# Massed Compute

Configure cloud GPU instances and manage hardware, images, billing, and VM lifecycle through a web console, REST API, or MCP.

- Findings24: <https://findings24.com/ja/products/massed-compute>
- ウェブサイト: <https://massedcompute.com>
- 料金: Paid
- カテゴリー: Developer Tools, GPUクラウド

## Massed Computeとは？

Massed Compute is a web-based platform for configuring and managing cloud GPU instances. Choose a GPU configuration and software image, deploy a virtual machine, then connect and manage it through the console or API. NVIDIA drivers and CUDA are included with on-demand instances.

The platform combines hourly compute with inventory, instance-management, and billing tools. Its MCP server lets compatible AI clients work with the same account, using read-only or full-access credentials. Dedicated bare metal and multi-node clusters are separate, custom-quoted options.

## Massed Computeでできること

### Configurable GPU virtual machines

Select hardware, GPU quantity, an image, and optional SSH keys or startup settings. The deployment interface shows specifications and the hourly price before launch.

### Prepared software environments

Start from images for PyTorch, vLLM, ComfyUI, Ollama, or JupyterLab, with virtual desktop access for supported configurations. Installed software and model licences still apply.

### Console, REST API, and MCP control

Check GPU inventory, inspect instances, launch or restart VMs, and manage SSH keys through documented interfaces. MCP also exposes billing settings and setup-recipe tools; write actions require full access.

### Optional persistent storage

After enabling the account's beta access, persistent storage can remain after a VM is deleted and attach through the website or REST API. It has separate storage and download charges and is not available through MCP.

## 利用シーン

### Model development and inference

Run your selected training or inference software on a GPU configuration sized for the workload.

### Rendering and technical workloads

Rent compute for rendering, notebooks, or simulations while supplying the software and licences your project requires.

### Automated compute management

Integrate provisioning and cost inspection with your own tooling or an authorized MCP-compatible assistant.

## よくある質問

### How much does an instance cost?

On-demand pricing is hourly and depends on the full configuration. Current public examples include one H100 80GB at US$2.73/hour and one RTX PRO 6000 Blackwell 96GB at US$2.19/hour. The homepage lists a one-hour minimum; inventory, coupons, and dedicated-service quotes can change the applicable offer.

### Does stopping the VM stop billing?

No. The VM guide says Stop leaves billing running. Delete ends the instance’s billing and destroys its local disk, so export needed files first. Separate Persistent Storage can retain data and continue accruing storage charges.

### Is all data transfer free?

On-demand VM egress is advertised as free. Optional Beta Persistent Storage is billed per GB stored per month; downloads beyond three times average monthly stored data cost US$0.05/GB. These are different services and allowances.

### How do I authorize API or AI-client access?

Create an API key in account settings, or approve an OAuth grant for a compatible MCP client. Use read-only access for inspection and full access for provisioning or deletion. The account controls which resources the client can manage.

## クイック情報

- **Product**: Massed Compute GPU management platform
- **Access**: Web console, REST API, and MCP
- **Purchase**: Hourly paid instances; dedicated services by quote
- **Data lifecycle**: VM deletion removes local disk; Beta storage is separate