Here is how most people store their digital lives in 2026. Photos go to Google. Email goes to Gmail. Documents go to whatever cloud came bundled with the phone. Passwords sit in a browser owned by an ad company. Every byte lives on someone else’s server, governed by terms nobody read, in a jurisdiction nobody checked. We hand it all over and we call it normal, because the alternative looks like a second job.

I want to talk about that alternative, and about why it is currently locked behind a wall most people will never climb.

The Problem: We’ve Lost Control

The business model behind most of these services depends on knowing you. Your email gets scanned. Your photos get analysed. Your location history becomes a product. None of this is a secret anymore, and most people have made a rough peace with it. The data is convenient where it sits, and moving it feels like more trouble than it is worth.

The technical crowd has an answer to this, and it is a good one. Self-hosting. Run your own services on your own hardware, and the surveillance problem mostly evaporates because there is nobody on the other end to do the surveilling.

I have run my own infrastructure for years. My own email, my own cloud storage, my own password vault, the lot. I genuinely love it, and I have written a fair bit about why on this blog. But I am not going to pretend my parents could do the same. Self-hosting today is a privilege. It costs technical knowledge and a steady supply of evening hours, and most people have neither to spare. The barrier is real, and it is high.

What If the Barrier Wasn’t There?

So picture the other side of that wall.

You pay a monthly fee, somewhere in the €15-30 range. In return you get your own private cloud for photos and documents and notes, your own email server, your own password manager, your own calendar and contacts. The usual stuff people rent from Big Tech, except the arrangement underneath is different in ways that actually matter:

  • Your data stays yours. No scanning, no profiling, no quiet resale to a data broker.
  • It runs on commodity hardware. That old PC in your closet works fine.
  • You can leave whenever you want. Take your data and walk. No lock-in.
  • The operations are transparent. Open source, auditable, no black boxes.

None of these pieces are new. Nextcloud, Mailcow, and a long list of mature open-source projects already cover the building blocks. The gap has always been the same one: somebody still has to assemble and babysit them. The missing piece is making that somebody not you.

The Cognitive Tax Nobody Prices In

Here is where the current reality gets heavier, because the barrier is not really about the initial setup.

Setup is a weekend, and a determined person can follow a tutorial. The cost shows up afterwards, every week, forever. Certificates expire. A disk fills up at 2am. An upstream image ships a breaking change and your contacts app stops syncing right before you need a phone number. Self-hosting is not a thing you build, it is a thing you keep alive. That maintenance load is exactly why I would never hand a Raspberry Pi to a non-technical friend and call it a gift. I would be handing them a pager.

This is the same friction problem that drives most of my technical choices. I want systems that fail gracefully and heal themselves, because the alternative is me being the single point of failure for someone else’s email at 2am. That does not scale, and frankly it is not kind to either of us.

So the question is not whether self-hosting is good. It is good. The question is who carries the operational weight.

How It Works: The Service Cluster Model

I have already started building an answer. Here is the shape of it.

flowchart TD
    subgraph service["Service Cluster<br/>(Managed by me, runs management plane)"]
        GitOps["GitOps<br/>(ArgoCD)"]
        Monitoring["Monitoring<br/>(Grafana)"]
        Updates["Updates<br/>(Auto)"]
    end
    service --> A
    service --> B
    service --> C
    subgraph A["Customer Cluster A (Home)"]
        A1["Nextcloud"]
        A2["Vaultwarden"]
        A3["Photos"]
    end
    subgraph B["Customer Cluster B (Office)"]
        B1["Nextcloud"]
        B2["GitLab"]
        B3["CI/CD"]
    end
    subgraph C["Customer Cluster C (Colo)"]
        C1["Email"]
        C2["Matrix"]
        C3["Nextcloud"]
    end

A central service cluster manages a fleet of customer clusters. Each customer cluster runs on hardware the customer owns, whether that is a mini PC at home, a server in an office, or rented hardware in a datacenter.

The service cluster handles the parts that hurt:

  • Deployment that is GitOps-driven, declarative, and reproducible
  • Monitoring so something gets caught before the customer notices
  • Updates that are automatic, tested, and rolled back when they misbehave
  • Backups that are encrypted, off-site, and actually restored once in a while to prove they work

The customer cluster handles two things only: running the services, and storing the data on hardware the customer can physically touch.

That split is the whole point. The service cluster never sees customer data. It manages the infrastructure, and the data stays on the customer’s side of the line. The person doing the babysitting is me and my tooling, not the person who just wanted their photos to stay private.

Running on Old Hardware

One principle I am not willing to give up: this runs on hardware people already own.

That old laptop, those retired office PCs, the Raspberry Pi gathering dust on a shelf, all of it qualifies. K3s runs comfortably on machines Big Tech would have written off as obsolete years ago. That matters for a few reasons at once. It keeps usable hardware out of the bin. It drops the cost of entry close to zero. And it makes the ownership real, because true sovereignty means the box is in your house, not a virtual lease in someone’s datacenter.

I have written before about sovereign infrastructure and why physical control is the part that counts. This is that same idea pushed one step further, out of the homelab and into the hands of people who would never build a homelab.

The Future: LLM-Managed Infrastructure

Now the part I find genuinely exciting, with the caveat that it is the least proven.

Today I am the operator. I read the alerts, I debug the failures, I make the calls. That works at the scale of friends and family. It does not work at the scale of everyone, because my attention is the bottleneck and there is only one of me.

Large language models are getting good at the kind of judgement that operations work demands: reading logs and metrics, diagnosing common faults, proposing a fix, weighing context. Picture a system that notices a customer’s disk creeping towards full, looks at the usage pattern, recommends a cleanup or an expansion, applies the chosen path, and remembers how it went for next time.

The building blocks for that exist right now. What does not yet exist, and what I am being careful about, is the guardrails. An LLM with write access to someone’s email infrastructure is a serious thing, and getting the autonomy-versus-safety balance wrong helps nobody. So this stays speculative until I can show it works on my own cluster first. I am not going to ask anyone to trust a thing I have not broken in myself.

If it does work, my role moves from operator to architect. I design the policies and the system designs the rest, and adding a customer stops adding a customer’s worth of pager duty.

The Business Model

Let me be pragmatic, because a vision that cannot pay for itself is just a hobby with extra steps.

Individuals: €15-20/month for cloud storage, a password manager, and photos.

Families: €25-35/month for shared services plus individual accounts.

Small businesses: €50-100/month for email, collaboration, and CI/CD at a quality you would actually run a company on.

Compare that to the going rate:

  • Google Workspace at €5-18/user/month, paid partly in your data
  • Microsoft 365 at €5-20/user/month, same arrangement
  • Self-hosting it yourself, “free” in euros and expensive in hundreds of hours

I want to be honest about the trade-off rather than sell past it. This is not the cheapest option on a spreadsheet. It asks you to pay money for something the surveillance economy gives away, and the thing you get back is sovereignty without the operational burden. For some people that is obviously worth it. For others it never will be, and that is a fair call to make.

What’s Next

I am building this in stages, on purpose, slowly enough to break things in private:

  1. Now: my own service cluster managing a handful of customer clusters for friends, family, and a few early adopters
  2. Soon: documenting the architecture and open-sourcing the management tooling
  3. Later: the LLM integration, once I trust it on my own infrastructure
  4. Eventually: something anyone can sign up for and actually own

If this resonates, whether you want to be an early adopter, help with the tooling, or just watch it develop, I will keep writing about it here, including the parts that go wrong.

Renting access to your own data from a corporation is one possible future for personal computing. It is not the only one, and it is not the one I want to live in. There is a version where the box is yours, the data is yours, and somebody competent keeps the lights on without ever looking inside. That is the one I am building.