Leaving a Major Cloud Platform, One Piece at a Time

I left my cloud platform slowly, in pieces, one service at a time, as each piece gave me a reason.

The reason that started it was their AI. I was paying for the AI future from the company that holds my files, my photos, and my email, and the flagship AI product kept failing at ordinary work in the ways I describe in Using AI Without Becoming Its Intern. At some point the thought stops being this tool is disappointing and becomes why is my entire digital life rented rather than owned? So I started with the piece that mattered most and was easiest to verify: my files.

What's gone, what stayed, and why

The cloud drive and its storage plan: gone. The storage subscription lapsed in October and I let it. Everything that lived there now lives in my house.

The AI subscription: cancelled. That one was an easy decision by the end.

The premium video subscription: staying. It's a television. Nothing of mine lives there, nothing of mine is held hostage there, and when it stops being worth the money I'll cancel it the way you'd cancel any TV package. A plain service like that is fine.

Email and calendar: still there, for now, on purpose. Your email address is your identity on half the internet: password resets, bank accounts, fifteen years of threads. You don't migrate your identity on a weekend because you're annoyed. Those are later phases, and they'll get the same treatment as this one: replacement built, replacement proven, then the cord gets cut.

The replacement: Nextcloud on a VM I own

The cloud drive's replacement is a self-hosted Nextcloud instance running on a VM in my homelab, on the same 3-host vSAN ESA cluster the rest of this site is about. It does the jobs the cloud drive did: files sync between my machines, I can share a folder when I need to, and my home automation uses it as its drop point. My receipt pipeline files its output there, and the bots that help me build things exchange files with me through a share of their own.

When I need my files away from home, I reach the lab over my tailnet. The file server has no public endpoint, which no cloud drive can say.

The tradeoff is that I'm the support desk now. When it breaks, it breaks on me. Updates are mine, maintenance is mine, 3 AM is mine. What I get for that is the thing no subscription sells: nobody scans these files, nobody gets to decide later what they're used for, nobody can retire the feature I depend on in a blog post, and nobody can hold them for rent.

Backups for the files

Moving your files home and stopping there only concentrates them in one place. So the Nextcloud data is backed up with restic to Backblaze B2, encrypted on my side before anything leaves the house. B2 stores ciphertext. The master copy of the key that can read it lives in my password manager; the VM holds a copy in a credential file so the backups can run unattended. The script itself carries nothing.

That gives the files two independent homes with two independent failure modes: the lab copy survives a cloud provider's bad day, and the encrypted off-site copy survives the house's bad day. It's the same doctrine as the lab's VM backups, applied to the files themselves.

Fencing the bots

Most write-ups about leaving a platform skip this part, because most people leaving one don't run automation against their files. I do. Bots and scripts read and write in my Nextcloud every day, and the moment your files come home, the question what can reach them, and with whose credentials stops having their answer and starts needing yours.

So nothing automated gets my credentials. Every bot gets its own Nextcloud account and app password: its own identity, revocable on its own, and it only sees the folders shared with it. On the network side, Tailscale ACLs decide which machines on my tailnet can talk to the file server at all. A confused or compromised bot can damage its own small corner. It cannot wander my files, because it was never given the keys to them.

It's the house rule for everything I build: any security risk, no matter how small, gets clamped down as part of the build. Cloud storage made that question invisible by answering it for you. Self-hosting makes you answer it yourself, and I'd rather answer it myself.

The method is the piece-by-piece part

People hear "leaving a platform" and picture one weekend: export everything, delete everything, announce it on the internet. That's how you lose data and lock yourself out of accounts. The method that works is boring:

Pick one service. Build its replacement. Live on the replacement until it has proven itself under your real life. Then cancel, let the subscription lapse, and move to the next piece.

The cloud drive went first because files are the easiest thing in the world to verify: they're either there, intact, and backed up, or they're not. It proved itself, so the storage plan lapsed and nothing bad happened, which is the entire review of most infrastructure you own: nothing bad happened, for months.

My files live on hardware I can point at now. The rest of the platform can wait its turn.

Sources

Updated Nov 17, 2026: edited the prose for clarity.


Daniel Barnes

Technical Support Engineer, VMware Cloud Foundation Division at Broadcom. Customer-facing work on VMware vSAN, ESXi, and VMware Cloud Foundation: storage, HCI, and virtualization troubleshooting. Views are my own and not necessarily those of Broadcom. I use AI tools for drafting and editing; the technical claims are mine, checked against public documentation and my own lab.

Started in 2018.

Homelab is a 3-host vSAN ESA cluster on ESXi 9.1 running self-hosted services (Nextcloud, Pi-hole DNS). That's where the write-ups come from.