Most graphics cards spend most of their lives doing nothing. A 4090 bought for gaming is idle while its owner sleeps, works, or is simply not playing anything. A workstation card in a studio is idle every night and every weekend. The hardware is already paid for and the depreciation is already running.
ParalonCloud turns that idle time into two kinds of income at once, and they are independent of each other. This is how each one works.
Track one: USDC when someone rents your card
When a renter takes your node, you keep 80% of the rental price. The platform keeps 20%.
That is not a promotional rate, a tier you unlock, or something that changes once you have earned enough. It is a constant. It is the same for the first node you connect and the fiftieth, and referral rewards come out of the platform's 20% rather than yours.
Billing runs per minute. You start earning in the first minute of a rental and keep accruing until the renter stops — no minimum session, no rounding up to the hour in either direction. The full mechanics are in Rewards & Earnings and Credits & Billing.
What each card is worth per hour
The renter pays the catalogue rate for your GPU, or whatever rate you set yourself. You receive 80% of it.
| GPU | VRAM | Renter pays | You keep |
|---|---|---|---|
| H100 PCIe | 80 GB | $1.80 | $1.44 |
| A100 80GB | 80 GB | $0.55 | $0.44 |
| RTX 5090 | 32 GB | $0.46 | $0.368 |
| RTX 4090 | 24 GB | $0.39 | $0.312 |
| RTX A6000 | 48 GB | $0.37 | $0.296 |
| RTX 5080 | 16 GB | $0.27 | $0.216 |
| RTX 3090 | 24 GB | $0.13 | $0.104 |
You are not stuck with the catalogue. You can price each node yourself, above or below the default — the reasoning, including when undercutting actually pays and when it just costs you, is in Set Your Own GPU Rental Price. The mechanics are in Node Pricing.
Earnings land in your balance as you go. You can withdraw them as USDC on Solana to your own wallet, or convert them into platform credits at 1:1 and spend them on the network yourself — both from the Payouts page.
Track two: points, whether or not anyone rents you
Rental demand is not something you control. Uptime is. So the second track pays you for the thing you can actually influence.
Season 0 points accrue on uptime alone. No rental required, no renter involved. The formula is deliberately simple:
1 GB of VRAM = 1 point per minute.
| GPU | VRAM | points/min | points/day |
|---|---|---|---|
| RTX 3090 / 4090 | 24 GB | 24 | 34,560 |
| RTX 5090 | 32 GB | 32 | 46,080 |
| RTX A6000 | 48 GB | 48 | 69,120 |
| RTX PRO 6000 Blackwell | 96 GB | 96 | 138,240 |
| H200 | 141 GB | 141 | 203,040 |
Multiple GPUs stack — a machine with two 5090s earns 64 points a minute for as long as it stays online and verified. CPU-only nodes earn on a separate formula.
Season 0 is running now and rewards early providers. Standings are public on the Leaderboard, and the season's terms are in The ParalonCloud Incentivized Testnet.
The two tracks run simultaneously. Points do not reduce your USDC and USDC does not reduce your points.
Why we want high-VRAM cards most
This part is a request rather than a description.
24GB has become the number that just barely does not work. We serve a 27B model at 4-bit on RTX 4090s, and the arithmetic is uncomfortably tight: the weights alone are 18.7 GB, peak activation adds another 1.76 GB, and what remains on a 24GB card is under two gigabytes for everything else.
That single constraint cascades. It forces --enforce-eager, which switches off CUDA graphs, which costs throughput on every token. It leaves a KV cache small enough to hold barely more than one full-context request, so the card cannot batch the way it should. We wrote the full measurement up in The Faster Card Was Half the Speed.
Every one of those compromises is a memory problem, and every one of them disappears on a larger card.
An RTX 5090 (32 GB) turns that model from a squeeze into a comfortable fit — CUDA graphs on, a KV cache that can actually batch, and roughly twice the memory bandwidth of a 3090.
An RTX PRO 6000 Blackwell (96 GB) is a different class of machine on a network like this. It holds models nothing else in our fleet can load at all, and it accrues 96 points a minute doing nothing but staying powered on.
If you own either — or an A6000, an A100, or anything with 48GB and up — you are the person this article is written for.
What we ask of your machine, and what we do not
No inbound port. Nothing on the internet ever connects towards your machine. The agent dials out and holds the connection open, which is why this works behind an ordinary home router with nothing configured and no static IP. The design is written up in A URL That Lives as Long as the Rental.
We measure your card rather than take its word for it. Every node runs a verification container that benchmarks the GPU it claims to have, because a network that pays for hardware has to check the hardware exists — see How We Prove a GPU Is Real. This protects honest providers considerably more than it inconveniences them.
The agent updates itself, in stages and with a rollback path, so you are not asked to run a command every time we ship. The mechanism, including the parts that went wrong on the way, is in A Field Guide to Updating a Network You Don't Own.
One piece of advice that costs you nothing and earns you more: if your card sits in a Windows machine, run the node on native Linux if you can. We measured two RTX 4090s that verify as near-identical hardware, one under WSL2 and one on native Linux. The WSL2 machine served inference 48% slower — same container, same model, same arguments. That is throughput you do not get paid for.
How to connect a card
Four pages, in order:
- Prerequisites — what your machine needs before you begin.
- Hardware Requirements — supported GPUs, driver versions, minimum specs.
- Add a Node — the actual install. It is a Docker container and a token.
- Node Pricing — set your rate, or keep the catalogue default.
Then Payouts to link a Solana wallet, and Referrals if you know other people sitting on idle cards.
The provider signup path starts at Become a Provider, and you can watch the live fleet — what is online, where it is, what it is running — on the Network page.
If you are running a 5090, an RTX PRO 6000, or anything with more memory than a 4090, we would like to hear from you directly — those are the cards the network needs, and we would rather talk than have you guess whether it is worth connecting.
