentsThe Terms name credit card and ACH or wire; PayPal and crypto are not mentioned. | Bitcoin, USDT, Ethereum and 12+ more | Not listed: card, ACH or wire |
| Card payments | No | Credit card; ACH or wire above 20,000 USDAhead |
| Minimum deposit | $25 | Not published |
| Platform | ||
| API and CLI | REST API, CLI, Python SDK | API with OpenAPI specs; CLI not verified |
| Terraform | Official provider | Official provider |
| Serverless GPUs | No | Serverless and dedicated inferenceAhead |
| Managed Kubernetes | No | Managed (CKS) on bare metal, plus SlurmAhead |
| Ready-made templates | 25 | Not published |
| Interruptible or spotSpot HGX H100 is $2.46 per GPU-hour; GPUBank's interruptible H100 SXM is $1.11. | Yes, 40 % off | Spot node pools, 8-GPU nodes |
| GPUs per instance | Up to 8 | 8-GPU nodes, NVL72 racks, multi-node clustersAhead |
| Multi-node interconnect | Not offered | InfiniBand, 3,200 GbpsAhead |
| Network and regions | ||
| Regions | 4 | 39 zone codes on the status pageAhead |
| Countries | France, Germany, US, Singapore | US, Canada; EU countries not published |
| EU regionsCoreWeave lists more EU zones; GPUBank names its cities, Paris and Frankfurt. | Paris, Frankfurt | 5 EU zones, locations not published |
| Trust and support | ||
| Availability targetCredits of 10 to 50 % of the bill, Spot excluded. GPUBank targets 99.9 % per region. | 99.9 % monthly | SLO 99.9 % multi-region, 99 % single region |
| Compliance reports | None published | SOC 2 Type II, ISO 27001, 27017, 27018Ahead |
| Public status page | Yes | status.coreweave.com |
| Support | Tickets from a $25 balance | Email and tickets; 24/7 statedAhead |
09The right fit
Different needs. A clear choice.
What CoreWeave does better
- GB200 NVL72, HGX B200 and B300 nodes, 21 GPU models in total
- Managed Kubernetes (CKS) and Slurm (SUNK), control plane free
- InfiniBand at 3,200 Gbps across multi-node clusters
- SLO with service credits, SOC 2 Type II and ISO 27001
- Spot HGX H100 at $2.46 per GPU, card or wire invoicing
Choose GPUBank if
- H100 SXM at $1.85 per GPU-hour, against $4.76 to $6.155
- Rent a single GPU, not an 8-GPU node at $49.24 per hour
- Self-serve sign-up with email and password, no sales form
- Pay in Bitcoin, USDT and 12+ other coins, prepaid from $25
- IPv4 included, where CoreWeave charges $4.00 per IP per month
Moving from CoreWeave takes an afternoon.
CoreWeave charges no egress and its object storage is S3-compatible, so exporting data costs nothing on its side. Workloads already run as containers on standard Kubernetes.
- 1Launch the GPUs you need on GPUBank, up to 8 per instance, with a persistent disk.
- 2Copy data from AI Object Storage with rclone or the S3 CLI, and file data with rsync.
- 3Run the same container images with Docker, then scale down CoreWeave node pools.
# 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.
Yes. CoreWeave's HGX H100 node costs $49.24 per hour, $6.155 per GPU, and its classic a la carte H100 is $4.76 per GPU-hour before CPU and RAM. GPUBank charges $1.85 per GPU-hour for an H100 SXM, with vCPU, RAM, NVMe, IPv4 and traffic included.
On demand, CoreWeave sells most GPUs as 8-GPU nodes; the GH200 is the exception. Per-GPU inference prices are limited to inference platform customers. GPUBank rents from one GPU up to 8 per instance.
CoreWeave's Terms name credit card and ACH or wire, with invoices of 20,000 USD or more paid by ACH or wire on Net 15. Crypto is not mentioned. GPUBank is paid from a USD balance topped up in Bitcoin, USDT, Ethereum and more than a dozen other coins.
CoreWeave. It runs HGX B200, H200 and H100 nodes and GB200 NVL72 racks on InfiniBand at 3,200 Gbps, with managed Kubernetes and Slurm. GPUBank rents up to 8 GPUs per instance and does not sell clusters.
There is no public self-serve sign-up page. The contact form asks for a business email, company name, job title and country, and access follows a sales contract. GPUBank opens an account with an email and password.
Without. Every price on this page excludes VAT and sales tax, for CoreWeave 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 CoreWeave was read on its own pages or official documentation. Prices last checked on 29 September 2026.
- 01coreweave.com/pricingGPU and CPU on-demand, spot and inference prices, storage, networking, Kubernetes prices, reserved discount
- 02coreweave.com/pricing/classicLegacy a la carte GPU component prices, classic storage prices
- 03docs.coreweave.com/policies/terms-of-serviceSLO and credits, payment terms (card, ACH/wire, Net 15, monthly in arrears), DPA, sanctions, data deletion
- 04docs.coreweave.com/policies/spot-tosSpot preemption, SLO exclusion, legal entity
- 05status.coreweave.com/Region and zone codes, status page
- 06trust.coreweave.com/SOC 2 Type II, ISO 27001/27017/27018 badges
- 07coreweave.com/securitySecurity practices, single-tenant nodes, SOC 2 and ISO 27001 alignment
- 08registry.terraform.io/providers/coreweave/coreweave/latest/docsOfficial Terraform provider
- 09docs.coreweave.com/Products (CKS, SUNK, Inference, Sandbox), docs access restrictions
- 10docs.coreweave.com/changelog/release-notes/spot-node-poolsSpot and capacity plans (Flex Reservations, Reserved Instances, Spot, On-Demand)
- 11docs.coreweave.com/support/platform/articles/how-do-i-choose-the-right-instance-typeCapacity plans, instance availability matrix
- 12coreweave.com/products/gpu-computeGPU catalogue, bare metal, InfiniBand 3200 Gbps
- 13coreweave.com/about-usHQ address, leadership
- 14coreweave.com/contact-usSales-led onboarding form fields
- 15coreweave.com/24/7 support claim, positioning
What we could not verify (10)
- Billing granularity (per hour vs per second or minute) not stated on public pages; the billing docs need an access code. 'per hour' reflects the published price unit only.
- Stopped-instance billing, minimum charge and free credits not published publicly.
- Region-to-city and EU region countries: 'Regions and Availability Zones' docs page needs an access code; region codes taken from the status page.
- Reserved pricing: only 'up to 60% discounts' published, no per-term prices.
- GB300 NVL72 on-demand and spot, HGX B300 on-demand and RTX PRO 6000 Standard Memory on-demand: 'Contact sales', no price rows recorded.
- A100 form factor (SXM vs PCIe) on the current pricing page not stated, recorded as model 'other'.
- Whether classic a la carte pricing is still orderable by new customers not stated.
- CLI, SDK, Jupyter and SSH access not verified on public pages.
- Founding year and listing or funding status not read on an accessible primary page (investor site blocked).
- Payment methods: Terms mention credit card and ACH/wire; PayPal and crypto not mentioned.
In short
GPUBank vs CoreWeave, in facts.
The essentials of this comparison in one table, for quick readers and AI assistants. Prices last checked on 29 September 2026.
- H100 SXM, per GPU-hour
- CoreWeave: $4.76 GPU only (classic); $6.155 on an 8-GPU node
GPUBank: $1.85, all-in - Cheapest GPU-hour
- CoreWeave: $0.24, Quadro RTX 4000, GPU only (classic)
GPUBank: $0.23, RTX 4090 - Billing granularity
- CoreWeave: Not published
GPUBank: Per second - Stopped instance
- CoreWeave: Not published
GPUBank: Disk only - Persistent storage
- CoreWeave: $0.07 per GB-month (file); object $0.06 Hot
GPUBank: $0.03 per GB-month - Egress traffic
- CoreWeave: Free
GPUBank: Included - Crypto payments
- CoreWeave: Not listed: card, ACH or wire
GPUBank: Bitcoin, USDT, Ethereum and 12+ more - Card payments
- CoreWeave: Credit card; ACH or wire above 20,000 USD
GPUBank: No - Minimum deposit
- CoreWeave: Not published
GPUBank: $25 - GPU models listed
- CoreWeave: 21
GPUBank: 10 - Regions
- CoreWeave: 39 zone codes on the status page
GPUBank: 4 - Serverless GPUs
- CoreWeave: Serverless and dedicated inference
GPUBank: No - Availability target
- CoreWeave: SLO 99.9 % multi-region, 99 % single region
GPUBank: 99.9 % monthly - Compliance reports
- CoreWeave: SOC 2 Type II, ISO 27001, 27017, 27018
GPUBank: None published - Support
- CoreWeave: Email and tickets; 24/7 stated
GPUBank: Tickets from a $25 balance - Headquarters
- CoreWeave: Livingston, United States
GPUBank: Tallinn, Estonia
How this page was made
Every price and policy of CoreWeave 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. CoreWeave, its logo and its website belong to their owner and are shown only to identify the service compared. GPUBank is not affiliated with CoreWeave.