PoolCompute 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
PoolCompute suits developers and small teams running fine-tunes or batch inference on datacenter parts, with a catalogue covering NVIDIA Tensor Core H100 SXM and AMD Instinct MI300X. The hourly plan is listed at $0.02 for a GPU hour and billed per second, so short jobs stop paying the moment they finish.
Key facts
- Best for
- Short, stop-and-go GPU jobs that want a low entry rate and per second billing.
- GPU models
- NVIDIA Tensor Core H100 SXM, NVIDIA Tensor Core H100 NVL, NVIDIA Tensor Core H100 PCIe, AMD Instinct MI300X↗
- GPU hour from
- $0.020↗
- Billing granularity
- Per second↗
- Legal entity
- PoolCompute Inc↗
- Headquarters
- United States↗
- Founded
- 2024↗
- Support channels
- Email, Community↗
- Pricing model
- Hourly billing
- Official website
- poolcompute.com
Facts checked September 26, 2026, 4 sources
Pros and cons
Pros
- The hourly plan is listed at $0.02 for a GPU hour, so a first fine-tune or batch inference run is cheap to try.
- Billing is per second, so a job that ends early stops accruing cost then; about 26.4% of listed providers meter at that granularity.
- The card list covers NVIDIA Tensor Core H100 SXM, NVIDIA Tensor Core H100 NVL and NVIDIA Tensor Core H100 PCIe, pairing NVIDIA Tensor Core H100 SXM and AMD Instinct MI300X under one provider.
- Contracts sit with PoolCompute Inc, a company headquartered in the United States, so procurement has a named counterparty.
- Entry GPU hour pricing places it at 3 of 95 listed providers on that measure.
Cons
- Support runs on email and community only, while help centres are common among the providers listed here.
- The card list is narrow: NVIDIA Tensor Core H100 SXM, NVIDIA Tensor Core H100 NVL and NVIDIA Tensor Core H100 PCIe covers fewer generations than H100 SXM, A100 SXM and H200 at ComputePrices.com.
- Older generations listed by SberCloud in B300, H100 and A100, such as A100 and V100 parts, have no counterpart in the NVIDIA Tensor Core H100 SXM, NVIDIA Tensor Core H100 NVL and NVIDIA Tensor Core H100 PCIe here.
- Founded in 2024, the company is a recent entrant, so a buyer signs up with a short operating history behind it.
Who it suits
- Teams running fine-tunes, batch inference and evaluation runs that end quickly and gain from per second billing.
- Developers who want an NVIDIA Tensor Core H100 SXM or AMD Instinct MI300X and are content with the NVIDIA Tensor Core H100 SXM, NVIDIA Tensor Core H100 NVL and NVIDIA Tensor Core H100 PCIe listed.
- Small teams that want a named counterparty, with contracts held by PoolCompute Inc in the United States.
Who should look elsewhere
- Teams chasing a specific card generation should check ComputePrices.com, whose H100 SXM, A100 SXM and H200 spans many more models.
- Buyers who need older accelerators alongside current parts should weigh SberCloud, with B300, H100 and A100.
- Workloads built around consumer and workstation NVIDIA cards fit Salad, with NVIDIA RTX 3060, NVIDIA RTX 3070 and NVIDIA RTX 3070 Ti, better than this list of NVIDIA Tensor Core H100 SXM, NVIDIA Tensor Core H100 NVL and NVIDIA Tensor Core H100 PCIe.
Who runs it
PoolCompute trades as PoolCompute Inc, a company headquartered in the United States that dates its GPU cloud work to 2024. Support is limited to email and community, so email carries account questions while a community channel carries the rest. For a buyer, that pairing reads as a young vendor with a domestic legal entity and no other listed support route. Procurement teams that require a phone line, live chat or a formal help centre will find none of those among these channels, and the operating history behind the service is short. Sizing a commitment should account for both points.
How PoolCompute compares
PoolCompute vs RunPod9.1 vs 9.5
PoolCompute vs ScalaHosting9.1 vs 8.3
PoolCompute vs Contabo9.1 vs 9.2
PoolCompute vs Kamatera9.1 vs 9.5
PoolCompute vs Cloudways9.1 vs 9.2
PoolCompute vs PlanetHoster9.1 vs 8.7
PoolCompute vs Acronis9.1 vs 9.9
PoolCompute vs Aruba9.1 vs 9.3
PoolCompute vs Decodo9.1 vs 9.1
PoolCompute vs Hostinger9.1 vs 9.1- PoolCompute vs MarsProxies9.1 vs 9.0
PoolCompute vs hide.me9.1 vs 8.7
PoolCompute vs Proton Drive9.1 vs 8.3
Frequently asked questions
What does a GPU hour cost on the hourly plan?+
PoolCompute lists the hourly plan at $0.02 for a GPU hour. That is the plan rate rather than a promotional one, so the entry figure and the figure that keeps applying are the same. Across the brands listed here, the median GPU hour price is €0.63.
How is usage billed?+
The pricing model is hourly and usage is billed per second. Per-second charging suits fine-tunes, batch jobs and evaluation runs that finish quickly, because a job stops accruing cost the moment it ends. Longer training runs sit on the same per-second meter.
Which GPUs can I rent?+
The catalogue covers NVIDIA Tensor Core H100 SXM, NVIDIA Tensor Core H100 NVL and NVIDIA Tensor Core H100 PCIe, pairing NVIDIA Tensor Core H100 SXM with AMD Instinct MI300X and adding H100 NVL and PCIe variants. ComputePrices.com lists a far wider set, H100 SXM, A100 SXM and H200, so a specific older card is more likely to appear there.
How do I reach support?+
Support runs on email and community, with email as the direct route and the community channel for open questions. About 90.2% of listed providers also run a help centre, which is the gap to weigh if you expect written documentation.
Who am I contracting with?+
The provider is PoolCompute Inc, headquartered in the United States, and was founded in 2024. A named legal entity in US keeps procurement, invoicing and vendor review in familiar territory, though the company is young by the standards of the market.
Is PoolCompute a fit for a small team?+
Short, experimental jobs are the natural fit here: an entry rate of $0.02 on the hourly plan plus per second billing keeps a test run inexpensive. The listed overall score is 9.6, and the cards in NVIDIA Tensor Core H100 SXM, NVIDIA Tensor Core H100 NVL and NVIDIA Tensor Core H100 PCIe are current datacenter parts rather than a deep catalogue.
How does it compare with SberCloud and Salad?+
SberCloud lists B300, H100 and A100, mixing current parts with older generations that have no counterpart here, while Salad lists NVIDIA RTX 3060, NVIDIA RTX 3070 and NVIDIA RTX 3070 Ti, built around consumer and workstation cards. PoolCompute stays narrow: NVIDIA Tensor Core H100 SXM, NVIDIA Tensor Core H100 NVL and NVIDIA Tensor Core H100 PCIe with per second billing.
Sources
- poolcompute.com checked September 26, 2026
Backs: Billing granularity, GPU hour from
“Save over 68% on your cloud bill with pay-by-the-second billing”
- poolcompute.com/blog/gpu-cloud-services checked September 26, 2026
Backs: Pricing model
“PoolCompute offers on-demand access to NVIDIA H100s at competitive rates”
- poolcompute.com/gpu checked September 26, 2026
Backs: GPU models
“Available NVIDIA GPU Servers on Rental ... NVIDIA Tensor Core H100 SXM ... NVIDIA Tensor Core H100 NVL ... NVIDIA Tensor Core H100 PCIe ... AMD Instinct MI300X”
- poolcompute.com/terms checked September 26, 2026
Backs: Headquarters, Legal entity, Support channels
“We are registered in Delaware, United States and have our registered office at 102 Lakeland Avenue, Dover, DE 19901.”
- linkedin.com/company/poolcompute checked September 26, 2026
Backs: Founded
“Founded 2024”
Compiled from 5 source pages, checked September 26, 2026. We did not open an account.
Not confirmed yet: Egress fee.
Related by the facts
PoolCompute
Short, stop-and-go GPU jobs that want a low entry rate and per second billing.