# Help and answers. > Answers about the Alpha Cube, from choosing a model and setting it up to security and how each industry uses it. ## General questions ### What is an Alpha Cube? An Alpha Cube is a complete private AI system for your business. It puts server-grade power in one compact Cube that sits on your premises. ### Who is it for? It’s for businesses that want to put powerful AI to work but have sensitive data or intellectual property that can’t be sent to cloud AI platforms. ### How is it different from using AI in the cloud? Your data, the models you train and your intellectual property stay under your own roof, on a system you own. Usage has no limits and nothing bills by the token, so costs don’t climb as your team uses it more. And you set the rules for how it’s used. ### What do I need to run one? Power and a wired ethernet connection to your network. It has no wireless connections. Setup is simple, and keeping it running takes no special staff. ### How do I order? Click Order and share a few details about your team. We’ll save your place on the waitlist and email you when your spot opens up. [Join the waitlist](https://vaultaisystems.com/order) ## Choosing an Alpha Cube questions ### Which Alpha Cube is right for our team? Alpha Cube Ion fits teams of up to 200, Alpha Cube up to 600, and Alpha Cube Pro up to 1,200. [Compare models](https://vaultaisystems.com/compare) ### How many people can use one Cube? An Alpha Cube serves up to 600 people at once, and more Cubes can be linked as teams grow. [Compare models](https://vaultaisystems.com/compare) ### Can we choose which models it runs? Yes. Load the open models you approve, including ones you’ve fine-tuned on your own data. ### What kinds of data can GEM work with? Documents, spreadsheets, datasets and reports, alongside your own analysis files. [Learn about GEM](https://vaultaisystems.com/system#gem) ### Can we train or fine-tune models on our own data? Yes. Train or fine-tune models on your own data right on your Alpha Cube, then switch to them in a click. Models you train stay yours. ## Setup and support questions ### Does it work offline? Yes. It can run fully offline for work that can’t touch the internet. ### Is there a limit on how much we can run? No meter. Run a job or rerun a search as often as you need, and costs don’t climb with use. ### Is it a purchase or a subscription? It’s a purchase, and costs don’t climb with use, so you can plan around it. ### Can we add more Cubes later? Yes. VaultLink links two or more Alpha Cubes so they run as one system and share the load. If one goes down, the others pick up the slack. [Learn about VaultLink](https://vaultaisystems.com/system#vaultlink) ### How do we get help? A real person on the Vault team. Email us at [hello@vaultaisystems.com](mailto:hello@vaultaisystems.com). ## Security questions ### Does any of our data go to the cloud? No. Models, VaultOS and GEM run on the Cube inside your firewall, so business data is never sent to the cloud. ### Does an Alpha Cube connect to the internet? Only through your network, by cable. There’s no Wi-Fi or Bluetooth, and it can run fully offline. ### What gets inspected? Every new file, dataset or model, before it reaches the Cube. ### What happens to a file that fails? It’s quarantined for admin review and never reaches the Cube. ### Who decides what each person can see? You do. Set access by person, team or department, so confidential work stays put. ### Can we see who used it and when? Yes. Every question is logged, so your team can review who asked what and when. ### Is our data used to train anyone else’s model? No. Your data only trains models you choose to train, and those stay on your hardware. ### Does an Alpha Cube send any data back to Vault? No. Nothing on the Cube is sent back to Vault. ### Does every Alpha Cube include this? Yes. Every model comes with file inspection, role-based access and wired-only networking, on your own network. [Compare models](https://vaultaisystems.com/compare) ## Healthcare questions ### Can we run AI on patient records without sending them out? Yes. Records stay on your network, and GEM processes them on the Cube, never on an outside provider’s servers. ### What kinds of data can GEM work with? Clinical notes, trial datasets, pathology reports and published studies, alongside your own analysis files. ### Can we train a model on our own study data? Yes. Fine-tune on your records right on the Cube, and the model stays on your hardware. ### Can every study team share one Cube? Yes. Access is set per study, so each team sees only its own records, and an Alpha Cube serves up to 600 people. [Compare models](https://vaultaisystems.com/compare) ### What do we tell our IRB? That identifiable data stays on hardware your institution controls and never goes to an outside AI provider. ## Education questions ### Can faculty and grad students share one Cube? Yes. Each lab gets its own access, and an Alpha Cube serves up to 600 people. [Compare models](https://vaultaisystems.com/compare) ### How does it fit a grant budget? It’s a fixed cost you can plan for. Costs don’t climb with how much your team runs, so you won’t run short mid-grant. ### Can we fine-tune models on our own research? Yes. Train on your lab’s data and the model stays yours to publish with. ### Can courses use it, not just research? Yes. Classes can run on the same Cube, with access keeping course work apart from research. ### Does it work offline? Yes. It can run fully offline for work that can’t touch the internet. ## Finance questions ### Does our strategy ever reach an AI provider? No. Prompts and data are processed on the Cube in your office, so nothing about your positions goes out. ### Can compliance see how it’s being used? Yes. Every question is logged, so compliance can review who asked what and when. ### Will thousands of backtests get expensive? No. There’s no per-token charge, so a heavy quarter of backtests doesn’t bring a bigger bill. ### Can we train a model on our own research? Yes. Fine-tune on your notes and trade history, and the model stays on your hardware. ### Can research, risk and client reporting share one Cube? Yes. Each team sees only its own files, and an Alpha Cube serves up to 600 people. [Compare models](https://vaultaisystems.com/compare) ## Legal questions ### Do our files ever leave the office? Only if you move them. The Cube runs on your network, and nothing is sent to Vault or to an outside AI provider. ### How is this different from cloud legal AI tools? Cloud tools process your files on the provider’s servers and charge by seat or usage. A Cube keeps the work in your office, with no per-token or per-page fees. ### Is there a limit on how much we can run? No meter. Review a full production or rerun a search as often as you need, and costs don’t climb with use. ### Can one client’s files show up in another client’s work? No. Access is set per matter, so GEM only surfaces files a person is cleared to see, and nothing is shared outside your Cube. ### Which Alpha Cube is right for our team? Alpha Cube Ion fits teams of up to 200, Alpha Cube up to 600, and Alpha Cube Pro up to 1,200. [Compare models](https://vaultaisystems.com/compare) ## Government questions ### Can we choose which models it runs? Yes. Load the open models your lab approves, including ones you’ve fine-tuned on your own data. ### Does it need an internet connection? No. It runs on your lab’s internal network. If you choose to connect it, it’s over wired ethernet only, with no wireless connection at all. ### How are files from outside the lab checked? Before a new file is added to the Cube, VaultShield inspects it. ### Is it a purchase or a subscription? It’s a purchase, and costs don’t climb with use, so you can plan around it. ### How many people can use one Cube? An Alpha Cube serves up to 600 people at once, and more Cubes can be linked as teams grow. [Compare models](https://vaultaisystems.com/compare) Have another question? [Get in touch](https://vaultaisystems.com/contact) - [How security works Security](https://vaultaisystems.com/security) - [What’s inside The Vault System](https://vaultaisystems.com/system) - [Compare models Alpha Cube lineup](https://vaultaisystems.com/compare) ## Need help with your Alpha Cube? Email our team at [hello@vaultaisystems.com](mailto:hello@vaultaisystems.com) Setting up a new Alpha Cube? Open the [Quickstart guide](https://vaultaisystems.com/start) Page: https://vaultaisystems.com/questions