On-demand hourly instances
Launch hourly GPU instances without contracts. The pricing page lists on-demand configurations with per-hour rates, and the homepage says setup is around 90 seconds for standard instances.
Massed Compute rents NVIDIA cloud GPUs on demand for AI, rendering, and compute-heavy workloads, with hourly instances, bare metal, clusters, and API provisioning.
Massed Compute is a cloud GPU service for renting NVIDIA hardware on demand. The site positions it as a way to deploy AI, rendering, big-data, and other compute-heavy workloads without contracts, with instances billed by the hour.
The product spans more than simple virtual GPUs: the homepage describes four compute modes — on-demand hourly GPUs, bare metal, GPU clusters, and a programmatic API/MCP workflow. It also states that instances ship with preinstalled CUDA, free egress, and a one-click VDI session, and that users can launch a GPU instance in about ninety seconds.
The pricing page publishes hourly rates for a range of configurations, including RTX PRO 6000 Blackwell, H200 NVL, H100, A100, L40S, RTX 6000 Ada, RTX A6000, RTX A5000, and A30. The partner page says Massed Compute is an NVIDIA Preferred Partner with access to NVIDIA's enterprise catalog and vendor-tested drivers and firmware.
Launch hourly GPU instances without contracts. The pricing page lists on-demand configurations with per-hour rates, and the homepage says setup is around 90 seconds for standard instances.
Choose among server types for different operational needs, including single-instance GPUs, bare metal, GPU clusters, and programmatic access for teams.
Use NVIDIA GPUs from the enterprise catalog, spanning Blackwell, Hopper, Ada Lovelace, and Ampere families, with listed models such as B300, H200, H100, A100, L40S, RTX 6000 Ada, RTX A6000, A40, and A30.
Start with preinstalled CUDA and vendor-certified drivers and firmware so instances are ready for training, inference, rendering, or development work.
Automate provisioning through a REST API or MCP server. The homepage shows bearer-key authentication, scoped tokens, webhooks, SDKs, and a Terraform provider.
Work from curated templates like PyTorch, vLLM, ComfyUI, Ollama, and JupyterLab, or bring your own image.
Launch a short-lived GPU instance for model training, experimentation, or inference when you want hourly billing and a fast start rather than a long contract.
Rent single-tenant hardware when workloads need more isolation than shared instances and the site’s bare-metal offering fits the security or operational requirement.
Run large multi-node jobs on GPU clusters when the workload needs higher scale, networked nodes, and dedicated support.
Provision GPUs from code or an AI agent using the REST API or MCP server, then connect the workflow to CI, notebooks, or internal automation.
Use curated templates such as PyTorch, vLLM, ComfyUI, Ollama, or JupyterLab to start with a known software stack instead of building one from scratch.
The site offers on-demand hourly GPU instances, plus bare metal, GPU clusters, and a programmatic API/MCP path for automation. The homepage also shows curated templates such as PyTorch, vLLM, ComfyUI, Ollama, and JupyterLab.
The pricing page shows hourly on-demand pricing for listed GPU configurations, and the homepage says instances are billed by the hour with no contracts. The contact page also invites users to reach out about pricing and configuration.
The homepage says instances can launch in about 90 seconds, and the contact page is aimed at users who want help choosing hardware or sizing a workload.
The homepage says the service includes preinstalled CUDA, and the partnership page says deployments ship with NVIDIA-certified drivers and firmware.
The site presents Massed Compute as an NVIDIA Preferred Partner and says it provides access to NVIDIA's enterprise catalog, but the sources do not document a public self-serve API reference beyond the examples shown on the homepage.