The Cloud: Hardware Stores for Computers

Your Laptop Is a Complete Computer

Open your laptop and you have everything: a processor (CPU) that runs instructions, memory (RAM) that holds what you're working on right now, storage (SSD) that keeps your files permanently, a network connection, and an operating system tying it all together.

A server is exactly the same thing (CPU, RAM, storage, network, OS) except it sits in a warehouse somewhere, has no screen, and is designed to run 24/7 serving requests from the internet instead of running Spotify for one person.

Same hardware, different job. A server is just a computer without a keyboard, sitting in a climate-controlled room, connected to a very fast internet pipe.

---

The Old Way: Rent the Whole Machine

Before the cloud, if you wanted to run a website or app, you rented (or bought) an entire physical server. This is called bare metal, a real, physical machine assigned entirely to you.

The problem? You pay the same amount whether your app is using 1% of the machine or 100%. If your automated report runs for 3 minutes each morning and the server sits idle the other 23 hours and 57 minutes, you still pay for all 24 hours.

It's like renting an entire office building because you need one desk. The space exists, you're paying for it, and nobody else is using it.

---

The New Way: Cloud Providers Are Hardware Stores

Here's the framing that makes it click: cloud providers are hardware stores.

A hardware store has an address. You walk in and see a menu of things you can buy. You pick the specific piece you need (a screw, a board, a bracket), pay for just that, and leave. You don't buy the whole store.

Cloud providers work the same way:

• Here's my address: https://cloud.google.com (or AWS, or Azure)
• Here are the questions you can ask: compute, storage, databases, schedulers, AI models
• If you ask me, I will give you a response: spin up a server, store this file, run this query

You buy individual pieces of a computer instead of the whole thing:

• Need a database? Buy just Cloud SQL, a managed database. No server to maintain.
• Need to run code for 3 seconds? Use Cloud Run. It spins up, runs your code, and shuts down. You pay for those 3 seconds.
• Need to store files? Use Cloud Storage. Pay per gigabyte stored and per download.
• Need a timer to trigger something? Cloud Scheduler, a cron job that costs fractions of a penny.

Amazon, Microsoft, and Google each run massive warehouses full of servers and let you rent tiny slices of them.

---

Financial Impact: Pay for What You Use

Here's where it gets real. A bare metal server costs roughly $150-300/month whether you use it or not. Compare that to cloud pricing for a small app:

Resource	Bare Metal (bundled)	Cloud (separate)
Compute	Included in $200/mo	Cloud Run: ~$4/mo
Database	Included in $200/mo	Cloud SQL: ~$15/mo
Storage	Included in $200/mo	Cloud Storage: ~$2/mo
Scheduler	Included in $200/mo	Cloud Scheduler: ~$0.10/mo
Secrets	Included in $200/mo	Secret Manager: free tier
Total	$200/mo fixed	~$21/mo actual

The cloud version costs roughly 90% less because you're only paying for what you actually use. And when traffic spikes (say everyone on the team opens a dashboard at the same time), the cloud automatically adds more capacity and scales back down when they leave. Bare metal can't do that.

---

The Big Three: AWS, Azure, GCP

Three companies dominate cloud computing:

AWS (Amazon Web Services) is the oldest and largest. Amazon started selling its own excess server capacity in 2006 and now runs about a third of the internet's infrastructure.

Azure (Microsoft) is tightly integrated with Microsoft's ecosystem (Office 365, Active Directory, Teams). If a company already lives in Microsoft's world, Azure is the natural choice.

GCP (Google Cloud Platform) is Google's cloud. Strong in data, AI/ML (they built TensorFlow and Vertex AI), and Kubernetes (they invented it). Tends to be developer-friendly with cleaner APIs.

They all sell the same fundamental ingredients (compute, storage, databases, networking, AI), just with different names and different menus. Amazon calls their compute service EC2; Google calls it Compute Engine; Microsoft calls it Virtual Machines. Same idea, different branding.

---

A Real Cloud Bill: What a Typical Small Team Uses

Here's what a small engineering team might buy from a cloud provider's hardware store to run their internal tools:

• Serverless compute (like Cloud Run or AWS Lambda) runs all the web apps (dashboards, internal tools, etc.) and background services. Scales to zero when nobody's using them.
• Managed database (like Cloud SQL or AWS RDS) is a PostgreSQL database shared across projects. Stores application data, user records, campaign information.
• Object storage (like Cloud Storage or S3) handles files, images, backups. The equivalent of a shared hard drive in the cloud.
• Scheduler (like Cloud Scheduler or EventBridge) is the alarm clock. Fires automated reports each morning, weekly digests on Fridays, sync jobs every few hours.
• AI model access (like Vertex AI or AWS Bedrock) gives you access to AI models for AI-powered features. Pay per API call.
• Secrets manager stores API keys and passwords securely. Free tier covers small-team usage.

Each of these is a separate line item, a separate service, running independently. If you stopped using one app tomorrow, its compute cost drops to zero. The database, scheduler, and everything else keeps running unchanged. That's the power of buying pieces instead of a whole computer.

---

Further Reading