What is a VPS? VPS vs cloud server pricing and specs
October 22, 2026
Aron Wagner
CEO & Co-Founder
A VPS (virtual private server) is a virtual machine you rent on a provider's physical server. You get your own operating system, root access, and a share of the host's CPU, memory and disk.
A cloud server is a virtual machine too. What usually comes with the "cloud" name is billing by the hour, an API, storage you can move between machines, and resizing without a support ticket, but the word doesn't guarantee any of it. Mostly it's a sales label. Read the spec sheet, not the name.
One kind of VPS really is different: some are containers, not virtual machines. OpenVZ, for example, makes isolated containers it calls "VPSs", and every one of them runs the same OS kernel as the host, so you can't bring your own kernel. A hypervisor-based VPS gives each machine its own virtual hardware, the same as a cloud server.
People who sell hosting don't agree on the label either. In a 2021 LowEndTalk thread asking what cloud hosting really means, one member called cloud "a marketing term at the end of the day", and a host replied, "It's not a marketing term and it's not high availability." The host is right about the second half. "Cloud" doesn't mean your server stays up, and that's the part to remember.
VPS and cloud server features to check
NIST's definition of cloud computing comes down to resources you take and give back yourself, on demand, from a shared pool, with the provider metering what you use. Here's how that shows up on a spec sheet:
| Check | What to look for | What it means for you |
|---|---|---|
| CPU | "Dedicated" or "shared" vCPU, and any burst or credit limits | Whether your app gets the same speed at 3 p.m. as at 3 a.m. |
| Disk | Local SSD on the host, or network storage | Local is fast and stays with that host; network storage can move with the machine |
| Billing | Monthly plan or by the hour, and whether stopped machines are billed | What a test server or a traffic spike costs you |
| Resizing | Change CPU, memory or disk in place | Whether growing means a migration |
| API | An API, SDKs or a Terraform provider | Whether you can script it or only click through a panel |
| Host failure | What the provider says happens when the physical host dies | Whether one server can take your app offline |
Cheap VPS plans: is the CPU shared or dedicated?
The biggest difference between two servers with the same "2 vCPU" label is whether those CPUs are yours. DigitalOcean's plan guide puts it plainly: a dedicated CPU plan gets "guaranteed access to the full hyper-thread at all times", while on a shared CPU plan the hyper-thread "may be shared between multiple other Droplets", and when the neighbors are busy, "a Droplet could receive fractions of hyper-threads."
AWS has a third kind. Its T instances give you a baseline with the ability to burst, paid for with CPU credits. In Standard mode, once the credits run out, the CPU comes down to the baseline until it earns more. T3 instances start in Unlimited mode instead, where sustained bursts cost extra.
You can check a server you already have. Run top and watch the st value, which the man page defines as "time stolen from this vm by the hypervisor". If it sits well above zero while your app is slow, something else on the host is getting your CPU time.
Our VMs come with dedicated vCPU and RAM. Standard VMs run on Intel hosts and keep their disk on local SSD on the host; Premium VMs run on AMD EPYC hosts with NVMe network storage clustered across those hosts.
VPS bandwidth and outbound traffic
Read how a plan charges for outbound traffic: an allowance with a price for every terabyte over it, or a bill for every gigabyte you send. A busy month shows up in this line before any other.
On American Cloud, outbound transfer from a VM is unmetered and there are no egress fees. What that saves at different traffic levels is in egress fees explained, and the providers that don't charge for it are compared in cloud providers with no egress fees.
Cloud server uptime: what an SLA guarantees
The word "cloud" doesn't keep one server online. A single virtual machine runs on a single physical host, whatever it's called, and when that host fails, your machine goes down with it.
Some providers move machines off a host before it fails. Google Compute Engine live-migrates a running instance to another host in the same zone during maintenance, with no reboot and the same IP address; when the hardware fails outright, the instance terminates and restarts automatically. DigitalOcean live-migrates Droplets for maintenance, upgrades and hardware failures. Those are features of those two providers. The word "cloud" doesn't include them.
An SLA is a promise about credits, not a second server. Ours aims for 99.9% network uptime and pays service credits when we miss it. If your app has to survive a host failure, run it on two VMs behind a load balancer, which is $12 a month on our rate card.
VPS deployment with an API, SDK or Terraform
A server you can only create by clicking through a control panel is hard to rebuild the same way twice. Before you pick a provider, check what you can script.
Our VMs come with a REST API, official SDKs for TypeScript, Go and Python, a Terraform provider, and an MCP server so an AI assistant can manage them for you. You can change a VM's CPU, memory or root disk in place from its Manage menu, and attach block storage volumes that outlive any one VM.
American Cloud VPS pricing
We don't sell anything called a VPS. We sell cloud VMs, billed by the hour or on a monthly term, priced from one rate card: $10 per vCPU, $5 per GB of memory and $1 per 16 GB of disk a month. List prices on October 22, 2026, Standard (Intel) hosts:
| VM size | Monthly term | Full month billed by the hour |
|---|---|---|
| 1 vCPU, 1 GB, 25 GB disk | $16.56 | $16.89 ($0.0235 an hour) |
| 2 vCPU, 4 GB, 50 GB disk | $43.13 | $43.99 |
| 4 vCPU, 8 GB, 50 GB disk | $83.13 | $84.79 |
Hourly billing costs 2% more and suits servers you start and delete.
A shared-CPU box with the same label is a different product, so compare at the same CPU type: cloud server pricing at 8 clouds puts the same dedicated-CPU server side by side. And when something breaks, a person answers: our median support response time is under two minutes.
Start with one VM. Add a load balancer, a second VM or Kubernetes when your traffic calls for it, not before. To price your own server, use the pricing calculator.