CUDOS (CUDOS Intercloud) review

Updated September 28, 2026 · DeployCue Editorial Team · Facts checked September 26, 2026

We earn a commission when you sign up through links on this site. Partner status can affect where a brand appears in our rankings.

Our verdict

CUDOS runs a focused GPU marketplace for developers and small teams that want current NVIDIA datacentre hardware without a long commitment. The catalogue spans NVIDIA B300, NVIDIA GB300 NVL72 and NVIDIA VR200 (Vera Rubin), billing granularity is hourly, and the ASI Mini model is free to try in the ASI family.

Key facts

Best for
Free entry GPU access for developers and small teams needing current NVIDIA datacentre cards.
GPU models
NVIDIA B300, NVIDIA GB300 NVL72, NVIDIA VR200 (Vera Rubin), NVIDIA H100 PCIe, NVIDIA H100 SXM, NVIDIA H200, NVIDIA HGX B200, NVIDIA GB200 NVL72 (Source, checked September 26, 2026)
Billing granularity
Hourly (Source, checked September 26, 2026)
Free credits
Free access to the ASI:1 Mini model (Source, checked September 26, 2026)
Legal entity
Cudos Ltd (Source, checked September 26, 2026)
Headquarters
United Kingdom (Source, checked September 26, 2026)
Founded
2019 (Source, checked September 26, 2026)
Support channels
Email, Help-centre, Community (Source, checked September 26, 2026)
Pricing model
Hourly billing, Free tier
Official website
cudos.org

VPS

Billing models
Hourly (Source, checked September 26, 2026)
Free credits
Free access to the ASI:1 Mini model (Source, checked September 26, 2026)
Server types
VPS (Source, checked September 26, 2026)
Legal entity
Cudos Ltd (Source, checked September 26, 2026)
Headquarters
United Kingdom (Source, checked September 26, 2026)
Founded
2019 (Source, checked September 26, 2026)
Support channels
Email, Help-centre, Community (Source, checked September 26, 2026)
Pricing model
Hourly billing, Free tier

Serverless

Legal entity
Cudos Ltd (Source, checked September 26, 2026)
Headquarters
United Kingdom (Source, checked September 26, 2026)
Founded
2019 (Source, checked September 26, 2026)
Support channels
Email, Help-centre, Community (Source, checked September 26, 2026)

Apps and profiles

Pros and cons

Pros

  • The catalogue in NVIDIA B300, NVIDIA GB300 NVL72 and NVIDIA VR200 (Vera Rubin) covers datacentre parts from the Hopper and Blackwell generations rather than consumer cards.
  • Billing models are hourly, so a job that ends stops the meter rather than running to a reserved term (hourly).
  • Access pairs hourly pricing with a free tier (hourly and free tier), and Free access to the ASI:1 Mini model covers a Mini model in the ASI family.
  • Newer rack scale Blackwell parts sit in the same list as the older Hopper cards, so a proof of concept and a long production run can share one catalogue (NVIDIA B300, NVIDIA GB300 NVL72 and NVIDIA VR200 (Vera Rubin)).

Cons

  • Support runs on email, help centre and community, covering email, a help centre and community, with no chat line or phone number.
  • Billing granularity is hourly rather than per second, so a short burst still counts a whole unit.
  • Server types cover virtual machines only (vps), so bare metal is not on the list.

Ready to try CUDOS (CUDOS Intercloud)?

Scored 9.2 / 10 by DeployCue.

Who it suits

  • Developers who need a current NVIDIA datacentre card for a single job instead of a long reservation.
  • Small teams testing a pipeline on the ASI Mini model before committing spend.
  • Anyone who prefers hourly billing to a reserved term and wants rack scale parts.

Who should look elsewhere

  • SberCloud shoppers who want older card generations, since B300, H100 and A100 reaches further back.
  • ComputePrices.com readers who want the widest catalogue, including consumer and AMD parts listed in H100 SXM, A100 SXM and H200.
  • Salad users after community supplied consumer cards, since NVIDIA RTX 3060, NVIDIA RTX 3070 and NVIDIA RTX 3070 Ti centres on those rather than datacentre racks.

Frequently asked questions

How does CUDOS meter GPU billing?+

CUDOS meters at hourly granularity, and hourly lists hourly as the only billing model. That suits jobs that stay up for a stretch, such as fine tuning or a serving node. A short burst still counts a whole unit, since per second metering is not part of the plan list.

Which GPUs can I rent on CUDOS?+

The list in NVIDIA B300, NVIDIA GB300 NVL72 and NVIDIA VR200 (Vera Rubin) runs from the Hopper generation cards to the newest Blackwell parts, including rack scale systems and the Vera Rubin card. Newer parts sit alongside the older cards, so a short proof of concept and a long production run can share one catalogue.

Is there a free tier or free credits?+

Yes. Free access to the ASI:1 Mini model covers a Mini model in the ASI family, and hourly and free tier lists free tier access next to hourly pricing. That gives a new account something to run before any spend, which suits a quick benchmark or a pipeline test.

What support do I get?+

Support runs on email, help centre and community, covering email, a help centre and community. That set carries no chat line or phone number, so a blocked job may wait for a written reply. The help centre holds the setup guides, with community threads for anything more specific.

Which server types are available?+

vps lists virtual machines, so the platform is aimed at virtual machines rather than bare metal, and bare metal servers are not part of that list. GPU work still runs on those virtual machines under the same hourly billing.

Who runs CUDOS and where is it based?+

The service runs as CUDOS Ltd, with its home country given as GB, and it was founded in 2019. Contracts, privacy terms and support requests all sit with that legal entity, which matters for teams that need a named counterparty on their invoices.

How does CUDOS compare with Salad for cheaper cards?+

Salad lists community supplied cards in NVIDIA RTX 3060, NVIDIA RTX 3070 and NVIDIA RTX 3070 Ti, which suits light inference and hobby work. CUDOS lists datacentre parts instead, so a team that needs a rack scale Blackwell system pays for a different class of machine rather than a different bill for the same one.

How does CUDOS compare with SberCloud?+

SberCloud covers B300, H100 and A100, reaching older generations of card that the CUDOS list does not carry. CUDOS in turn holds the newer rack scale parts, so the two catalogues lean in opposite directions on generation and suit different workloads.

Sources

  • cudos.org checked September 26, 2026

    Backs: Billing granularity, Founded, Free credits, Legal entity, Pricing model

    “Access high performance GPUs per hour ... spin up GPUs by the hour”

  • cudos.org/legal/privacy-policy checked September 26, 2026

    Backs: Headquarters

    “Means Cudo Ventures Ltd which is a company registered in England and Wales under company number 11065412”

  • docs.cudos.org/docs/asi-cloud/introduction/overview checked September 26, 2026

    Backs: Billing models, Server types

    “GPU Cloud: Per-hour VM/GPU rentals, private networks and pre-set VM templates”

  • docs.cudos.org/docs/welcome checked September 26, 2026

    Backs: Support channels

    “Support ... Discord ... Twitter ... Telegram”

  • cudocompute.com/pricing checked September 26, 2026

    Backs: GPU models

    “H100 PCIe, H100 SXM, H200, HGX B200, GB200 NVL72, VR200, B300, GB300 NVL72 - Quote on request”

Compiled from 5 source pages, checked September 26, 2026. We did not open an account.

Not confirmed yet: GPU hour from, Egress fee.

CUDOS (CUDOS Intercloud)

Free entry GPU access for developers and small teams needing current NVIDIA datacentre cards.

9.2/10See alternatives