People use "VPS" and "cloud hosting" like they're different categories of thing. Mostly, they aren't. Here's what's actually different, and where the real distinctions are.
VPS, plainly
VPS means Virtual Private Server. Take one physical machine, split it into several isolated virtual ones, and rent you one of those slices. You get root access, a fixed amount of CPU/RAM/disk, and it's yours until you cancel it. Nothing fancy, just a server that happens to be virtual instead of physical.
Cloud hosting, plainly
"Cloud hosting" almost always means the same underlying thing, a virtual machine, sold with a different pricing and provisioning model. Instead of one fixed box you keep forever, you can usually spin more of them up or down on demand, get billed more granularly (sometimes by the hour or even by usage), and the provider handles more of the underlying hardware failover for you.
So a big part of the answer to "what's the difference" is marketing and billing model, not the underlying technology. A VPS from a provider like UpCloud or DigitalOcean and a "cloud instance" from the same provider are frequently the exact same kind of virtual machine, sold under different plan names.
The distinction that's real: shared vs. dedicated resources
Here's the part that actually is a technical difference, not just naming. Most entry-level VPS and cloud instances give you shared vCPU: your virtual machine's CPU cores are physical cores that other customers' virtual machines are also scheduled onto. You get your slice when you need it, but the underlying core is time-shared with strangers. That's genuinely different hardware allocation, not just a label.
Hetzner is a clean real example because they sell both side by side. Their CX line (shared vCPU) starts at €5.49/month. Their CCX line (dedicated vCPU, meaning the physical cores are reserved to your instance only) starts at €42.99/month, roughly 8x the price at the entry tier. Same company, same underlying hardware family, genuinely different resource guarantee, and the price gap is exactly what you'd expect for actually reserving hardware instead of sharing it.
This is where "noisy neighbor" problems can happen even before you get to hosting multiple clients yourself: if you're on a shared-vCPU instance, someone else's spike on the same physical host can affect you, and you have no visibility into who that is or when it'll happen.
Where the other real difference shows up
Separate from shared-vs-dedicated hardware, there's the question this article started with.
It shows up at the platform layer, not the VM layer. A single VPS, however you bought it, is still one machine you're responsible for. If you're running 3 client apps or sites on it and one crashes the box, all 3 go down together. Nothing about calling it "cloud" instead of "VPS" changes that.
The real distinction that matters for hosting other people's sites is whether you're running one machine per client (VPS-style, however it's billed) or a shared platform layer, like Kubernetes, that spreads workloads across several machines and can move things around when one fails. That's an orchestration difference, not a VPS-vs-cloud naming difference.
I've written what that orchestration layer actually is if you want the next layer down.
Why this matters if you're hosting client apps or sites specifically
If you're choosing hosting for one site, the VPS-vs-cloud-hosting naming genuinely doesn't matter much. What does matter, even for one site, is whether you picked shared or dedicated vCPU for what that site actually needs.
It starts mattering once you're hosting several client apps or sites and comparing "one VPS per client" against a shared setup. That comparison isn't about VPS vs cloud at all, it's about whether each client's app or site is an island you maintain separately, or part of one platform where the maintenance work doesn't multiply by client count. I've gone deeper on the actual maintenance-cost argument for that specific comparison.
The short version
VPS and cloud hosting are usually the same product with different billing labels, that part's mostly naming. Shared vs. dedicated vCPU is the part that's a real technical and price difference, worth checking regardless of which label a provider uses. And if you're deciding how to host multiple client apps or sites, the real question isn't VPS vs cloud at all, it's whether each one gets its own isolated machine to maintain, or a shared platform that handles more of that for you.
If this is useful, tell me
I'm building a course and template repo around the shared-platform side of this, the actual setup I run client apps and sites on. Still figuring out how much terminology-level groundwork like this belongs in it versus assumed knowledge.
If you're trying to work out what actually fits your situation before committing hours to learning any of this, get on the list below. You'll get real build notes as I hit them, not a sales sequence.