Sell everyRTX 5090up to 2.5 times.Same hardware, same rack, but more revenue.
Your GPUs, your call. Place them where you want to earn: tokens, GPU per hour, GPU per month, reserved or offtake.
Sell your GPU five ways at once
Powering AI teams that ship
01Five ways to earn
One GPU. Five ways to sell it.
Offtake locks your hardware into one buyer at one price. On Packet.ai the same GPU can sell five ways: as tokens, by the hour, by the month, reserved and as offtake. The higher up the list, the more margin per GPU-hour. You decide where each GPU earns.
1Your GPUs
2The scheduler
3Demand channels
- Packet.ai
- Neocloud partners
- Offtakers
4Five ways to earn
Relative margin, not a rate card.
02Process
From application to first payout.
Typically under seven days to your first marketplace revenue. Nothing takes public traffic until it has passed hardware testing.
Apply
Submit your infrastructure details. Reviewed within 2 business days.
Verify and agree terms
We check GPU model, network, power and environment, and agree plain-language, month-to-month terms before technical access.
Install the agent
The lightweight hosted·ai agent goes on your hosts. You keep root and physical access.
Pass hardware tests
Every node is tested before it takes public traffic.
Go live
Your GPUs list with a New Host badge. Track occupancy, $/hr and earnings per GPU.
Get paid
Bi-weekly payouts in dollars, by direct deposit to your bank account.
03Who it is for
Built for operators who count idle hours.
If you think in payback, cost per MWh and fill rate, this is written for you.
Racks that wait between tenants
You think in payback, cost per MWh and idle time.
How do I earn on capacity between anchor tenants?
- Earnings and utilization per GPU, so your model runs on real hours
- Revenue per kW in the calculator, because power is your binding constraint
- No sales team and no long-term contracts
Capacity without more customers to manage
You think in capacity planning and fill rate.
Can I add GPUs without adding accounts to run?
- Add from 1 to thousands of GPUs and scale down any time
- Keep your anchor tenants; the scheduler sells the gaps between them
- Utilization, $/hr and occupancy for every GPU in the dashboard
A home for the spare capacity
You think in reserved customers first, spare capacity second.
Where do I put what my customers are not using?
- Keep root, BMC and physical access
- Packet.ai handles customers, billing and support
- A lightweight agent, not a platform migration
04Packet.ai vs competitors
Packet.ai vs competitors.
How supplying GPUs to Packet.ai compares with listing them on other GPU marketplaces, based on published provider terms.
| Topic | Packet.ai | Others |
|---|---|---|
| Ways to earn from one GPU | Five ways: tokens, GPU per hour, GPU per month, reserved and offtake | Usually hourly rental, sometimes with reserved or interruptible tiers |
| Revenue making | 2.5 times on average, proven over six months on Packet.ai (hosted·ai) | Utilization is usually not published openly |
| Customers you get | Verified businesses with KYC and acceptable-use enforcement, brought to you | Often an open marketplace where renters sign up with a card or crypto |
| Payouts | Bi-weekly payouts in dollars. Predictable income every two weeks, by direct deposit to your bank account | Varies by platform. Minimum payout balances are common, and bank transfers are not always offered |
| Who sets the price | Tier-based rate band. Your rate card is shared at onboarding | Varies. Some let you set your own price, others set it for you |
05Where providers are
Your racks, wherever they are.
Packet.ai already has providers in 10+ major countries.
06FAQ
Questions before you apply.
How much will I earn per RTX 5090?
Provider rate cards are tier-based by GPU and shared during onboarding, based on your hardware and region. Providers earn the majority of customer-billed revenue after platform fees. The provider calculator shows your net with your own power costs.
Will I really sell every GPU 2.5 times?
Not necessarily. 2.5 is the average number of times each GPU has been sold on Packet.ai. It is not a guarantee for any single card or rack, because it depends on hardware, location, demand and contract terms. The provider calculator lets you lower the multiple and shows your net, including where you break even.
What do I need to qualify?
Datacenter-grade NVIDIA GPUs in a real datacenter or comparable hosting environment, with stable 1 Gbps+ networking, redundant power (N+1 is ideal), stable cooling, 99%+ uptime capability and hosts running Ubuntu Server 22.04 or 24.04 LTS. Packet.ai onboards the hardware operator, not resellers or brokers, and does not accept home-lab or consumer-grade setups. Every node is then tested, including a 24 to 48 hour stress test, before it takes public traffic.
Do I keep control of my hardware?
Yes. You keep full root, BMC and physical access. The hosted·ai provisioning agent orchestrates VMs through a defined API, and you can pause it, audit it or pull the plug at any time. It does not transmit customer workload data back to Packet.ai. Packet.ai's 24/7 team handles all tier-1 and tier-2 customer issues and only contacts you for hardware-level incidents, such as a node down or a fan failure.
How and when do I get paid?
Payouts are bi-weekly and paid in US dollars, so your income is predictable every two weeks. Packet.ai handles billing, invoicing, payment collection and currency conversion, and pays you by direct deposit to your bank account.
What if I want my GPUs back?
Withdraw capacity at any time with no financial penalty. If a GPU is occupied, Packet.ai asks for 14 days of notice so it can migrate the customer workload smoothly. There are no minimums or long-term contracts: terms are plain-language and month-to-month, agreed before technical access.
Ready when you are
Fill the hours.
Applications are reviewed within 2 business days. Want to ask something first? Email providers@packet.ai. Replies typically arrive within one business day.
- No minimums or long-term contracts
- Withdraw any time, 14 days notice if occupied
- A live dashboard of your whole portfolio