TOOL 001Free tool
LLM VRAM Calculator
Estimate model weights, KV cache, runtime overhead, and GPU count for inference.
- Useful for
- Sizing a local server, inference node, or GPU request
- Output
- GiB required and minimum GPU count
AI infrastructure reference
Plan with inputs you control.
Use transparent calculators for memory, power, and data movement. The provider directory links to official provisioning documentation without assigning invented scores or implying an integration.
Status3 free tools · research directoryEvery Run Any AI resource declares what is real today.
Start with the workload
Calculate before comparing vendors.
These tools run locally and show the arithmetic. They do not fetch live pricing, inventory, quota, or account data.
TOOL 002Free tool
AI Cluster Power Calculator
Turn a GPU count into IT load, facility load, cooling, energy, and electricity cost.
- Useful for
- Planning a rack, lab, cluster expansion, or facility conversation
- Output
- kW, kWh, cooling load, cost, and rack density
TOOL 003Free tool
Dataset Transfer Calculator
Estimate transfer time, effective throughput, replicas, egress, storage, and growth.
- Useful for
- Moving training data, checkpoints, embeddings, or backups
- Output
- Wire time, storage footprint, and user-supplied cost estimate
Source-linked provider research
Six provisioning surfaces, without a ranking.
Documentation links last reviewed August 2, 2026. Prices, inventory, quotas, regions, and product names can change; verify them directly before spending.
ResearchGPU cloud
RunPod
A container-oriented GPU Pod API suited to experiments, inference workers, and direct GPU selection.
- Provisioning
- REST Pods API · POST /v1/pods
- Credentials
- Bearer API token
ResearchGPU cloud
Lambda Cloud
A focused REST API for launching and managing GPU instances, images, SSH keys, filesystems, and firewalls.
- Provisioning
- Lambda Cloud REST API · instance-operations/launch
- Credentials
- Bearer API key
ResearchHyperscaler
AWS EC2
A broad infrastructure API with GPU instance families, networking, IAM, storage, Spot, and mature lifecycle controls.
- Provisioning
- EC2 Query API · RunInstances
- Credentials
- IAM credentials with Signature Version 4
ResearchHyperscaler
Google Compute Engine
A REST and client-library surface for creating VMs with attached accelerators, IAM, networks, disks, and reservations.
- Provisioning
- Compute Engine REST API · instances.insert
- Credentials
- OAuth access token through IAM or workload identity
ResearchHyperscaler
Azure Virtual Machines
An Azure Resource Manager API for creating GPU VMs within subscriptions, resource groups, networks, and identity controls.
- Provisioning
- Azure Compute REST API · Virtual Machines Create or Update
- Credentials
- Microsoft Entra OAuth through a managed identity or service principal
ResearchAI infrastructure
CoreWeave
A Kubernetes-first AI cloud whose CKS API creates and manages clusters before GPU node pools and workloads are attached.
- Provisioning
- CKS REST API · POST /v1beta1/cks/clusters
- Credentials
- Bearer API access token
Clear boundary
Run Any AI does not provision infrastructure.
The calculators and provider research require no Run Any AI account, AI API key, provider key, or database. No provider is scored or recommended by an unpublished formula.
Creating infrastructure happens with the provider and requires an account, current price and capacity checks, scoped credentials, spend controls, monitoring, and dependable cleanup. The source notes are a starting point for that review—not a substitute for it.