If you own an NVIDIA GPU, it spends most of the day doing nothing. The card sits warm and idle while AI teams, researchers, and 3D studios pay every hour to access exactly the kind of hardware you already have. Paralon turns that idle time into income: the moment someone rents your GPU, the meter starts; the moment they stop, it stops. You are paid per minute, in crypto, with no KYC.
This guide walks through what you can earn, how the marketplace works, and how to add your card.
What is Paralon
Paralon is a decentralized NVIDIA GPU marketplace. Renters launch containerized workloads — AI training, inference, rendering, mining, scientific compute — on real verified hardware contributed by independent providers around the world. Providers list their GPU once and earn whenever a renter is on it.
Everything runs in an isolated Docker container with the renter's own image. The renter never touches your host filesystem and you never touch theirs. Your only role is to keep the node online and healthy.
Which GPUs can earn
Paralon supports the full NVIDIA range, from consumer gaming cards to data-center accelerators. Anything from an RTX 3060 upward is listable today. A practical baseline:
- NVIDIA GPU with 8 GB VRAM or more
- Linux (Windows hosts work via WSL2, but Linux is cleaner)
- Docker with the NVIDIA Container Toolkit installed
- Stable internet, ideally 50 Mbps upload or better
- The node stays plugged in and online
Older or low-memory cards may still list, but most renter demand sits on cards with 12 GB+ VRAM, where modern models actually fit.
How much your GPU earns per hour
Listed prices on Paralon are set on the marketplace and shown live on the pricing page. Providers receive 80% of the rental price; Paralon keeps 20% as platform fee. There are no other deductions and there is no minimum payout.
Here is what that looks like across the catalog today. The "Provider per hour" column is your net after the platform fee.
Consumer GPUs
| GPU | Listed price | Provider per hour | At 100% utilization (24h) |
|---|---|---|---|
| RTX 3060 | $0.078/hr | $0.062/hr | $1.50/day |
| RTX 3090 | $0.180/hr | $0.144/hr | $3.46/day |
| RTX 4070 Ti SUPER | $0.170/hr | $0.136/hr | $3.26/day |
| RTX 4080 | $0.180/hr | $0.144/hr | $3.46/day |
| RTX 4090 | $0.540/hr | $0.432/hr | $10.37/day |
| RTX 5070 | $0.190/hr | $0.152/hr | $3.65/day |
| RTX 5070 Ti | $0.220/hr | $0.176/hr | $4.22/day |
| RTX 5090 | $0.720/hr | $0.576/hr | $13.82/day |
Workstation and data-center GPUs
| GPU | Listed price | Provider per hour | At 100% utilization (24h) |
|---|---|---|---|
| RTX A6000 | $0.370/hr | $0.296/hr | $7.10/day |
| L40S | $0.680/hr | $0.544/hr | $13.06/day |
| A100 40 GB | $0.420/hr | $0.336/hr | $8.06/day |
| A100 80 GB | $0.550/hr | $0.440/hr | $10.56/day |
| H100 PCIe | $1.900/hr | $1.520/hr | $36.48/day |
| H100 SXM | $2.600/hr | $2.080/hr | $49.92/day |
| H200 | $3.400/hr | $2.720/hr | $65.28/day |
These are gross figures at full utilization. Real income depends on how often your card is rented, which depends on listing price, location, network speed, and how well your node passes verification. A reasonable working assumption when planning is 40 to 70% utilization once the node has built a track record on the network. Multi-GPU nodes earn proportionally — a node with four RTX 5090s lists each card independently and is paid for each rental in parallel.
Per-minute billing — paid for what you actually serve
Paralon bills by the minute, not the hour. The moment a renter starts a container on your GPU, the meter begins. The moment they stop, it ends. There is no rounding up to the hour, no minimum billing window, and no charge to either side for idle time between rentals.
For you as a provider this matters in two ways:
- Short jobs still pay. A renter who spins up a card for a 14-minute inference test pays for 14 minutes and you earn for 14 minutes. Nothing is lost to rounding.
- You can take the GPU back at any time. You decide when your node is online. There is no contractual lock-in.
Crypto payouts, no KYC
Earnings accumulate in your provider balance every minute a rental runs. Payouts are sent in crypto — USDC or USDT on Solana — directly to a wallet address you control.
- No KYC, no ID verification, no bank account. You sign in with email or wallet, set a Solana payout address, and you get paid.
- No fiat off-ramp on Paralon's side. What you do with the USDC/USDT after it lands is up to you.
- Payouts are processed from your
earnings_balanceindependently of any credits you also use to rent other GPUs. The two balances do not mix.
This is the part most providers actually care about: the income exits the platform the same way it was earned, in crypto, without a bank or a KYC form in the way.
Referral program — stacking extra income on top
Every rental on Paralon goes through a 20% platform fee. 15% of that platform fee is paid out as referral commission — in other words, 3% of the total rental price is paid to the referrer of each side of the trade.
It works like this:
- The renter can apply a referral code within 15 days of signup. Whenever that renter pays for compute, the person whose code they applied earns 3% of the rental price, for 12 months from the day the code was applied.
- The provider can do the same. If you joined Paralon with a referrer's code, that referrer earns 3% of every rental on your nodes for 12 months.
- Both sides can stack. If the renter and the provider were both referred, both referrers earn 3%, the platform keeps 14%, and the provider still receives the full 80%. Referral commission is paid out of the platform fee, never out of provider earnings.
A few things to know:
- The 15-day window to apply a referral code starts at signup. Apply it early.
- A user cannot apply their own code, and circular referrals (A refers B who refers A) are blocked.
- Referral earnings credit to the referrer's account per-minute, the same way provider earnings do, and are withdrawable in crypto.
If you bring other providers onto the network — friends with a 4090 at home, a mining rig owner who wants to repurpose hardware, a small studio with a spare RTX A6000 — every rental on their nodes pays you a slice for a year.
Practical notes before you list
- Stable power and cooling are more important than peak performance. A node that crashes mid-rental is the single biggest reason a card stops getting picked.
- Upload bandwidth matters. Renters pull container images, datasets, and model weights through your connection. 50 Mbps up is workable, 200 Mbps+ is comfortable.
- Multi-GPU rigs earn more efficiently per kWh. Splitting the host overhead (CPU, RAM, NVMe, network) across four cards lifts the margin per GPU.
- Run the node on a dedicated machine where you can. You can run on a daily-driver desktop if you pause earning during gaming sessions, but a dedicated headless box is the lower-friction setup.
Ready to monetize your GPU
Sign in at paraloncloud.com, add a node from your dashboard, and your card is on the marketplace within minutes of passing verification. Check the pricing page for the live rate on your specific GPU, and start earning USDC or USDT on Solana from the next rental on.
