Why Nodegrove exists.

The gap we are building for

If you want to run open models today, there are three honest options and none of them is a place you own.

Rent a GPU by the hour and you get a bare machine and a running meter. Every session starts from nothing: install the tools, pull the weights, rebuild the workflow, and lose all of it when you stop paying. The meter runs while you set up, and it runs while you think.

Buy the hardware and you get permanence, at the cost of a large amount of money, a machine that is obsolete in two years, a power bill, noise in the room, and a card sized for the model you wanted last year rather than the one released last month. Our build versus rent calculator exists because that decision deserves your own numbers rather than our opinion.

Use somebody else's hosted model and most of this goes away, along with the thing you wanted: control over the model, the workflow and where your work lives.

What is missing is the ordinary middle. A place that is yours, that stays as you left it, where compute attaches when you need it and stops costing you when you do not. Somewhere between a rented computer and an owned one.

What we are actually building

A private environment in the cloud that persists between sessions. Your files, your models, your installed tools and your settings stay put. A GPU attaches when you start working and detaches when you stop. You pick what you want to do — create images and video, run open models privately, fine-tune something — rather than picking a GPU specification and working backwards.

We are not building a ChatGPT replacement, and we are not reselling raw GPU hours. Both of those exist and are done well by other people. What we sell is control, privacy and the absence of surprises.

How we work, and how you can check

Anyone can write a paragraph about being honest. These are the specific things we have already committed to in public, each on a page you can read now:

• We say where your data lives. Not "your data is safe" — the actual layers, what we can see, what the infrastructure provider can see, and what nobody sees. Where your data lives.
• We publish estimates as estimates. Every figure on this site is a published specification, a formula written out on the page, or a number you typed in. Nothing here is a benchmark, and our own terms tell you not to buy hardware on the strength of one of our numbers alone.
• We published the rules before asking for money. The acceptable use policy is live before the service is, so it is something you can read before deciding rather than discover afterwards. It gives reasons, not just rules.
• We tell you what we collect. The privacy notice describes what the code actually does, field by field, including the things we deliberately do not store.
• We would rather be cited than protected. The guides and calculators are free, need no sign-up, and you may quote them anywhere with a link back. They are useful whether or not you ever become a customer.

The test of a company is not what it claims when things go well. It is what it published before it had customers, and whether that changed later. This page and those four have version dates and a public history for exactly that reason.

What we will not say

A few words you will not find on this site, deliberately:

• "Unlimited." Nothing is unlimited. Where there is a limit we would rather write the number.
• "Only you can see it." A rented environment runs on somebody's hardware. There is always an operator and always a provider, and pretending otherwise is the easiest lie in this industry.
• Benchmarks we did not run. When we have measured something on real hardware, we will say so and show the method. Until then, estimates stay labelled as estimates.
• A provider we have not named. Before anyone runs anything on our infrastructure, we will say whose machines they are.

Where this stands today

Plainly, because the stage matters more than the ambition: the service is not open. There is no product to buy, nothing is charged, and no environments are running. What exists today is this site, the reference material on it, four free calculators, and a waitlist.

Next is a narrow pilot: a small number of invited people, one supported setup, real hardware, a real bill, and a price that is provisional until it is on a page you can accept. If that pilot shows people finishing real work without needing a call from us, understanding their bill, and coming back, we continue. If it does not, we would rather say so than quietly keep selling.

We will not go quiet. If the project stops, everyone on the waitlist gets told, and their email address is deleted rather than kept for something else.

Who we are

A small independent team, not a venture-funded platform, and not a reseller with a logo on somebody else's dashboard. Small is a deliberate constraint rather than a stage we are trying to escape: it is why there are no sales calls, why the person who reads info@nodegrove.io is the person who builds the thing, and why we would rather support a few hundred people properly than sign up thousands and answer none of them.

It also sets what we can honestly promise. We are not going to claim enterprise support hours or a global footprint we do not have. We would rather be a company that does one thing carefully and tells you exactly what that thing is.

Questions, disagreements, or a number on this site you think is wrong: info@nodegrove.io. Corrections get made in public, in the same place as the error.