Core Concepts

Learn the basics of how Akash Network works in 5 minutes.


How Akash Works

Akash is a decentralized marketplace connecting people who need compute resources with data centers that provide them.

How Akash Works


Key Terms

Deployment

A request for compute resources. You describe what you need (CPU, memory, storage) using SDL (Stack Definition Language).

Provider

A data center offering compute resources on Akash. Providers bid competitively to host your deployment.

Lease

An agreement between you and a provider. Locks in pricing and resources for your deployment.

SDL

Stack Definition Language - a YAML file that describes your deployment:

  • What container image to run
  • How much CPU, memory, and storage you need
  • What ports to expose
  • How much you’re willing to pay

Escrow

Funds held to pay for your deployment. Provider is paid automatically per block from your escrow account.


Basic Workflow

1. Create Deployment

Write an SDL file describing your app and resources needed.

2. Review Bids

Providers submit bids with their pricing. Choose the one you prefer.

3. Accept Bid

Create a lease with the selected provider.

4. App Runs

Provider pulls your container image and starts your app.

5. Automatic Payment

Your escrow pays the provider per block (~6 seconds).


Deployment Options

Akash Console (Easiest)

  • Visual web interface
  • $100 free trial (no wallet needed)
  • Perfect for beginners

CLI (Most Control)

  • Command-line interface
  • Use your own wallet
  • Best for automation

SDK (Programmatic)

  • JavaScript/TypeScript or Go
  • Integrate into your apps
  • Build custom tools

Resource Units

  • CPU: Measured in units (1 unit = 1 thread)
  • Memory: Measured in Mi (Mebibytes) or Gi (Gibibytes)
  • Storage: Measured in Mi or Gi
  • GPU: Measured in units with specific model requirements

Example: 0.5 CPU, 512Mi memory, 1Gi storage


Pricing

Akash is significantly cheaper than traditional cloud providers:

  • Basic web app: ~$1-3/month
  • Small API: ~$5-10/month
  • GPU workload (RTX 4090): ~$50-150/month

Prices vary by provider and demand.


What You Can Update

Container image version
Environment variables
Container commands

Cannot update: CPU, Memory, Storage amounts - Must create new deployment


Learn More

Want to understand Akash in depth?

→ Detailed Core Concepts - Deep dive into deployments, providers, SDL, private containers, persistent storage, IP leases, and more.


Next Steps

footer-logo-dark

© Akash Network 2025 The Akash Network Authors Documentation Distributed under CC BY 4.0

Open-source Apache 2.0 Licensed.

GitHub v0.38.2

Privacy