When I first heard about "cloud computing" I thought it was just marketing nonsense. Servers in someone else's building, accessible over the internet — how is that different from hosting? Well, it turns out there's a bit more to it, and after years of working with cloud services I can safely say: it's still partly marketing nonsense, but it's also genuinely useful.

What the cloud actually is

Forget the buzzwords for a minute. The cloud is basically renting computers and services from someone else (AWS, Azure, Google, etc.) instead of buying and maintaining your own hardware. That's it. The "revolutionary" part is that you can rent exactly what you need, when you need it, and let someone else worry about the physical servers.

The three main models you'll hear about:

  • IaaS (Infrastructure as a Service): You rent virtual machines. You do everything else. Think EC2.
  • PaaS (Platform as a Service): You deploy code, the platform handles the rest. Think Heroku, Vercel.
  • SaaS (Software as a Service): You use someone else's application. Think Gmail, Slack.

As a developer, you'll mostly deal with IaaS and PaaS. SaaS is what your non-technical colleagues use.

The big three providers

AWS is the one everyone uses. It has the most services, the most documentation, and the most tutorials. It's also the most complicated and the documentation quality varies wildly. Some AWS docs are excellent, others look like they were written in 2014 and never updated.

Azure is what you end up with if your company uses Microsoft stuff. The integration with Active Directory, Office 365, and .NET is good. For JavaScript and Python development, it's fine but not special.

Google Cloud Platform is technically solid but has a smaller market share. The data and ML tools (BigQuery, Vertex AI) are genuinely good. The problem is that Google has a habit of killing products, which makes you nervous about building anything important on their platform.

They all offer free tiers, by the way. If you're learning, use the free tiers to experiment. You can do a lot without spending anything.

Things that are actually useful

  • Object storage (S3, Blob Storage, Cloud Storage): Put files in, get files out. Cheap, reliable, infinitely scalable. I use this for everything.
  • Managed databases (RDS, Cloud SQL): Running your own database is a pain. Let someone else handle backups, patches, and replication. Worth every penny.
  • CDN (CloudFront, Cloudflare): Serve static content from locations close to your users. Essential for anything with a global audience.
  • Serverless functions (Lambda, Cloud Functions): Good for specific use cases, not for everything. I covered this in detail in another post.

Things that are overhyped

  • "Going cloud-native": Just means using cloud services properly. Not as revolutionary as the conference talks suggest.
  • Multi-cloud: Running your application on multiple cloud providers simultaneously. Sounds smart, but it doubles your complexity for very little benefit. Very few companies actually need this.
  • "Serverless everything": Already covered this. Some things need servers and that's fine.

The real benefits

For me, the real value of the cloud is speed. I can spin up a database in 5 minutes instead of ordering hardware and waiting for it to arrive. I can deploy an application globally without negotiating contracts with data centres. I can experiment with a new service for a weekend and throw it away if it doesn't work.

The cost? You pay a premium for that convenience. Cloud is almost always more expensive than running your own hardware if you're running things 24/7 at scale. But for most of us, the flexibility is worth it.

Where to start

If you're new to cloud stuff, pick one provider and learn the basics. I'd suggest AWS because there are more tutorials and community resources. Create a free account, deploy a simple application, set up a database. Don't try to learn everything at once — the cloud providers each have 200+ services and nobody knows all of them.

Start with EC2 (virtual machines), S3 (storage), and RDS (databases). That's 90% of what you'll use anyway.