Thunder Compute 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. Partners with a live offer in your country come first, then every other brand by score.

Our verdict

Thunder Compute suits developers and small teams who want an hourly GPU at $0.35 and pay nothing to move data out, No data egress fees. Billing is per minute, so short jobs stop costing a full hour, and support runs on help centre and community.

Key facts

Best for
Short, bursty GPU jobs where keeping running cost down matters
GPU models
RTX A6000, L40, A100 80GB, H100 PCIe↗
GPU hour from
$0.35↗
Billing granularity
Per minute↗
Egress fee
No data egress fees↗
Interconnect
7-10 Gbps networking↗
Support channels
Help-centre, Community
Pricing model
Hourly billing
Official website
thundercompute.com

Facts checked September 26, 2026, 2 sources

Apps and profiles

Pricing

  • Thunder Compute NVIDIA L40 (USD)

    compute: $0.79 / hour

  • Thunder Compute NVIDIA H100 (USD)

    compute: $2.95 / hour

  • Thunder Compute NVIDIA RTX A6000 (USD)

    compute: $0.35 / hour

Pros and cons

Pros

  • GPU hour from $0.35, under the site median of €0.63
  • per minute billing, so a job that ends early is not billed for a full hour
  • No data egress fees, which matters once a trained model starts shipping results out
  • Interconnect of 7-10 Gbps networking keeps data moving between nodes

Cons

  • The listed range covers RTX A6000, L40 and A100 80GB, a shorter hardware list than the wider catalogues
  • Support runs on help centre and community only, with no live channel in the list
  • Pricing is hourly, so a fixed monthly rate is not part of the model
Thunder Compute

Ready to try Thunder Compute?

Scored 9.1 / 10 by DeployCue.

Who it suits

  • Developers running short training or inference jobs that start and stop often
  • Small teams that need RTX A6000, L40 and A100 80GB without a long commitment
  • Anyone whose pipeline ships data out and would otherwise pay for egress

Who should look elsewhere

  • Teams that need a broad hardware catalogue, where SberCloud lists B300, H100 and A100 and ComputePrices.com lists H100 SXM, A100 SXM and H200
  • Buyers who want a monthly rate, since pricing is hourly
  • Workloads on consumer-only cards, where Salad covers NVIDIA RTX 3060, NVIDIA RTX 3070 and NVIDIA RTX 3070 Ti

Core features

The lineup covers RTX A6000, L40 and A100 80GB, spanning workstation and data centre parts for training and inference. Networking runs at 7-10 Gbps networking, so datasets and checkpoints move quickly between instances. Hardware is the main thing to check against your workload, since the list is narrower than the aggregated catalogues on comparison sites.

Plans, egress and billing

GPU hour from $0.35 is the entry figure, and the model is hourly. Billing granularity is per minute, so a run that finishes early stops counting quickly. Egress matters for GPU work, and No data egress fees here. Together these keep the total predictable for short jobs.

Where Thunder Compute is available

Available
Algeria, Angola, Bahrain, Colombia, Côte d'Ivoire, Georgia, India, Iraq, Jordan, Kazakhstan, Kuwait, Moldova, Morocco, North Macedonia, Oman, Paraguay, Qatar, South Korea, Switzerland, Tanzania, Tunisia, Turkey, Ukraine, United Arab Emirates, Uzbekistan, Yemen

Countries from the brand (Source, checked September 26, 2026)

Frequently asked questions

What does a GPU hour cost on Thunder Compute?+

GPU hour from $0.35, with the model listed as hourly. That entry price sits below the site median of €0.63, and the brand ranks 19 of 95 on this measure.

How is time billed when a job ends early?+

Billing granularity is per minute. Short runs therefore stop counting sooner than under whole-hour billing, which suits experiments and fine-tuning passes that finish without warning.

Is there a charge for moving data out?+

Egress is listed as No data egress fees. For GPU work that means shipping checkpoints, datasets or finished predictions does not add a transfer line, which is worth weighing against providers charging per gigabyte.

Which GPUs can I rent?+

The catalogue covers RTX A6000, L40 and A100 80GB. Check the model you need against that list, since it is shorter than the aggregated catalogues used elsewhere on comparison sites.

How fast is the interconnect between instances?+

Interconnect is 7-10 Gbps networking. That sets the ceiling on how quickly a dataset reaches the GPU and how fast results move between nodes during a run.

How do I get help?+

Support is listed on help centre and community. That means documentation and the community rather than a live channel in the list, so plan on solving straightforward issues from the help centre first.

Does Thunder Compute sell a monthly plan?+

The model here is hourly, so a fixed monthly commitment is not what the pricing is built around. Budgets should be planned as usage that scales with the work rather than a subscription.

How does this compare with the widest GPU catalogues?+

Thunder Compute lists RTX A6000, L40 and A100 80GB, while SberCloud lists B300, H100 and A100 and ComputePrices.com lists H100 SXM, A100 SXM and H200. The trade-off is a short hardware list against a low entry rate of $0.35 and No data egress fees.

Sources

  • thundercompute.com/cloud checked September 26, 2026

    Backs: Interconnect

    “7-10 Gbps networking”

  • thundercompute.com/pricing checked September 26, 2026

    Backs: Billing granularity, Egress fee, GPU hour from, GPU models, Pricing model

    “Billed per minute with expandable configurations for any project”

Compiled from 2 source pages, checked September 26, 2026. We did not open an account.

Thunder Compute

Thunder Compute

Short, bursty GPU jobs where keeping running cost down matters

9.1/10See alternatives