Compute for Autonomous AI Agents — pay in dollars, no crypto

Rent GPU and CPU by the hour, priced in dollars. No AKT, no ACT, no YAML — we run the decentralized cloud underneath.

Powered by Akash Network — verifiable, decentralized compute.

Fast resource allocation

Akash matches your SDL to providers automatically, so you're never waiting on a human to provision a VM. Containers are typically live within a couple of minutes of submitting the order.

1–5 min
typical order → running container
~30–60 s
fastest — small or cached images
seconds
order → lease, on-chain (1–3 blocks)

What actually takes the time

  1. Order & bid (~seconds). Your deployment create hits the chain; providers submit matching bids within 1–3 blocks (~6–7 s each).
  2. Lease & manifest (~seconds). Accept a bid, send the SDL manifest — the provider validates it and schedules your workload immediately.
  3. Image pull & start (the dominant cost). Small images (10–50 MB) start in seconds; large AI images (several GB) can take 5–20+ min on first pull, or near-instantly if already cached on the provider.

Tip: keep your container image small (multi-stage builds, a distroless base) and pin the exact digest — repeat deployments on a warm provider consistently hit the ~1-minute end of the range.

Pricing & pre-order

Pay in dollars — no AKT, ACT or escrow to figure out. We run the Akash side for you and email you access within 24h.

Small Agent
$3.99/mo
per instance

1 vCPU · 2 GB RAM · 20 GB for agents that need a little more room.

Pre-order
Own hardware
Compute
$16/mo
full server on our own nodes

8 vCPU · 16 GB RAM · 160 GB — the Hetzner-class equivalent, priced in dollars.

Pre-order
Agent GPU · 24 GB
$25
50 GPU-hours ≈ $0.50/hr

A10G / L4 / 4090 or equivalent. Inference, embeddings and autonomous-agent workloads.

Pre-order
Train GPU · 80 GB
$199
50 GPU-hours ≈ $3.98/hr

A100 / H100 or equivalent. Fine-tuning and large-model inference with headroom to spare.

Pre-order
Most compute
Train GPU · 320 GB
$410
25 GPU-hours ≈ $16.41/hr

4× H100 / 4× A100 (320 GB) or 2× H200 (282 GB). For full fine-tuning of 70B-class models.

Pre-order

The models listed are examples of the class you get — we guarantee VRAM and performance, not a specific chip. Need a different size? Use the form below and we'll size + quote it before you pay.

What are you deploying?

Tell us your workload and we'll match it to a tier (or quote a custom size) before you pay anything.

1 · How to publish your Docker / Kubernetes container on Akash

You describe your workload in a small YAML file (the Stack Definition Language, SDL); Akash matches it to a provider and runs it. Prefer to skip all this? Buy a managed plan and we'll deploy it for you.

  1. Get a wallet. Install Keplr or Leap, create an Akash account, and copy your akash1… address.
  2. Fund it. You need AKT for gas fees, plus escrow in USDC or ACT — see our top-up guide.
  3. Write the SDL. Save this as deploy.yaml:
version: "2.0"
services:
  web:
    image: nginx:latest
    expose:
      - port: 80
        as: 80
        to:
          - global: true
profiles:
  compute:
    web:
      resources:
        cpu: { units: 1 }
        memory: { size: 1Gi }
        storage: [ { size: 10Gi } ]
  placement:
    akash:
      pricing:
        web: { denom: uakt, amount: 100 }
deployment:
  web:
    akash:
      profile: web
      count: 1
  1. Deploy. Use Akash Console (console.akash.network) for point-and-click, or run the CLI sequence below. When a bid is accepted, your container starts and a public URL is issued.

CLI — one-time setup

provider-services config set node http://<NODE>:26657
provider-services config set chain-id akashnet-2
provider-services keys add deploy-key          # or: keys import / keys add --recover

The examples use --keyring-backend test for simplicity. For production, use --keyring-backend os and a hardware wallet.

CLI — deploy, bid, manifest

# 1. create the deployment
provider-services tx deployment create deploy.yaml \
  --from deploy-key --keyring-backend test \
  --chain-id akashnet-2 --gas auto --gas-adjustment 1.5 --gas-prices 0.025uakt -y

# 2. list bids (use the DSEQ printed in step 1)
provider-services query market bid list \
  --owner $(provider-services keys show deploy-key -a --keyring-backend test) \
  --dseq <DSEQ> --output json

# 3. accept a bid
provider-services tx market lease create \
  --owner <AKASH_ADDRESS> --dseq <DSEQ> --gseq 1 --oseq 1 \
  --provider <PROVIDER_ADDRESS> \
  --from deploy-key --keyring-backend test \
  --chain-id akashnet-2 --gas auto --gas-adjustment 1.5 --gas-prices 0.025uakt -y

# 4. send the manifest
provider-services send-manifest deploy.yaml \
  --owner <AKASH_ADDRESS> --dseq <DSEQ> --provider <PROVIDER_ADDRESS> \
  --from deploy-key --keyring-backend test --output json

# 5. get status + public URL
provider-services lease-status --dseq <DSEQ> --provider <PROVIDER_ADDRESS> \
  --from deploy-key --keyring-backend test

Your container is typically live 1–5 minutes after the manifest is accepted (see fast allocation).

2 · Ways to top up the account

Escrow is paid in USDC or ACT; gas is always paid in AKT. Pick whichever suits you.

USDC / USDT / ETH (Ethereum blockchain)

USDC is accepted directly for escrow. Deposit it through Akash Console: Wallet → Deposit → USDC — the flow bridges USDC to Noble and transfers it to Akash over IBC automatically.

USDT and ETH are not accepted as escrow on Akash. Swap them to USDC or AKT first (on any DEX, e.g. Osmosis), then bridge/transfer the result to your Akash address.

ACT (Akash blockchain)

ACT is Akash's native compute token (escrow currency, pegged ~$1). Mint it by burning AKT:

In Console Air: Wallet Actions → Mint ACT.

Or via CLI: provider-services tx bme mint-act 500000uakt --from mykey.

Note: the minimum mint is 10 ACT.