GPUBank vs TensorDock
TensorDock is a US GPU marketplace run by TensorDock.com, Inc, a Delaware company, that rents KVM virtual machines on vetted third-party hosts and partner data centers at prices each host sets. Its public pages list the H100 SXM at $2.12 to $2.25 per GPU-hour; the $2.25 table, dated July 2024, covers the GPU alone, with vCPU, RAM and NVMe billed on top. GPUBank charges $1.85 all-in. TensorDock offers Windows VMs, cheaper storage and H100 bare metal on long terms; GPUBank publishes current prices without an account.
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
TensorDock
Lewes, United States
H100 SXM
$2.12 / h
- Full KVM VMs with root access, including Windows 10
- 8x H100 bare metal reserved from $1.50 per GPU-hour on 3 years
- Pay by card from $5; RTX A4000 from $0.10 per GPU-hour
TensorDock
Marketplace
28 points compared24 of their pages readprices last checked on 29 September 2026page updated 2 November 2026
The verdict
Full VMs and long-term H100 reservations at TensorDock. Current all-in prices and more tooling at GPUBank.
TensorDock suits users who want full KVM VMs, including Windows, and are ready to pick hosts and pay a la carte for CPU and RAM. Its live prices appear only after sign-in, and the public tables date from 2024 and conflict between pages. On those pages, GPUBank costs 13 to 42 % less for the GPUs both of us rent, and TensorDock's table prices exclude vCPU, RAM and storage. TensorDock wins on storage price, card payments and 8x H100 bare metal on 1-month to 3-year terms.
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 | TensorDock per hour | GPUBank per hour | Difference | 100 h at TensorDock | 100 h at GPUBank |
|---|---|---|---|---|---|
| $2.12 | $1.85 | −13 % | $212.00 | $185.00 | |
| $1.99 | $1.35 | −32 % | $199.00 | $135.00 | |
| $1.65 | $0.95 | −42 % | $165.00 | $95.00 | |
| $0.35 | $0.23 | −34 % | $35.00 | $23.00 |
The RTX 4090 figure is from a table dated July 2024 that covers the GPU only; the others are benchmark chart labels with no date or inclusions stated. Live prices are set by each host and shown after sign-in.
A fine-tune, 40 h on an H100 SXM
TensorDock
$84.80
GPUBank
$74.00
Saves $10.80 at GPUBank (−13 %)
Image generation, 200 h on an RTX 4090
TensorDock
$70.00
GPUBank
$46.00
Saves $24.00 at GPUBank (−34 %)
A training run, 100 h on an A100 SXM
TensorDock
$165.00
GPUBank
$95.00
Saves $70.00 at GPUBank (−42 %)
02Their catalogue
Every GPU TensorDock lists.
TensorDock lists 14 GPU models. The 13 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).
H100 SXM80 GB · on demand$2.12
H100 PCIe80 GB · on demand$1.99
A100 SXM 80GB80 GB · on demand$1.65
A100 PCIe 80GB80 GB · on demand$1.50
L4048 GB · on demand$0.95
V10016 GB · on demand$0.17
RTX 6000 Ada48 GB · on demand$0.75
RTX A600048 GB · on demand$0.43
RTX A500024 GB · on demand$0.27
RTX A400016 GB · on demand$0.10
Quadro RTX 800048 GB · on demand$0.32
RTX 409024 GB · on demand$0.35
RTX 309024 GB · on demand$0.20At GPUBank
vCPU, RAM, NVMe and traffic included. All prices
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.
TensorDock
2 h 14 m 03 s on H100 SXM
- GPU · $2.12 / h
- $4.74
- vCPU and RAM
- Billed separately
- Storage
- $0.036 per GB-month, host-set
- Egress
- Included, no ingress or egress fees
- Public IPv4
- Port forwarding; dedicated IP in some sites
- Billing granularity
- Per second
- Minimum charge
- None
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.
TensorDock
Storage only; GPU released.
Restart depends on GPU availability on that host.
GPUBank
Only the persistent disk, $0.03 per GB-month. Compute, IPv4 and scratch cost nothing.
04Where the GPUs run
4 countries against 4, and where they are.
Countries where TensorDock 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.
TensorDock4
United States
United Kingdom
Czech Republic
Germany
GPUBank4
Paris
Frankfurt
Ashburn
Singapore
TensorDock · 4 countriesGPUBank · 4 regions
05Paying and signing up
How each one takes your money.
TensorDock
Payments and sign-up
Payment methods
Crypto: Implied; coins not published
Minimum top-up $5
To start
- 1Email and password
- 2Card (3D Secure attempted on every card transaction)
Billing modelPrepaid balance
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 | Not published | Not published |
| ISO 27001 | Not published | Not published |
| HIPAA | Not published | No |
| GDPR DPA | Not published | Published DPA, applies automatically |
| Availability target | No customer SLA; downtime credits | 99.9 % monthly, per region |
| Status page | None found | Public status page |
| Support | Email, live chat, Discord | 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 TensorDock home page captured on 29 September 2026, and GPUBank's live board. Cookie banners and chat bubbles are closed; nothing else is edited.

- 1tensordock.com/pricing returns a 404; the only public price table is marked Last Updated: July 24, 2024, with a pointer to the dashboard.
- 2The deploy page loads signed out but shows Authentication Required instead of available GPUs.
- 3The home page claims 45 GPU models, the GPU pages 27 and the comparison pages 44; public pages name 14.

- 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
28 points, read off both sites.
GPUBank aheadTensorDock aheadEven
| Criterion | GPUBank | |
|---|---|---|
| Price and catalogue | ||
| H100 SXM, per GPU-hour$2.25 is GPU only, from a July 2024 table; $2.12 is a benchmark chart label. Hosts range from $1.90 to $2.50. | $1.85, all-in | $2.12 to $2.25, 2024-era public pages |
| Cheapest GPU-hour | $0.23, RTX 4090 | $0.10, RTX A4000, GPU onlyAhead |
| GPU models listed | 10 | 14 named; 44 to 45 claimedAhead |
| What the GPU price includesvCPU at $0.003 per hour and RAM at $0.002 per GB-hour. | vCPU, RAM, NVMe, IPv4, traffic | GPU only; vCPU, RAM, NVMe extra |
| Billing | ||
| Billing granularity | Per second | Per second |
| Minimum chargeThe API needs a $1 balance and 100 GB of storage per VM. | None | None |
| Stopped instanceRestart depends on GPU availability on that host. | Disk only | Storage only; GPU released |
| Persistent storageStorage can be increased, never decreased. | $0.03 per GB-month | $0.036 per GB-month, host-set |
| Egress traffic | Included | Included, no ingress or egress fees |
| Public IPv4Core Compute locations give a static IP per VM. The price is not published. | Included | Port forwarding; dedicated IP in some sites |
| Billing modelServers are deleted automatically when the balance reaches $0. | Prepaid balance | Prepaid balance |
| Payment and sign-up | ||
| Crypto payments | Bitcoin, USDT, Ethereum and 12+ more | Implied; coins not published |
| Card payments | No | Cards, 3D SecureAhead |
| Minimum deposit | $25 | $5Ahead |
| Platform | ||
| API and CLI | REST API, CLI, Python SDK | REST API; no CLI or SDK documented |
| Terraform | Official provider | Not documented |
| Serverless GPUs | No | No |
| Managed Kubernetes | No | No |
| Ready-made templates | 25 | Ubuntu, Windows and an ML image |
| Interruptible or spotHost minimum bids are typically around 50 % of the on-demand price. | Yes, 40 % off | Spot bidding, in beta |
| GPUs per instance | Up to 8 | Up to 8 per VM; 8x bare metal |
| Multi-node interconnect | Not offered | Not published |
| Network and regions | ||
| RegionsThe live location list requires login. | 4 | 7 named locations; 97+ claimedAhead |
| CountriesIndia hosts CPU locations only. | France, Germany, US, Singapore | US, UK, Czechia, Germany |
| EU regions | Paris, Frankfurt | Prague and Germany |
| Trust and support | ||
| Availability targetCredits of 10x or 5x the hourly rate by host tier, 40x for data loss. | 99.9 % monthly | No customer SLA; downtime credits |
| Compliance reportsPartner data centers are said to hold SOC 2, ISO 27001 and other certifications. | None published | None at company level |
| Public status page | Yes | None found |
| Support24-hour response standard, staff available 24/7 per the about page. | Tickets from a $25 balance | Email, live chat, DiscordAhead |
09The right fit
Different needs. A clear choice.
What TensorDock does better
- Full KVM VMs with root access, including Windows 10
- 8x H100 bare metal reserved from $1.50 per GPU-hour on 3 years
- Pay by card from $5; RTX A4000 from $0.10 per GPU-hour
- Published downtime credits of 5x to 10x the hourly rate
Choose GPUBank if
- H100 SXM at $1.85 all-in, against $2.12 to $2.25 on TensorDock's pages
- Current prices visible without an account
- Public IPv4 included, where TensorDock uses port forwarding by default
- CLI, Python SDK, Terraform provider, 25 templates and a status page
Moving from TensorDock takes an afternoon.
TensorDock VMs are standard KVM images and it charges no egress fees, so moving data out costs nothing on its side. You can also request an export of the VM disk image.
- 1Launch the same GPU on GPUBank with a persistent disk in the region you need.
- 2Copy your data from the TensorDock VM with scp or rclone.
- 3Rebuild the environment from a template or your Docker image, then delete the TensorDock VM.
# 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. TensorDock's public pages list the H100 SXM at $2.12 to $2.25 per GPU-hour. The $2.25 table is dated July 2024 and covers the GPU alone, with vCPU, RAM and NVMe billed on top. GPUBank charges $1.85 with vCPU, RAM, NVMe, IPv4 and traffic included.
Live per-host prices appear only in the dashboard and API after sign-in. The public table is marked Last Updated: July 24, 2024, and prices differ between pages, for example $1.91 and $1.30 for an H100 spot bid. GPUBank publishes its prices without an account.
Its security page implies that cryptocurrency is accepted, but it does not name the coins or the processor. GPUBank takes Bitcoin, USDT, Ethereum and more than a dozen other coins.
Yes. A stopped VM pays only for storage at the host's rate, but its GPUs are released and a restart depends on availability. On GPUBank a stopped instance also costs only its persistent disk, at $0.03 per GB-month.
Yes. It lists 8x H100 SXM hostnodes at $2.00 per GPU-hour on a monthly term, $1.90 annual and $1.50 on 3 years, with an 8-GPU minimum. GPUBank reserves any instance for 1 week or 1 month at 20 % off.
Without. Every price on this page excludes VAT and sales tax, for TensorDock 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 TensorDock was read on its own pages or official documentation. Prices last checked on 29 September 2026.
- 01tensordock.com/Headline prices, billing FAQ, $5 start, location counters, 45 models claim
- 02tensordock.com/cloud-gpus.htmlOn-demand price table, a la carte resource prices, no egress fees
- 03tensordock.com/gpu-h100.htmlH100 on-demand, spot, monthly/annual/3-year and bare metal prices, Dallas location
- 04tensordock.com/gpu-a100.htmlA100 prices, Devon/Prague/Oregon locations
- 05tensordock.com/gpu-4090.html4090 on-demand and spot prices
- 06tensordock.com/benchmarks.htmlPrice labels for 11 GPUs incl. H100 PCIe, V100 32GB, Quadro 8000
- 07tensordock.com/comparison-runpod.htmlEquivalent configuration prices, 5% refund fee, support channels
- 08tensordock.com/comparison-vast.htmlGPU type count, 1 Gbps minimum
- 09tensordock.com/security.html3D Secure, compliance statements
- 10tensordock.com/about.html24/7 staff, same pricing for all, no referral program
- 11tensordock.com/host.htmlHost requirements, 25% revenue-share fee
- 12tensordock.com/cloud-cpus.htmlCPU locations in US, Europe, India
- 13dashboard.tensordock.com/deployPer-second billing, stopped cost estimate, login requirement
- 14dashboard.tensordock.com/api/docs/getting-startedBearer auth, 100 req/min rate limit, base URL
- 15dashboard.tensordock.com/api/docs/fetch-hostnodesAPI structure, max_count 8, per-host pricing fields
- 16dashboard.tensordock.com/api/docs/instance-creationMinimum $1 balance, 100 GB storage minimum, OS images
- 17dashboard.tensordock.com/api/docs/instance-managementStop/start/modify, storage cannot decrease
- 18dashboard.tensordock.com/api/v2/locationsReturns empty list when signed out
What we could not verify (6)
- Live per-host GPU prices and the full current catalogue: dashboard and API require login; public endpoints return empty data
- Dedicated IPv4 price and network storage price
- Whether PayPal or wire/invoice are accepted
- Founding year and funding
- Any CLI, SDK or Terraform provider (only REST API v2 documented; an older Postman v0 collection is linked from the site)
- Whether the 2024-dated price tables are still current
In short
GPUBank vs TensorDock, 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
- TensorDock: $2.12 to $2.25, 2024-era public pages
GPUBank: $1.85, all-in - Cheapest GPU-hour
- TensorDock: $0.10, RTX A4000, GPU only
GPUBank: $0.23, RTX 4090 - Billing granularity
- TensorDock: Per second
GPUBank: Per second - Stopped instance
- TensorDock: Storage only; GPU released
GPUBank: Disk only - Persistent storage
- TensorDock: $0.036 per GB-month, host-set
GPUBank: $0.03 per GB-month - Egress traffic
- TensorDock: Included, no ingress or egress fees
GPUBank: Included - Crypto payments
- TensorDock: Implied; coins not published
GPUBank: Bitcoin, USDT, Ethereum and 12+ more - Card payments
- TensorDock: Cards, 3D Secure
GPUBank: No - Minimum deposit
- TensorDock: $5
GPUBank: $25 - GPU models listed
- TensorDock: 14 named; 44 to 45 claimed
GPUBank: 10 - Regions
- TensorDock: 7 named locations; 97+ claimed
GPUBank: 4 - Serverless GPUs
- TensorDock: No
GPUBank: No - Availability target
- TensorDock: No customer SLA; downtime credits
GPUBank: 99.9 % monthly - Compliance reports
- TensorDock: None at company level
GPUBank: None published - Support
- TensorDock: Email, live chat, Discord
GPUBank: Tickets from a $25 balance - Headquarters
- TensorDock: Lewes, United States
GPUBank: Tallinn, Estonia
How this page was made
Every price and policy of TensorDock 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. TensorDock, its logo and its website belong to their owner and are shown only to identify the service compared. GPUBank is not affiliated with TensorDock.