Affiliate disclosure: Some links on this page may be affiliate or sponsored links. We may earn a commission at no extra cost to you. This does not change our editorial criteria.

Rent GPU cloud

This page is a beginner guide to renting GPUs in the cloud. A GPU is a graphics processing unit: a chip that is very good at doing the same math thousands of times at once. GPU cloud means you rent those chips by the hour from a company, then turn them off when the work ends. People also call this GPU rental or GPU as a service. Same product: extra chips you do not own.

Who it is for: founders, lab leads, studio producers, and the person who was told "we need GPUs" and wants the plain-English version first. You do not need to be an engineer to follow this page. You do need a job a laptop cannot finish.

You rent GPU cloud when you need that class of chip this week and you can shut the meter off. You do not buy a room of servers for a campaign that may last six weeks. If those same machines stay busy for years, the next problem is usually electricity and a site. That is powered land: land with a credible path to large-load power.

What GPU rental actually is

GPU rental is extra compute you do not own. A provider gives you a machine with one or more data-center GPUs. You log in, copy your files, run the job, save the results, and stop paying.

It is not a laptop graphics card. Data-center GPUs sit in servers with their own fast memory and special links between chips. CUDA is NVIDIA's software layer that lets programs use NVIDIA chips. Most of the AI, science, and render software you will hear about expects CUDA drivers on the machine.

A reservation is still rental. You may pay for a block of weeks. You still do not control the building, the utility, or the next density jump. The useful questions are which chip, how many, whether they can talk to each other, whether you can turn them off, and what else is on the bill.

You rent GPUs to teach or run a language model

A large language model (LLM) is a computer program that predicts text. Training means showing it huge amounts of examples so it learns patterns. That job usually wants many matching GPUs that can share work. The checklist is rent GPUs for LLM training.

Inference means using a trained model to answer questions. Fine-tuning means teaching an existing model a smaller job, often in days, then shutting the GPUs off.

You rent for these jobs when the model size is still moving, you do not know how many hours a run will take, or the traffic is still a guess. Paying a premium per hour to learn that is rational. Buying a hall of cards for a run that might fail in week two is not.

Save checkpoints (snapshots of the model) often. Ask where those files live. If every save crosses the public internet, the run will stall.

You rent GPUs for science and industry bursts

Drug-research teams rent GPUs to score candidate molecules, teach protein models, rebuild 3D shapes from a special microscope, or simulate how atoms move. That beginner checklist is GPU cloud for life sciences.

Physics, climate, and airflow jobs rent GPUs when the solver actually has a GPU path. Studios rent GPUs to finish a sequence of frames. Those are bursts: a window of work, then quiet. A farm that never sleeps is a different product.

If the campaign is one-off, rent. If the machines are busy unless someone is on vacation, you are already describing a cluster you might need to own or house.

Which GPU you are actually buying

Cloud menus name chips. The name is not the whole product. Start from memory and count, not from a famous label.

The A100 is NVIDIA's previous generation (Ampere), still widely rented; ask 40 GB or 80 GB. The H100 is the Hopper workhorse people mean by "we need H100s." The H200 is Hopper with a bigger memory tank (NVIDIA lists 141 GB). The B200 is Blackwell. B200 is the GPU, not automatically a whole rack.

NVLink is NVIDIA's fast link between GPUs in the same server. SXM is the dense tray. PCIe is the plug-in card. MIG (Multi-Instance GPU) means the provider can slice one GPU into smaller pieces. A slice is not a whole GPU.

What to check before you rent

Start with the software, not the famous name. Confirm NVIDIA drivers and a CUDA version your code will actually run. Confirm the work you intend to put on one GPU will fit in that GPU's own memory.

Ask whole GPU vs MIG slice, SXM vs PCIe, and how NVLink is wired if several chips train together. Spot (interruptible) GPUs are cheaper and can vanish. They are fine if you checkpoint. They are a bad fit for a deadline. Egress is the fee for copying data out. Keep working files next to the chips.

How to read a quote lives on the GPU cloud pricing index. How provider types differ lives on GPU cloud providers. We will not invent hourly prices on this hub.

GPU cloud partners (placeholder)

Partner rows below are placeholders until terms are signed. No live rates. No outbound URLs. When links go live, they will use rel="sponsored".

This table is not a ranking. It will reflect partner relationships once partners exist.

GPU cloud pricing varies. Verify live rates.

There is no honest single dollar-per-hour number on this page. List prices move with GPU type, region, form factor, and whether you pay only when the machine is on. Quoted "from" rates often skip storage, data copies, and idle reservations. See the GPU cloud pricing index for how we talk about rates. We will not invent hourly prices here. Treat every public figure as stale until you re-check the provider.

Compare hours you will actually run. A three-week training job is not the same product as a chatbot that answers all year.

When GPU rental stops working

Rental stops working when utilization is high and predictable, the hardware mix is stable, and the next constraint is megawatts and a site, not another login. A megawatt is a large unit of electrical power. You are then running a cluster on someone else's floor.

The honest alternative is powered land: a development site with a credible path to large-load electricity. Browse the U.S. data center directory, then read when GPU clusters need powered land. For the rent-versus-build fork in plain English, see rent GPUs vs build a data center. Hybrid is normal. Rent the bursts. Site the work that never turns off.

Find powered land when you need a site. Rent GPUs when you need capacity now.

Related

• GPU cloud for life sciences and drug discovery
• Rent GPUs for LLM training
• Rent NVIDIA H100 GPUs

FAQ

What is GPU cloud rental in plain English?

You pay a company for time on data-center GPUs, run your job, and stop paying when you shut the machines off. It is the same idea as GPU as a service. It is not a laptop graphics card, and it is not a building you own.

Which GPU should I rent?

Start from memory and count, not from the newest name. Many jobs still fit on an A100 or one H100. Full training of a serious language model usually wants several matching H100-class chips. If the model no longer fits, look at the H200. Rent B200 when Hopper is the real bottleneck and a provider actually has the chips.

How much does GPU rental cost per hour?

There is no honest single number. Quotes move with chip, region, and whether the machine can sit idle on your bill. We do not invent dollar-per-hour rates. Use the pricing index to learn how to read a quote, then re-check the provider.

Is a reserved GPU the same as owning a cluster?

No. A reservation is capacity on someone else's floor. You still do not control power, cooling, or the next generation's density. See rent vs build.

When should I stop renting and look at land?

When the same GPUs would stay busy for a planning horizon in years, and the limiter is electricity and a site. Start with when GPU clusters need powered land and what powered land is.

Do I need to be an engineer to rent GPU cloud?

No. You need a clear job, a way to save progress, and a person who can confirm the software will run. This hub is written for the buyer who has to make that call.

Sources

1. NVIDIA H100 GPU
2. NVIDIA H200 GPU
3. NVIDIA A100 Tensor Core GPU
4. NVIDIA DGX B200
5. NVIDIA NVLink
6. NVIDIA CUDA Toolkit
7. What is powered land
8. U.S. data center directory
9. Powered Lands data center map

Need land or power instead? See the map or contact.