Sandbox
An engineering team wants to test an AI assistant on approved sample questions before using it in everyday work.
What would make the experiment useful?
Storm AI Factory™
An enterprise AI appliance.
Bring generative AI and agents into a customer-controlled system. Storm AI Factory™ is an all-in-one inference and training appliance built on open-source software, with Storm™ cluster management included.
Deploy on premises or in colocation, alongside your existing compatible infrastructure. Size the system for inference, training, or both.
Workload examples
Start with the AI workflow.
Begin with a task your team needs to accomplish. These examples help define the models, tools, data, and capacity your AI Factory needs.
Train
A team needs to run a model-training job with an approved dataset and an allocation of compute and storage.
What data and capacity does the job need?
Inference
An application team wants to use a model to classify incoming documents, with capacity sized for the arriving requests.
How many requests must it serve, and how quickly?
Agents
A team wants an agent to gather information from approved sources and draft a report for a person to review.
Which actions can proceed, and which need a person's approval?
Deployment options
Add AI to your infrastructure.
Install the standalone appliance at your site or in a colocation facility and connect it to the approved applications and data sources in your existing environment. The deployment plan defines those connections, access controls, capacity, and support responsibilities.
Also available within NetThunder platforms.
The AI operations module is included with CloudSpawn™ and optional with Spark™ and SecureEnclave™. Choose one of these deployments when its broader operating model fits your needs.
Spark™
Available in selected configurations. Run private AI on one supported node when the workflow fits its capacity and operating boundary.
Review Spark™SecureEnclave™
Available in selected configurations. Place approved private AI inside assessment scope, or connect through an explicitly approved integration path.
Review SecureEnclave™CloudSpawn™
Included. Run AI within an elastic, supercomputer-class tenant environment with qualified bare-metal capacity, high-performance data fabrics, and parallel storage.
Review CloudSpawn™Supported models, runtimes, data services, topology, capacity, performance, isolation, and operations differ by platform and configuration.
Define data and operating controls
The environment specification identifies where source data, model artifacts, prompts, outputs, logs, and retained state can live and move. It also defines identities, permitted connections, administrative paths, and responsibility for access, monitoring, change, updates, incidents, backup, restoration, and retirement.
Metering and limits
Use your assigned capacity.
Match capacity to demand.
Size compute and storage around model requirements, concurrent users, and training jobs. Your subscription scope and any separately licensed software or services still apply.
Plan capacity changes around the selected deployment and its operating requirements.
Included with your appliance
Storm™ manages the underlying infrastructure.
The appliance includes Storm™ to provision its supported infrastructure and manage configuration as requirements change. Your team chooses the AI applications and requested changes; Storm™ calculates and applies the underlying configuration from service definitions, dependencies, and required state.
Keep control outside workload execution
The included Storm™ software configures supported compute, storage, and services through its out-of-band control path. In a bare-metal configuration, AI workloads run directly on qualified processors and accelerators. The architecture is designed to preserve bare-metal performance characteristics.
Size compute and storage together
Select compute, accelerator memory, storage, and data paths together around the models and jobs you intend to run. Supported configurations can use high-performance fabrics and shared-root storage; the selected topology determines capacity and performance.
Already operate an HPC or AI cluster? Storm™ is also available as cluster-management software and can run AI applications without the AI Factory appliance. Explore Storm™ for your cluster
Architecture review
Start with the AI task.
Bring the task, its data needs, and where you want the system to run. The first conversation identifies an appliance configuration or platform deployment to explore, open questions, and the next design decisions.