a saved workflow, run on 48 GB, outputs kept in your grove
Create images and video
Run the workflows you already use, with your own models and LoRAs, on more memory than a desktop card has. Your workflows stay saved.
Early access · invite-only pilot
Your private AI server.
No hardware to buy.
Create images and video with your own models, run open models on your own documents, and fine-tune when you're ready. On a private server in the cloud, with your files and setups saved between sessions.
Your grove is your saved AI workspace. Wake it when you need a GPU. It sleeps when you're done.
My creative studio
interactive preview
What did the landlord agree to about the deposit?
Returned within 14 days of the inspection, minus documented damage only.
lease-2026.pdf · p.4
Saved in your grove: 3 documents · 2 models · 1 workflow · 12 GB
- Saved
- Starting
- Ready · 48 GB
GPU time starts when your app is ready. Storage stays while it sleeps.
What you can do
Three things people do in a grove.
Summarise the three risks the report flags for Q4.
1. Supplier concentration in one region. 2. A pricing clause that resets in November. 3. Head-count above the hiring plan. q3-report.pdf · pp. 12, 19, 23
Run open models, privately
Any open model from Hugging Face, on your own machine. Chat with contracts, notes or research and get the page it came from. No provider decides what you can ask.
dataset1,200 support replies
trainQwen3 8B · LoRA
resulta model that answers like you · export as GGUF
Fine-tune a model
Teach an open model your tone, your product or your data. Bring a dataset, train, evaluate, export. Advanced, and worth it.
How it works
What stays when you stop.
Your grove keeps your files, models and app settings. A GPU attaches when you wake it and detaches when you're done. Storage stays either way. GPU hours only count while the GPU is attached and your app is ready.
- always
- files, models, workflows, and the history your apps keep
- on wake
- a 24, 48 or 96 GB GPU, attached in under a minute
- while asleep
- storage only, no GPU hours
How billing works
One flat plan. The meter only runs while a GPU is ready.
Your plan includes GPU hours and storage. Hours count while a GPU is attached and your app answers. Starting doesn't count. Asleep doesn't count. Storage is kept the whole month. Pilot prices are set with the first invitees; the rules are fixed now.
- Ready counts against your included hours.
- Starting never counts. A failed start costs nothing.
- Asleep costs nothing beyond the storage your plan includes.
An example month
30 days · one grove · hours are illustrative
62 GB kept all month
1102030
- Counted
- 19.5 h ready
- Not counted
- 0.6 h starting · 699.9 h asleep
- Storage
- 62 GB · kept, within plan
- Extra hours
- none this month
Providerruns the hardware · named and located before the pilot
Nodegrovesees GPU minutes, storage size, health, billing
GPUdedicated or shared by plan · shown on your card
Your environmentisolated · only your processes
Your prompts, outputs, filesnot logged by us · on your disk · yours to export and delete
Data & privacy
Where your data lives. Who can access it.
A rented environment still has an operator and a provider. Read it inside out: the centre is yours alone, and each ring outward can see less of what matters. Nothing you type goes to a third-party model API unless you add your own key on purpose.
Every layer, spelled out →Questions people ask
Before you ask for an invite.
- Is this a private AI server or a shared cloud?
- Both, honestly. Your environment is yours alone: nobody else's processes run inside it. Whether the GPU is dedicated or shared depends on the plan, and your card always says which. The physical machine belongs to a provider, like any cloud. The data and privacy page says exactly who can see what.
- What happens when I stop using it?
- Your grove goes to sleep after a period without GPU activity. Files, models and settings stay. GPU hours stop counting. Storage continues within your plan.
- Can I run any model?
- Any open model that fits the memory you chose: 24, 48 or 96 GB. Hugging Face and Civitai downloads work with your own tokens.
- Do you log my prompts?
- No. We see GPU minutes, storage size and health checks. Your apps keep their own history on your disk, and you can delete it.
- How is this different from renting a GPU by the hour?
- Hourly GPU clouds give you a fresh machine and a running meter. A grove is one saved place that persists, with the apps installed, a flat plan, and hours that only count while your app is ready.
Early access
Tell us what you'd like to try.
We're inviting a small group into the pilot. One email when your invite is ready. No newsletter.