Skip to content

Decentralized GPU computing power marketplace

Every idle GPU.
One supercomputer.

Rent Resource pools graphics cards sitting idle in homes, offices and small data centers into a single on-demand marketplace. Rent one in a few clicks, or get paid while yours rests.

Rent a GPUEarn with your GPU
  • Hourly pricing, metered to the second
  • No servers to configure
  • No contracts

The problem

The world of GPUs has a two-sided dilemma.

$1,500–$40,000

Per modern GPU. Buying is out of reach for most individuals and developers.

Millions of GPUs

In gamers' rigs, offices and small data centers, sitting completely idle.

  • High cost to purchase

    GPU prices and maintenance costs are extremely high.

    Rent by the hour and pay only for the time you use.

  • Cloud queues and complexity

    High prices, strict quotas and long-term contracts on the big clouds.

    Independent providers compete for your job. No contracts, no servers to set up.

  • Idle hardware

    GPU owners have no simple way to monetize their computing power.

    Connect your card and earn passive income while it sits idle.

  • A single GPU is not enough

    Training AI models takes multi-GPU clusters.

    Spin up a ready-made multi-GPU Ray cluster in one request.

The solution

The Airbnb model, built exclusively for computing power.

Provider

The GPU owner connects a graphics card and earns passive income while it remains idle.

Rent Resource

Securely connects both sides and takes a commission on each transaction.

Customer

Rents the GPU they need in a few clicks. No servers, no configuration, no contracts.

Competition among independent providers is what keeps prices below the cloud giants.

For builders

From marketplace to running code in a few clicks.

Compare on real numbers

Every listing shows the GPU model, measured TFLOPS, the hourly price and a reputation score.

Sample listings. Live prices are set by providers.

A ready JupyterLab, one click away

A token-secured environment opens in your browser. Write code straight on the GPU.

Multi-GPU Ray clusters

Ask for N GPUs. We pick the cheapest available, start a head and workers, and hand you a ray:// address.

Or script the whole thing

A REST API, a zero-dependency Python SDK and a CLI, for when clicking stops scaling.

# see what is available
$ curl https://apis.rentresource.cloud/gpus

# start a 4-GPU Ray cluster
$ curl -X POST https://apis.rentresource.cloud/clusters \
    -H "X-API-Key: $KEY" -H "Content-Type: application/json" \
    -d '{"gpu_count": 4}'

For GPU owners

Your GPU sleeps most of the day. Let it pay rent.

Provider app · preview
Detected GPUNVIDIA RTX 4090 · 24 GB
StatusStopped

No desktop? One command on Linux:

curl -fsSL https://apis.rentresource.cloud/install.sh | sudo PROVIDER_API_KEY=<your-key> bash
  • Desktop app for Windows, macOS and Linux with a single Start/Stop button
  • Set your own price and watch earnings on a live dashboard
  • Withdraw through Stripe, or as CGT reward tokens on Solana
Open the provider dashboard

Estimated gross per month

$120

A calculator, not a promise. Before platform commission and electricity; real earnings depend on demand for your GPU.

Core product

How a GPU rental works, in five steps.

  1. 01

    Selection

    You choose a GPU from the marketplace and the image you want to run.

  2. 02

    Verification

    Your payment method and the security policy are checked before anything starts.

  3. 03

    Startup

    The provider's agent opens a secured Docker container.

  4. 04

    Connection

    You connect to a ready JupyterLab environment through a secure tunnel.

  5. 05

    Billing

    When you stop, billing settles automatically for the time you used.

session.log · illustrative

rent gpu=12 image=pytorch/pytorch:latest

payment method on file

image policy passed

agent: docker run --cap-drop=ALL ...

container running

reverse tunnel open

JupyterLab ready, token issued

rental stopped

billed for time used, provider credited

Trust and security

How do you trust a stranger's computer?

You make it keep proving itself. The results build a trust score for every machine, shown right on its marketplace listing.

  1. 01

    Ed25519 signature verification

    Each agent has its own cryptographic key. The server sends a random challenge, and the agent signs and returns it.

  2. 02

    Proof-of-compute

    A SHA-256 chained deterministic problem proves the node actually performed the computation.

  3. 03

    GPU benchmark

    A real CUDA test during registration, so advertised hardware is shown with measured TFLOPS.

attestation · illustrative

serverchallenge nonce=7c1e04ab…

agented25519 sig=9fd2a61c…

serververified signature matches registered key

servercompute sha-256 chain, 200,000 rounds

agentresult e40b77d2…

serververified result reproduced

Shown on the listingTrust score updated

Customer code runs in a locked-down sandbox, protecting the provider's machine:

  • --cap-drop=ALL
  • no-new-privileges
  • CPU, memory and process limits
  • Optional gVisor isolation
  • Crypto-miner images blocked
  • Provider LAN and cloud metadata guarded

Technical maturity

Built for the day a node disappears.

  • Fault tolerance

    If a node dies, its rentals are closed, its GPUs are freed and the customer is not charged.

  • NAT and tunnels solved

    FRP reverse tunnels make a provider on home internet reachable from anywhere.

  • Persistent storage

    A /workspace disk that survives restarts, plus S3-compatible storage for datasets and model weights.

  • Horizontal scale

    Multiple backend replicas over Redis pub/sub, with no sticky sessions.

  • Observability

    Prometheus-style metrics and structured logs across the control plane.

  • Encrypted by default

    Automatic TLS on every endpoint, with hashed API keys and optional two-factor sign-in.

Traditional payments

Stripe

Customers pay by card through PaymentIntents and providers are paid through Stripe Connect. An internal ledger tracks who earned what, commissions and balances.

Blockchain settlement

CGT on Solana

Providers can settle in CGT, an SPL reward token. Withdrawals are on-chain transactions confirmed with a signature.

CGT is a platform reward mechanism. It is not offered as an investment.

New direction · Coming soon

AAA games at maximum graphics, on the simplest laptop.

A gaming PC costs $1,000 to $3,000, and millions of people are left with old machines. With Rent Resource, the game runs on a powerful remote GPU. A low-latency video stream comes to you and your controls go back.

  1. 01

    Select a machine

    Choose a PC with a GPU like an RTX 4080 from the Gaming section.

  2. 02

    One click, one session

    The existing rental, tunnel and billing infrastructure does the work.

  3. 03

    Connect and play

    In a browser or lightweight client. No drivers, nothing to configure.

  4. 04

    Hourly billing

    When the session ends, billing works just like a GPU rental.

The foundation is already built

  • GPU fleet and agent
  • Hourly billing
  • NAT reverse tunnels
  • Isolation and trust system
  • Low-latency video streaming layerin development

For gamers

Instead of a $1,500 computer, a top-tier experience at an affordable hourly rate. A local network of providers means play happens closer to you.

For providers

A whole new audience of renters for your GPU, far beyond AI developers.

Roadmap

Where we are and where we are going.

Ready

The platform, working end to end

  • Single GPU rental
  • Multi-GPU Ray clusters
  • Stripe payments and payouts
  • CGT settlement on Solana
  • Attestation and trust scores
  • NAT reverse tunnels
  • Desktop app, CLI and SDK
  • Admin console

Coming soon

Cloud Gaming

Renting a GPU PC for gaming, on a new low-latency video-streaming layer. It opens a second, mass market on the same infrastructure.

See how it works

Planned

Going deeper

Multi-host Ray clusters
Real clusters spanning many separate machines.
TEE and zkML
Cryptographic proof of the computation itself.
Token governance
Platform decisions shaped through the CGT token.
Staking and slashing
Providers back their reliability with a stake.

Put idle power to work.

Rent a GPUEarn with your GPU