The phrase "Binance funds to rent io.net" needs to be broken down first: Binance does not offer GPU rental services, and io.net is not a product under Binance. What you can do through Binance is buy IO tokens, but IO tokens cannot directly pay for GPU rental fees. You need to first convert them into IO Credits or complete payment through a payment channel. This path involves two asset conversions and at least one intermediate step, so it is not a "one-click rent a GPU on Binance" experience.
The world's largest cryptocurrency exchange by trading volume,leading in security and liquidity.
New user benefit: Enjoy 20% off trading fees upon registration!
The Actual Path of Funds from Binance to io.net
io.net's PAYG deployment does not directly accept IO tokens as payment. The official FAQ clearly states: PAYG bookings can only be deducted from IO Credits. IO Coin, USDC, or credit cards all require purchasing Credits first before they can be used for deployment.
IO Credits are prepaid USD-equivalent credits. 1 IO Credit = 1 USD. The minimum purchase amount is $10, and payment is supported via USDC or credit/debit card (USD).
The path starting from Binance is:
Convert assets to USDC on Binance. If you hold IO tokens, you need to first sell them on Binance spot and convert them to USDC or stablecoins.
Purchase IO Credits on io.net. Choosing USDC payment will redirect you to Spherepay to complete the transaction. io.net also supports paying directly with IO Coin through Spherepay. This is the official channel for "paying with IO tokens," but it requires a separate payment process and cannot be deducted from the IO Credits balance.
Deploy an instance using IO Credits.
Important note: IO Credits are valid for 12 months. Unused balances cannot be refunded or restored after expiration. If you do not plan to use them within a year, you need to calculate your actual needs before purchasing.
Deployment Models: Choosing Between VM and Containers
io.net Cloud offers two deployment forms: VM on Demand (virtual machine) and CaaS (containers). Both use PAYG billing.
VM on Demand provides a complete virtual machine environment, suitable for tasks that require SSH login, custom environments, or long-running operations. There are two image options:
General Purpose Image: Ubuntu 24.04, includes CUDA Toolkit and drivers, suitable for fully custom environments.
Data Science Image: Pre-installed with Python 3.12, Conda, CUDA 12.1, RAPIDS, ready to use out of the box for training or inference.
CaaS is suitable for standardized workflows where images are already packaged. When deploying, you need to fill in the container image address, startup command, traffic port (default 8000), and environment variables. One common pitfall: when referencing environment variables in the entrypoint, $ needs to be escaped as $$, otherwise deployment may fail.
Regardless of which one you choose, you need to specify the GPU model, quantity, and location when deploying. io.net's GPU pool covers consumer-grade to enterprise-grade. The official recommended selection logic is: use H100 for training, A100 for fine-tuning, and L40S for inference.
Stop Billing: When Is Money Actually Deducted
This is the real question that "stop billing" needs to answer. io.net's PAYG billing rules are different from most cloud platforms: the moment you deploy, one hour's fee is deducted.
At the moment of deployment: The system deducts the first hour's fee from IO Credits. If the balance is not enough for this first hour, the deployment will not start.
During operation: After the first hour, fees are deducted in batches at intervals of a few minutes, with deductions occurring slightly before actual usage time.
When you stop it yourself: After stopping from the instance page, billing stops immediately. However, the prepaid first hour is not refunded, whether you only ran it for a few minutes or used the full hour. The official FAQ states: if you stop it yourself, unused time is not refunded.
When it stops due to exhausted balance: When Credits are insufficient to cover the next deduction, the instance is terminated. Termination is different from stopping — termination removes the instance, and unsaved data will be lost. When multiple PAYG instances share one balance, the newest instance will be terminated first.
When it fails due to an io.net-side issue: The prepaid fee will be refunded. This is the only situation where the first hour's fee can be refunded.
So the essence of "stop billing" is: stopping no longer charges fees, but the prepaid hour at deployment will not be returned just because you stopped early. If you expect a task to run only for a dozen minutes, this one-hour prepayment is the actual cost floor.
How to Avoid Forced Termination Due to Exhausted Balance
io.net sends reminders via email and in-app notifications when the balance is low, with two levels: a reminder when the balance is low, and a final warning before termination. After receiving a reminder, you have time to top up or manually stop the instance to preserve your data.
You can also proactively check the consumption rate. The API endpoint Get Burn Rate returns three key numbers: the total hourly cost of all current PAYG instances (hourly_rate), how many more hours the Credits can last at this rate (runway_hours_estimate), and the number of running instances (cluster_count). Before deploying a task that will run for more than an hour, checking the runway estimate is more reliable than waiting for notifications to react.
Regarding data security, the official requirement is clear: back up the data you need before the balance runs out, because termination will directly remove the instance.
An Easily Overlooked Cost Factor
Prices for io.net PAYG instances may change during operation. If the hardware supplier adjusts prices, the new rate takes effect from the next deduction, and the portion already paid is not affected. Official notifications will be sent regarding price changes.
This means the hourly rate you see at deployment is not necessarily a fixed rate for the entire operating period. For tasks planned to run for several hours or more, leaving some margin in your budget is more realistic than deploying right at the balance limit.
Completion Criteria: What Counts as a Successful Deployment
The final criterion for a usable deployment is not "the page shows deployment successful," but that you can actually access the running instance via SSH (VM) or port (container), and the IO Credits deduction records match your expectations.
After deployment, it is recommended to do two things immediately: check the amount of the first deduction and the Credits balance change on the Usage & Billing page; if running a container, test whether the service responds using the port configured at deployment. Only after confirming these two points should you put your main tasks and data in.
If your balance is deducted by a large amount quickly after deployment, first check whether you accidentally selected a multi-container/multi-GPU configuration — the deployment summary page shows Total Cost, and it is worth reviewing once more before confirming.
The world's largest cryptocurrency exchange by trading volume,leading in security and liquidity.
New user benefit: Enjoy 20% off trading fees upon registration!
References
- io.net · Official Homepage, no update date indicated on the page; verification date: 2024-12-01.
- io.net · Pay-As-You-Go Usage Guide, no update date indicated on the page; verification date: 2024-12-01.
- io.net · IO Credits Usage Guide, no update date indicated on the page; verification date: 2024-12-01.
- io.net · Cluster Deployment and Payment Guide, no update date indicated on the page; verification date: 2024-12-01.
- io.net · Deploy VM on Demand Guide, no update date indicated on the page; verification date: 2024-12-01.
- io.net · Container Deployment Guide, no update date indicated on the page; verification date: 2024-12-01.
- io.net · Multi-GPU Workload Deployment Cheat Sheet, no update date indicated on the page; verification date: 2024-12-01.
- io.net · Burn Rate Query API Documentation, no update date indicated on the page; verification date: 2024-12-01.
- io.net · Usage and Billing Query Guide, no update date indicated on the page; verification date: 2024-12-01.
