GPUBank vs Fluidstack
Fluidstack is a New York-headquartered, UK-incorporated AI infrastructure company that builds and runs large dedicated GPU clusters for frontier AI labs under contract, with Anthropic as its named anchor customer. It publishes no GPU prices: its pricing page returns a 404 and rates are set in each customer Order. GPUBank rents GPUs self-serve by the second at published rates, for example an H100 SXM at $1.85 per GPU-hour, so the two serve different buyers.
GPUBank
Tallinn, Estonia
- vCPU, RAM, NVMe, IPv4 and traffic included
- Per-second billing after the health check
- Top up in Bitcoin, USDT and 12+ other coins
Fluidstack
New York, United States
Shared GPUs
It rents none of our 10 models.
- Dedicated multi-thousand-GPU clusters under a single contract
- Managed Kubernetes and managed Slurm on bare metal
- GPU health checks and auto-remediation in its Lighthouse suite
Fluidstack
Enterprise
18 points compared25 of their pages readprices last checked on 29 September 2026page updated 2 November 2026
The verdict
Dedicated clusters at data-center scale at Fluidstack. Published prices and self-serve access at GPUBank.
Fluidstack sells capacity by contract: managed Kubernetes and Slurm on bare metal, a Technical Account Manager per customer, and campuses planned up to 1.5 GW. It is not a place to rent one GPU for an afternoon; there is no public price list and clusters are provisioned through support. GPUBank publishes all-in hourly rates, bills by the second and needs only an email and password. Price criteria below are marked not published because Fluidstack does not disclose them.
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.
Fluidstack publishes no price for any of the 10 GPU models GPUBank rents: its prices are quoted per contract. See the spec sheet below for how it bills.
02What the price includes
What appears on the bill besides the GPU.
The same session of 2 h 14 m 03 s, line by line, as each provider describes its billing.
Fluidstack
2 h 14 m 03 s
- vCPU and RAM
- Not published
- Storage
- Not published
- Egress
- Not published
- Public IPv4
- Not published
- Billing granularity
- Not published
- Minimum charge
- Not published
GPUBank
2 h 14 m 03 s
- 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.
Fluidstack
Not published.
GPUBank
Only the persistent disk, $0.03 per GB-month. Compute, IPv4 and scratch cost nothing.
03Where the GPUs run
1 country against 4, and where they are.
Countries where Fluidstack 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.
Fluidstack1
United States
GPUBank4
Paris
Frankfurt
Ashburn
Singapore
Fluidstack · 1 countryGPUBank · 4 regions
04Paying and signing up
How each one takes your money.
Fluidstack
Payments and sign-up
Payment methods
Minimum top-up Not published
To start
- 1Email (or Google login)
- 2Signed Order naming the customer individual
- 3Customer sanctions/export-control representations
Billing modelPrepaid per Order, non-refundable
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
05Trust 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 | DPA available | Published DPA, applies automatically |
| Availability target | 99% monthly network availability | 99.9 % monthly, per region |
| Status page | status.fluidstack.io | Public status page |
| Support | Technical Account Manager per customer | Tickets from a $25 balance |
06Their site, our site
What you see before you pay.
Both sites as a visitor sees them, without an account: the Fluidstack home page captured on 29 September 2026, and GPUBank's live board. Cookie banners and chat bubbles are closed; nothing else is edited.

- 1The home page is a mission and recruiting page and does not describe a GPU product or prices.
- 2The pricing URL returns a 404 page.
- 3Product docs live on docs.fluidstack.io, and the console sits behind a login with email or Google.

- 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.
07The spec sheet
18 points, read off both sites.
GPUBank aheadFluidstack aheadEven
| Criterion | GPUBank | |
|---|---|---|
| Price and catalogue | ||
| H100 SXM, per GPU-hourPricing is set per customer Order. The pricing page returns a 404. | $1.85, all-in | Not published |
| Cheapest GPU-hour | $0.23, RTX 4090 | Not published |
| GPU models listedOther models offered under contract are not listed publicly. | 10 | 1 documented, H100 SXM5 in an API example |
| What the GPU price includes | vCPU, RAM, NVMe, IPv4, traffic | Not published |
| Billing | ||
| Billing granularity | Per second | Not published |
| Minimum charge | None | Not published |
| Stopped instance | Disk only | Not published |
| Persistent storage | $0.03 per GB-month | Not published |
| Egress traffic | Included | Not published |
| Public IPv4 | Included | Not published |
| Billing modelRenewals may rise by 3 % compounded or CPI, with 30 days notice. | Prepaid balance | Prepaid per Order, non-refundable |
| Payment and sign-up | ||
| Crypto payments | Bitcoin, USDT, Ethereum and 12+ more | Not mentioned; terms name bank transfer, card |
| Card payments | No | Cards via Stripe, bank transferAhead |
| Minimum deposit | $25 | Not published |
| Platform | ||
| API and CLI | REST API, CLI, Python SDK | Infrastructure and Management APIs, fluidctl |
| Terraform | Official provider | None found |
| Serverless GPUs | No | No |
| Managed KubernetesProvisioned through support, not through the API. | No | Managed, on bare metalAhead |
| Ready-made templates | 25 | None documented |
| Interruptible or spot | Yes, 40 % off | Preemptible flag, no price published |
| GPUs per instance | Up to 8 | 8 per instance; dedicated large clustersAhead |
| Multi-node interconnect | Not offered | NVLink and NVSwitch; fabric not published |
| Network and regions | ||
| RegionsRegion names are visible to customers only. Announced sites are in New York State and Texas. | 4 | Not published |
| Countries | France, Germany, US, Singapore | United States |
| EU regions | Paris, Frankfurt | None announced |
| Trust and support | ||
| Availability targetService credits apply only where an Order provides them. | 99.9 % monthly | 99% monthly network availability |
| Compliance reportsNo SOC 2, ISO 27001 or HIPAA status is named publicly. | None published | GDPR DPA; certifications on request |
| Public status page | Yes | status.fluidstack.io |
| Support | Tickets from a $25 balance | Technical Account Manager per customerAhead |
08The right fit
Different needs. A clear choice.
What Fluidstack does better
- Dedicated multi-thousand-GPU clusters under a single contract
- Managed Kubernetes and managed Slurm on bare metal
- GPU health checks and auto-remediation in its Lighthouse suite
- Enterprise SSO and a named Technical Account Manager
- Card or bank transfer payment and a published GDPR DPA
Choose GPUBank if
- Published all-in rates, such as H100 SXM at $1.85 per GPU-hour
- Self-serve sign-up with an email and password, no contract
- Per-second billing with a published $0.03 per GB-month stopped rate
- 10 public GPU models in France, Germany, the US and Singapore
- Crypto top-ups from $25, a Terraform provider and templates
Moving from Fluidstack takes an afternoon.
Fluidstack workloads run as containers on managed Kubernetes or as Slurm jobs with Pyxis and Enroot, so images and job scripts are portable. Customer data may be deleted on termination, so move it before the contract ends.
- 1Copy datasets and checkpoints out of the cluster with rsync or an S3 client before the Order ends.
- 2Launch GPUBank instances of up to 8 GPUs with persistent disks in the regions you need.
- 3Run the same container images on GPUBank and replace Slurm or Kubernetes job specs with your own scheduler or scripts.
# 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.
Fluidstack does not publish it. Its pricing page returns a 404, and rates are set in each customer Order. GPUBank publishes its H100 SXM at $1.85 per GPU-hour on demand, with vCPU, RAM, NVMe, IPv4 and traffic included.
Not self-serve. Fluidstack sells dedicated capacity through contracts, and its Kubernetes and Slurm clusters are provisioned through support. GPUBank lets anyone with an email and password launch 1 to 8 GPUs and pay by the second.
Fluidstack. It builds dedicated clusters for AI labs, including custom data centers for Anthropic and a campus in Cameron County, Texas planned up to 1.5 GW, with managed Kubernetes or Slurm. GPUBank rents up to 8 GPUs per instance and does not sell clusters.
Crypto is not mentioned. Its terms name bank transfer in USD and credit card via Stripe, with fees paid in advance and non-refundable. GPUBank is topped up in Bitcoin, USDT, Ethereum and more than a dozen other coins, from $25.
Fluidstack does not publish a region list. Its announced sites are in New York State and Texas. GPUBank runs in Paris, Frankfurt, Ashburn and Singapore.
Without. Every price on this page excludes VAT and sales tax, for Fluidstack 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 Fluidstack was read on its own pages or official documentation. Prices last checked on 29 September 2026.
- 01fluidstack.io/Positioning, site observations
- 02fluidstack.io/pricingProof that no public pricing page exists
- 03fluidstack.io/legal/terms-conditionsLegal entity, HQ address, fees and payment (USD, advance payment, deposits, Stripe cards, bank account), SLA 99%, termination
- 04fluidstack.io/legal/terms-of-useCompany number, VAT number
- 05fluidstack.io/legal/data-protectionGDPR DPA, certifications on request
- 06fluidstack.io/legal/non-circumventionMigration constraints
- 07fluidstack.io/legal/acceptable-useUsage restrictions
- 08fluidstack.io/news/series-a-announcementFunding
- 09fluidstack.io/news/fluidstack-expands-u-s-investment-by-relocating-global-headquarters-to-new-york-cityHQ, offices in Austin and San Francisco
- 10fluidstack.io/news/fluidstack-selected-by-anthropic-to-deliver-custom-data-centers-in-the-usCustomer, sites in New York and Texas
- 11fluidstack.io/news/fluidstack-awarded-clustermax-gold-by-semianalysisSelf-reported rating
- 12fluidstack.io/community/building-compute-in-cameron-countyData center site
- 13docs.fluidstack.io/Managed Kubernetes, managed Slurm, bare metal, GPU/Network Operator
- 14docs.fluidstack.io/kubernetes/overviewProvisioning only via support
- 15docs.fluidstack.io/slurm/overviewPyxis/Enroot, provisioning via support
- 16docs.fluidstack.io/lighthouse/overviewHealth checks, NVLink/NVSwitch monitoring, auto-remediation preview
- 17docs.fluidstack.io/getting-started/clifluidctl install
- 18docs.fluidstack.io/getting-started/ssoSSO providers, TAM
What we could not verify (9)
- All GPU prices (on-demand, preemptible, reserved): not published on fluidstack.io, docs.fluidstack.io or any public official page on 2026-09-29; pricing is set in customer Orders.
- Full GPU catalogue: only H100 SXM5 80GB appears (as the h100.8x example in the Infrastructure API reference); other models offered under contract are not listed publicly. catalogue_count=1 reflects only this documented example.
- Billing granularity actually applied, storage products and prices, egress, IPv4, stopped-instance charges: not published.
- Crypto payments: Terms mention bank transfer in USD and credit card via Stripe; crypto is not mentioned.
- Named region list: the Management API exposes GET /regions but region names are not published; only announced data center sites (New York State, Texas, Cameron County) are public.
- SOC 2 / ISO 27001 / HIPAA status: not published; DPA offers certifications on request.
- Founding year: not stated on the official pages read.
- Terraform provider: none found in docs or the fluidstackio GitHub organisation (no provider repository listed on 2026-09-29).
- Persistent volumes, snapshots, Jupyter: not documented.
In short
GPUBank vs Fluidstack, 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
- Fluidstack: Not published
GPUBank: $1.85, all-in - Cheapest GPU-hour
- Fluidstack: Not published
GPUBank: $0.23, RTX 4090 - Billing granularity
- Fluidstack: Not published
GPUBank: Per second - Stopped instance
- Fluidstack: Not published
GPUBank: Disk only - Persistent storage
- Fluidstack: Not published
GPUBank: $0.03 per GB-month - Egress traffic
- Fluidstack: Not published
GPUBank: Included - Crypto payments
- Fluidstack: Not mentioned; terms name bank transfer, card
GPUBank: Bitcoin, USDT, Ethereum and 12+ more - Card payments
- Fluidstack: Cards via Stripe, bank transfer
GPUBank: No - Minimum deposit
- Fluidstack: Not published
GPUBank: $25 - GPU models listed
- Fluidstack: 1 documented, H100 SXM5 in an API example
GPUBank: 10 - Regions
- Fluidstack: Not published
GPUBank: 4 - Serverless GPUs
- Fluidstack: No
GPUBank: No - Availability target
- Fluidstack: 99% monthly network availability
GPUBank: 99.9 % monthly - Compliance reports
- Fluidstack: GDPR DPA; certifications on request
GPUBank: None published - Support
- Fluidstack: Technical Account Manager per customer
GPUBank: Tickets from a $25 balance - Headquarters
- Fluidstack: New York, United States
GPUBank: Tallinn, Estonia
How this page was made
Every price and policy of Fluidstack 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. Fluidstack, its logo and its website belong to their owner and are shown only to identify the service compared. GPUBank is not affiliated with Fluidstack.