A home for your company data — and a message when something stops

On site: the working data stays physically in the company, you set access yourself, and after the one-off purchase you pay no provider a monthly rent for that machine; the costs that remain are power, the copy kept off the device, and replacing the hardware later. The start is slower, because hardware has to be bought and connected. A rented machine: you start faster and buy nothing, but you pay continuously, and the data sits with someone whose rules you do not set. Where the backup goes is settled separately — a copy sitting next to the original survives neither fire nor theft, so one of those places is often outside the company, and I say so before we start. By default I propose the computer at your premises, because that is why you are here: so the data stays in the company. I propose the rented machine when there is nowhere to put a computer, or you want to start right away — and I say plainly what you give up.

Usually not to begin with. One small machine covers the first tasks. AI models running locally may need a stronger graphics card, and I say so before the purchase, not after. Hardware follows the task, not a stock of spare capacity. It can also be done differently: the small machine stays with you and an outside provider's model does the heaviest part. Cheaper at the start, a monthly fee, and some of the content leaves the company. I propose the on-site variant by default, but the choice is yours.

It will eventually, and that is planned for. This is why the backup is checked by an actual restore: you then know the result for the scenario that was tested and the measured time of that attempt, not a guarantee for the duration of every future failure. Without that attempt a backup is only an assumption — and it is the most common thing I find when somebody says they have backups.

No. A spot with power and network is enough, where a device can sit continuously and nobody switches it off or moves it. If there is no such spot, the rented machine is the sounder option — and that is what I will advise.

I start by looking over what already runs — that is part of the conversation before the quote. Whatever can stay, stays; I replace only what gets in the way. If the new setup needs something fixed that worked before, that is covered by the price of the implementation. Taking over the day-to-day running of your existing server is separate work and a separate quote. I look after the place where the data lives and the services on it — not staff computers, printers or accounts, and I do not rebuild your network. I do not buy hardware in your name: you buy it yourself, and I say what and why. And I do not rule on whether you comply with the regulations.

Up to handover I answer for putting it into service, for testing the backup by restoring from it, and for setting up access the way we agreed. After that it is run by the person your company names, or by me — within the care package you choose. The message still arrives either way, but it is not a promise of how soon somebody will fix the fault. What is in the data, and who should see it, is your company's call. Whether your case needs further formal steps is a question for your own adviser — I do the technical part and say where it ends.

In Bavaria the Digitalbonus is running; I describe its conditions on the home page. One thing worth knowing early: only a project that has not yet started can be funded, so the conversation about funding belongs before the start, not halfway through.

Yes — after a conversation about what is meant to run on it. Hardware or the rented machine is billed separately from my work, so you see two figures rather than one. I do not give a range beforehand: a small machine for documents and a machine for AI models are very different money. No open-ended hourly billing.