Skip to content
DevOps devops cloud 5 min read

Cloud Providers

A cloud provider is a company that rents you computers, storage, and networking over the internet, so you never have to buy or rack your own hardware. You pay for what you use, by the hour or by the second, and you can create or destroy servers in minutes. Picking the right provider matters because it shapes your costs, your learning curve, and how much “plumbing” you have to manage yourself. This page walks through the big three (AWS, Azure, GCP) and the simpler, developer-friendly providers (DigitalOcean, Linode, Hetzner, Vultr) so you can choose with eyes open.

The two families of cloud provider

Cloud providers roughly split into two camps.

The hyperscalers — AWS (Amazon Web Services), Azure (Microsoft’s cloud), and GCP (Google Cloud Platform) — offer hundreds of services covering compute, databases, machine learning, queues, and more. They are powerful and run most of the internet’s largest companies. The trade-off is complexity: the dashboards are huge, pricing is hard to predict, and you can accidentally run up a big bill.

The developer-friendly VPS providers — DigitalOcean, Linode (now part of Akamai), Hetzner, and Vultr — focus on one core thing: giving you a simple Linux server (a VPS, or Virtual Private Server — a slice of a physical machine that acts like your own dedicated computer). Pricing is flat and predictable, the dashboards are clean, and you can have a running Ubuntu box in under a minute.

When you are learning DevOps and Linux, start with a developer-friendly VPS provider. The hyperscalers add a mountain of abstractions that get in the way of actually learning how a server works.

Comparing the providers

ProviderTypeCheapest server (approx, 2026)Pricing modelBest forWatch out for
AWSHyperscaler~$4/mo (t4g.nano)Per-second, complexEnterprises, huge scale, broad service catalogSteep learning curve, surprise bills
AzureHyperscaler~$4/mo (B1ls)Per-second, complexWindows shops, Microsoft/.NET ecosystemsPortal complexity, licensing
GCPHyperscaler~$5/mo (e2-micro)Per-second, complexData/ML workloads, Kubernetes (GKE)Fewer regions, smaller community
DigitalOceanVPS$4/mo (Droplet)Flat monthly/hourlyLearners, startups, small appsFewer managed services
Linode (Akamai)VPS$5/mo (Nanode)Flat monthly/hourlyLearners, predictable budgetsSmaller than DO ecosystem
HetznerVPS~€4/mo (CX22)Flat monthlyCheapest power-per-euro, EU hostingEU-centric data centres, stricter signup
VultrVPS$5/mo (Cloud Compute)Flat monthly/hourlyMany global locations, bare metalSmaller community

A few notes to read the table honestly:

  • The “cheapest server” figures move over time and exclude bandwidth and storage add-ons. Always check the provider’s current pricing page.
  • The hyperscalers bill per second and break every feature into a separate priced line item (compute, disk, IP address, data transfer). That granularity is great at scale but makes a $4 box quietly become a $30 bill.
  • The VPS providers bundle a fixed amount of CPU, RAM, disk, and bandwidth into one flat monthly price. You know your bill before you start.

When each one fits

Choose AWS when your team needs the widest possible catalogue of managed services, you are hiring engineers who already know it, or you are building something that must scale to millions of users. It is the default for large enterprises.

Choose Azure when your organisation already lives in the Microsoft world — Active Directory, Windows Server, .NET, Office 365. Azure integrates tightly with those, and your existing Microsoft licensing may carry over.

Choose GCP when your workload is data-heavy or machine-learning-heavy (BigQuery, Vertex AI) or when you want best-in-class managed Kubernetes (GKE). Google’s networking and data tooling are strong.

Choose a VPS provider (DigitalOcean, Linode, Hetzner, Vultr) when you want a plain Linux server you fully control, a bill you can predict, and a dashboard you can understand in five minutes. This covers the vast majority of personal projects, side businesses, and learning.

My recommendation for learners

Start with DigitalOcean or Hetzner. DigitalOcean has the friendliest dashboard and the best free tutorials on the internet (their Community articles are excellent). Hetzner is the cheapest for the raw power you get, with strong hardware, if you are comfortable with a slightly stricter signup and EU-based data centres. Either lets you spin up Ubuntu 24.04 LTS for a few dollars a month and learn apt, systemd, and ufw on a real internet-facing server without fear of a runaway bill.

A quick taste: creating a server with the DigitalOcean CLI

Most VPS providers offer a command-line tool so you can create servers without the web dashboard. DigitalOcean’s is called doctl. Here is the full flow on Ubuntu.

First, install the tool using snap (a packaging system built into Ubuntu):

sudo snap install doctl

Output:

doctl 1.104.0 from Digital Ocean installed

Authenticate with an API token (create one in the DigitalOcean control panel under API):

doctl auth init

Output:

Please authenticate doctl for use with your DigitalOcean account. You can generate a token in the control panel at https://cloud.digitalocean.com/account/api/tokens
Enter your access token:  ****************************************
Validating token... OK

Now create a small Ubuntu Droplet (DigitalOcean’s name for a VPS):

doctl compute droplet create learn-box \
  --image ubuntu-24-04-x64 \
  --size s-1vcpu-1gb \
  --region nyc1 \
  --ssh-keys 12345678 \
  --wait

Output:

ID           Name        Public IPv4      Memory    VCPUs    Region    Status
411223344    learn-box   203.0.113.45     1024      1        nyc1      active

Within about a minute you have a live Ubuntu server. From here you SSH in (ssh [email protected]) and start configuring it — see the next pages for that.

Always create your server with an SSH key (--ssh-keys) rather than a password. Password logins on a public IP get brute-forced within hours of the server coming online.

Best Practices

  • Start simple. Use a flat-rate VPS provider while learning; move to a hyperscaler only when a specific managed service actually justifies the complexity.
  • Set a billing alert on day one, especially on AWS/Azure/GCP, so a misconfigured resource cannot silently drain your wallet.
  • Pick Ubuntu LTS (Long Term Support) images — 22.04 or 24.04 — for the longest security-update window and the best community documentation.
  • Authenticate with SSH keys, never passwords, and disable root password login as soon as the server is up.
  • Choose a data-centre region close to your users to cut latency, and one that meets any data-residency rules you must follow.
  • Destroy servers you are done with. Cloud billing keeps running until you delete the resource, not just power it off.
  • Read the actual current pricing page before committing — the numbers above are 2026 estimates and change often.
Last updated June 15, 2026
Was this helpful?