Why Cloud Computing Exists

A launch date you cannot un-guess

Picture a team preparing to launch a new product in 6 months. Before they write a line of code, someone has to answer a question nobody can answer accurately: how many servers will they need on launch day? Guess low, and the site buckles the moment real traffic shows up. Guess high, and the company just spent tens of thousands of dollars on machines that will spend most of their life idle, humming in a rack, drawing power, waiting for a traffic spike that may never come.

That guess used to be unavoidable. Buying a server was not a same-day decision. A purchase order, an approval chain, a vendor shipment, a rack installation, and a network configuration could easily add up to weeks before the hardware did anything useful. Once it arrived, someone on staff had to keep it running: apply security patches, replace failed drives, keep the room cool, and eventually retire and replace it every few years whether the business had grown into it or not.

This is the world cloud computing was built to escape: a large upfront purchase (capital expense, or CapEx) made months before you actually need the capacity, based on a guess you will often get wrong in one direction or the other. You will see this cost model contrasted with the cloud's pay-as-you-go alternative in detail later in this domain.

An idea older than the personal computer

It is tempting to think of the cloud as a brand-new invention from the 2000s. It is not. The core idea, letting many users share one large, expensive computer instead of each owning a small one, dates back to the 1960s. Mainframe computers of that era cost more than most companies could justify owning outright, so vendors and universities built time-sharing systems: dozens of people connected to the same physical mainframe through remote terminals, each believing they had the machine to themselves. By the mid-1960s, hundreds of companies were selling access to shared computing time this way.

Time-sharing did not survive the 1970s and 1980s. Minicomputers, and later personal computers, became cheap enough that businesses could simply own their computing instead of renting a slice of someone else's. The shared-mainframe market collapsed, but the underlying idea, that many users could share one large pool of computing power, never went away. It just needed better technology to come back.

The 3 pieces that had to fall into place

Modern cloud computing needed 3 separate advances to converge before it could work at the scale you see today.

The first was virtualization: software that lets one physical machine safely run several independent, isolated systems at once. IBM had proven this was possible on mainframes back in the 1970s, but it stayed a mainframe-only trick for decades. That changed in 1999, when VMware brought practical virtualization to the ordinary x86 servers that most businesses actually used, turning a mainframe specialty into something any data center could run. The next lesson walks through exactly how this works.

The second was reliable, affordable internet access. Time-sharing in the 1960s meant dialing into a specific machine over a dedicated line. A cloud has to be reachable from anywhere, on ordinary internet connections, which only became realistic as broadband spread through the 1990s and 2000s.

The third was a business willing to rent out spare capacity at scale. Amazon had already built massive data centers to handle its own retail traffic spikes, like the holiday shopping season, and realized that capacity sat underused most of the rest of the year. In 2006, it opened that infrastructure to the public as Amazon Web Services, launching with Amazon EC2 for compute and Amazon S3 for storage. Google followed in 2008 with Google App Engine, and Microsoft launched Microsoft Azure in 2010. Within 4 years, the 3 providers that still dominate the market today were all live.

What the cloud actually removes

Put the pieces together and you can see exactly what changed. The guessing game around capacity mostly disappears, because you can add or remove servers in minutes instead of weeks. The multi-week wait for hardware to ship and get installed disappears, because there is no hardware to ship to you. The large upfront purchase disappears, because you pay for what you use as you use it instead of buying equipment before you need it. And the ongoing maintenance work, patching, cooling, replacing failed drives, shifts to the provider, whose entire business is doing that at a scale no single company could match.

None of this makes the 1960s idea new again. It makes it, for the first time, practical: safe to run at scale, reachable from anywhere, and offered as a business you can simply sign up for instead of building yourself.

Where this leaves you

The problems in this lesson, expensive guessing, slow procurement, and constant maintenance, are the "why." The next lesson gives you the formal "what": the definition cloud computing actually has to meet to earn the name, so you can tell a real cloud service apart from a data center someone just calls one.