A cloud GPU platform that has been operating as Paperspace since around 2014, acquired by DigitalOcean in 2023 and now integrated into DigitalOcean’s AI/ML product lineup. It lets you rent GPUs including the NVIDIA H100 by the hour, and covers the full workflow — development in Jupyter notebooks (Gradient Notebooks), model training, and deployment — within a single cloud. It suits individuals and teams who want to spin up an AI development environment only when needed, without buying expensive GPUs of their own.
Key Features
- Wide GPU lineup including NVIDIA H100: From high-end options such as the H100, A100-80G, and A6000 down to affordable classes like the A4000 and RTX4000, you can choose by the hour according to your workload and budget
- Gradient Notebooks running in the browser: A zero-setup Jupyter notebook environment. Even the free plan lets you try some GPU-backed notebooks, making it an easy entry point for learning and experimenting with machine learning
- Integrated development, training, and deployment: Prototype in notebooks, run full training on GPU machines, and serve models as inference APIs with Gradient Deployments — all on the same platform
- GPU machines usable as virtual desktops (CORE): Launch GPU virtual machines with a full OS, useful not only for machine learning but for any GPU-heavy work such as 3DCG and rendering
- Integration with the DigitalOcean ecosystem: Since the acquisition, consolidation with DigitalOcean’s own AI products such as GPU Droplets and serverless inference has progressed, making it easier to combine with existing DigitalOcean infrastructure
Pricing
Pricing combines a Gradient (ML platform) subscription with hourly billing for GPU instances.
| Plan | Monthly Price | Key Features |
|---|---|---|
| Free | $0 + instance usage | Public projects, auto-shutdown (12 hours) |
| Pro | $8 + instance usage | Private projects, 15GB storage |
| Growth | $39 + instance usage | 50GB storage, expert support |
Reference hourly GPU rates (on-demand): H100 at $5.95/hr, A4000 at $0.76/hr, RTX4000 at $0.56/hr, among others. Significant discounts apply with long-term commitments (H100 at $2.24/hr on a 3-year contract). Team plans (T0–T2) are also available separately.
Pricing is as of August 2026. For the latest pricing, please check the official site.
Pros & Cons
✅ Pros
- Use H100-class compute for only the hours you need, without purchasing expensive GPUs
- Gradient Notebooks let you start with just a browser, keeping the setup barrier low
- Development, training, and deployment complete within one platform, with little tool-switching
- Entry-level GPUs start at under $1/hr, allowing small experiments at low cost
- Being part of DigitalOcean brings the reassurance of an established infrastructure company
⚠️ Cons
- Popular GPUs can be out of stock at certain times and may be hard to secure
- The two-tier structure of subscription plus hourly billing can make cost estimation somewhat complex depending on usage
- With the product lineup migrating to the DigitalOcean side after the acquisition, documentation and dashboard layouts may change
- On-demand H100 rates can be pricier than other GPU-focused clouds, so price comparison is essential
Comparison with Similar Services
| Comparison | DigitalOcean (Paperspace) | RunPod | Lambda Cloud | Google Colab |
|---|---|---|---|---|
| Main use | Integrated development–training–deployment | Low-cost GPU rental, serverless inference | GPU cloud for training | Notebook-based learning and experiments |
| Notebook environment | Yes (Gradient) | Yes | JupyterHub provided | Core feature |
| Free tier | Free plan + some free GPUs | None (pay-as-you-go only) | None | Yes (with limits) |
| Deployment features | Gradient Deployments | Serverless available | Limited | None |
| Distinctive point | Integrated with DigitalOcean products | Cheap community cloud | Research and training focused | Simplicity first |
Who Is It For
- Individuals and students who want a machine learning and deep learning environment without buying a GPU
- Small teams that want to go from notebook prototyping to model deployment within one cloud
- Researchers and developers who need H100-class GPUs for short-term, spot usage
- Existing DigitalOcean users who want to keep AI workloads in the same ecosystem
Summary
DigitalOcean (Paperspace) is an integrated platform for AI development, training, and deployment where cloud GPUs, including the NVIDIA H100, are billed by the hour. It combines the ease of starting with just a browser with a lineup that reaches up to high-end GPUs, and since the DigitalOcean acquisition it has continued to evolve as part of a larger AI product family. A good approach is to first try the Free plan and lower-priced GPUs, then scale up gradually while estimating hourly costs for serious training workloads.