01
Explore live GPU inventory
Browse available GPU instances by price, region, and hardware specs. Compare options before committing to an hourly rate.
GPU cloud
High-performance compute,integrated inside your platform.
Provision NVIDIA GPU instances for AI model training, inference, and rendering straight from Hubfly space — attached to your projects and protected by a dedicated wallet that stops instances before you owe a dime.
Project workflow01 / 03
01
SourceJupyter and SSH ready
02
ControlSeparate GPU wallet
03
ReleaseStops at zero balance
Hubfly space projectNetwork · volumes · domains · team access
One project model
How it works
From offer selection to running notebook in five steps.
Hubfly space integrates with GPU provider markets, offering live inventory rates rather than fixed static markups.
- 01
Search available offers
Filter market offers by GPU model, region, storage, and hourly cost.
H100 · A100 · RTX 4090 - 02
Select template and storage
Choose your disk size and select a Jupyter or SSH-ready environment image.
Jupyter · SSH - 03
Fund your GPU wallet
Transfer balance intentionally from your main account to fund compute hours.
gpu_wallet_transfers - 04
Provision instance
Your instance boots up connected to your project network.
provider: vast - 05
Start, pause, destroy
Pause instances when you finish training runs. You only pay for active runtime.
Micro-invoiced
Offer example
NVIDIA A100 · 80 GB
4× GPU · EU-West · 500 GB disk
$3.20/hr
provider rate × 1.20
JupyterSSHWallet required
Wallet separation
Account balance
$48.20
Containers, storage, bandwidth
GPU wallet
$15.00
GPU instances only
Transferring funds between balances is always a conscious choice, never automatic.
Capabilities
Everything your AI workloads need, pre-configured.
02
Jupyter and SSH templates
Launch notebook-ready or SSH-accessible environments from pre-configured templates without spending hours setting up CUDA drivers manually.
03
Dedicated GPU wallet
GPU costs are funded independently through a separate GPU balance, topped up deliberately so heavy AI workloads never drain your primary account balance.
04
Project or global scoping
Attach GPU instances to specific projects so backend containers connect over private network aliases, or manage them globally at the account level.
05
Transparent pricing markup
We display raw provider prices alongside our exact platform markup, so you always know what you're paying.
1.20× default
06
Full lifecycle control
Provision, start, pause, sync, and delete GPU instances right from the dashboard — using the same control flow you use for app containers.
Spend safety
Never get hit with runaway GPU bills.
GPU hours are often the easiest place to accidentally rack up large cloud bills. Our separate wallet system and automated shutoff rules protect you against accidental overspending.
01
Low-balance alerts
Our scheduler monitors your GPU wallet balance every five minutes and triggers notifications as funds run low.
02
Automatic shutoff at zero
If your GPU wallet balance hits $0, running instances receive an automatic stop signal, preventing surprise charges overnight.
03
Micro-invoices and daily summaries
Usage is calculated in micro-invoices aggregated into daily summaries, allowing you to trace every penny spent down to the hour.
Because GPU workloads draw exclusively from their dedicated wallet balance, an active instance can only ever consume what you've allocated to it. Once that wallet hits zero, the instance stops automatically without drawing funds from your main account.
GPU cloud
Train AI models with complete budget confidence.
Search live hardware offers, fund your GPU wallet, and spin up instances that automatically pause before causing surprise charges.