Arrakis: rent GPUs at the lowest live price, safe for AI agents

What GPUs cost right now

Every search covers the markets Arrakis connects and lists the cheapest machines first. These are the lowest prices per GPU-hour at this moment. Compare every GPU.

Prices come from independent markets and change minute to minute. Updated 13 Oct 2026, 16:19 UTC.
GPU	Memory	From	Available	Rent
H100 80GB HBM36 machines	up to 80 GB	$1.30/GPU-hr	29GPUs	Rent
H2005 machines	up to 140 GB	$3.52/GPU-hr	16GPUs	Rent
B2004 machines	up to 179 GB	$5.60/GPU-hr	18GPUs	Rent
B300 SXM6 AC4 machines	up to 269 GB	$6.99/GPU-hr	30GPUs	Rent
A100 80GB PCIe6 machines	up to 80 GB	$0.65/GPU-hr	20GPUs	Rent
RTX PRO 6000 Blackwell Server Edition2 machines	up to 96 GB	$1.77/GPU-hr	2GPUs	Rent
RTX 50905 machines	up to 32 GB	$0.72/GPU-hr	5GPUs	Rent
RTX 40905 machines	up to 24 GB	$0.46/GPU-hr	5GPUs	Rent
RTX A60008 machines	up to 48 GB	$0.42/GPU-hr	9GPUs	Rent
RTX 6000 Ada6 machines	up to 48 GB	$0.70/GPU-hr	6GPUs	Rent
RTX 30902 machines	up to 24 GB	$0.22/GPU-hr	8GPUs	Rent

Hand it to your agent. Keep the budget.

Agents connect over MCP or the REST API. Every rental they make is bounded before it starts, so you can let them run unattended.

A token with a ceiling

Give your agent a token that can search and rent, with a spend cap it cannot pass.
```
arrakis tokens create --name eval-agent \
  --scope read --scope rent --spend-limit 50
```
Every machine shuts itself down

Rentals carry a removal time. A capped token cannot rent without one.
```
"termination_hours": 6
```
Retries never rent twice

The same idempotency key returns the same machine, not a second bill.
```
"idempotency_key": "eval-run-0412"
```

Connect over MCP

claude mcp add --transport http arrakis $ARRAKIS_URL/mcp \
  --header "Authorization: Bearer ark_..."

{
  "mcpServers": {
    "arrakis": {
      "url": "$ARRAKIS_URL/mcp",
      "headers": { "Authorization": "Bearer ark_..." }
    }
  }
}

Transport  Streamable HTTP
URL        $ARRAKIS_URL/mcp
Header     Authorization: Bearer ark_...

$ARRAKIS_URL is this site's address. Replace ark_... with an API token from the console.

Then ask it something like

Rent the cheapest H100 under $2 an hour for three hours, run eval.py, copy the results back, then shut it down.

Read the agent guide

From sign-in to SSH in three steps

No cloud accounts to open and no card to add per provider. One balance covers every market.

Sign in and add funds
Use your email, then request a bank transfer. Your balance grows when it arrives.
Pick a GPU
Cheapest first, filtered by memory, count, price and country. Renting is one dialog with a shutdown time already set.
SSH in as root
Your Docker image, your keys. The machine removes itself on schedule, or whenever you say.

One console for every market

Your balance and how long it lasts, what is running and what it costs, and the cheapest GPUs right now. Agents get the same view over MCP.

The Arrakis console overview: a balance of $3,756.43 spending $23.72 an hour across two machines, the machines running now with their hourly cost and a copy SSH button, and the cheapest GPUs right now. — The console overview, with an example account and example prices.

Compute is the scarcest input in AI. It should flow to whoever needs it, person or agent.

Supply is scattered

The cheapest GPUs sit in independent markets, each with its own account, card, CLI and prices that move by the minute.

Agents are becoming buyers

Coding agents now run experiments end to end. They need a budget they cannot exceed, not a credit card.

Spend needs a ceiling

Caps per token, a removal time on every machine and retry-safe rentals make GPU spend something you can delegate.

Everything a training run needs

Bring your image, keep your checkpoints and see what you spend, from the console, the CLI or the API.

Templates from any image

Start machines from any Docker image, with its environment, entrypoint and ports saved as a template.

Volumes

Keep datasets and checkpoints on volumes that outlive the machine.

SSH keys

Add your public keys once and every machine gets them.

Removal schedules

Set when a machine goes away at rent time, move it later, or cancel it.

Usage and billing

Cost by week or month, per machine, against one prepaid balance.

One spec behind everything

The console, the CLI and the MCP tools all run on the same OpenAPI spec.

Pay for GPU time. Nothing else.

Arrakis is prepaid. Add funds, rent at the hourly price shown, and each machine pays for itself an hour at a time while it runs.

No subscription: No seats and no minimum commitment.
The price you see: The hourly price shown when you rent is what the machine is billed at.
An hour at a time: A rental pays its first hour up front and each later hour as it begins.
When it runs out: The machine that cannot pay its next hour stops and rentals pause, so you never owe more than you added.
Paying: By bank transfer while in early access. Card payments are next.

Built for tenants who don't trust each other

Each guarantee below is enforced in the API and covered by tests. Machines run with independent hosts, so keep regulated data off them.

Hashed, scoped tokens

Stored as SHA-256 hashes, scoped to read, rent, manage or billing, and revocable at any time.

No silent escalation

A token can only create tokens with scopes it already holds, and only if it may manage.

Single-use bootstrap

A new account's first token needs a one-time secret that expires after 24 hours.

Tenant isolation

Every request is scoped to the account behind its token. Other accounts' machines are invisible and untouchable.

Funds nobody can fake

A balance grows only when our team confirms a payment. No account can credit itself.

Spend caps for agents

A capped token cannot rent past its ceiling or without a removal time.

Arrakis at a glance

The short version, for people and for the assistants that answer their questions.

What it is: A GPU rental service that finds the lowest live price across independent GPU markets and rents machines on one prepaid balance.
Who it is for: AI engineers, researchers and the coding agents they work with.
GPUs: H100, H200, B200, B300, A100, RTX PRO 6000, RTX 5090, RTX 4090, L40S, RTX 6000 Ada, RTX A6000, RTX 3090 and more, depending on live supply.
How you rent: Web console, CLI, REST API, or MCP for agents at /mcp.
Pricing: Pay the hourly price shown when you rent, from a prepaid balance, an hour at a time. No subscription and no minimum.
Agent safety: Spend-capped tokens, a required removal time on every rental, and retry-safe rentals with idempotency keys.
Access: SSH as root into machines running any Docker image. Volumes for data and checkpoints.
Live market: 148 GPUs on 53 machines, H100 from $1.30 per GPU-hour (13 Oct 2026, 16:19 UTC).
Billing: From start to removal at the price shown when you rent. A started hour counts as a full hour.
MCP endpoint: https://arrakis.goodailabs.com/mcp, streamable HTTP with a bearer token.
Status: Early access. Built by Good AI Labs (goodailabs.com).

Questions people and agents ask

Anything else is in the API reference or the agent guide.

What is Arrakis?

Arrakis is a GPU rental service that finds the lowest live price across independent GPU markets and rents machines on one prepaid balance. People rent from a web console or CLI, and AI agents rent over MCP or the REST API with spend-capped tokens.

Can my agent overspend?

Not past what you allow. Give it a token with a spend cap. Every rental it makes must carry a removal time, and a rental that would pass the cap is refused before it starts.

How much does it cost to rent a GPU?

Prices change with the market and are shown live per GPU-hour on the GPU prices page and in the console. Right now (as of 13 Oct 2026, 16:19 UTC), 148 GPUs are available on Arrakis across 53 machines. The lowest H100 price is $1.30 per GPU-hour and the lowest RTX 4090 price is $0.46 per GPU-hour. You pay the hourly price shown when you rent, from a prepaid balance, with no subscription. A machine bills from start to removal, and a started hour counts as a full hour.

What is the cheapest way to rent an H100?

Search every market at once. Arrakis lists every available H100 across independent GPU markets, cheapest first, and rents it on one prepaid balance. As of 13 Oct 2026, 16:19 UTC, the lowest price for an NVIDIA H100 on Arrakis is $1.30 per GPU-hour, with 29 GPUs available on 6 machines.

Who makes Arrakis?

Good AI Labs builds and runs Arrakis. It is in early access.

How do I pay?

Request a bank transfer in the console, and your balance grows when the payment arrives. Card payments, with automatic top-ups from a saved card, are next.

Where do the machines run?

With independent GPU hosts in many countries. Search shows each machine's country and region, so you can choose.

What if a machine fails?

Each machine shows a reliability score. Terminate it and rent another, and keep checkpoints on a volume so nothing is lost.

Can I bring my own image?

Yes. Any Docker image works, saved as a template with its environment, entrypoint and ports. You SSH in as root.

Is there an SLA?

No. Arrakis is in early access and the hosts are independent. Don't run anything on it that needs one.

How do I get an API token?

Sign in to the console and create one under API & agents, or bootstrap your first one from the CLI with your account's one-time secret.

What happens when my balance runs out?

Each machine pays for one hour at a time from your balance, as the hour begins. When the balance cannot pay a machine's next hour, Arrakis stops that machine and pauses new rentals, and your other machines stop when the hour they already paid for ends, so you never owe more than you added. The console warns you before that happens, adding funds opens rentals again, and volumes are kept.

We're hiring people who like hard infrastructure.

Take the Spice Test. All it needs is the API.

See open roles

Rent your first GPU in minutes.

Start renting See GPU prices

The GPU market, made safe for agents.