It has appeared earlier. In fact, this was the original model of getting access to computing resources back in the 1950s till well into the 1980s, when it was called “time sharing”, then in the early 1990s it re-appeared under the name “Client/Server”, then in the late 1990s again under the name “Thin Client”, then “Application Service Provider”.

However, in the exact form we see it today it requires high quality, high reliability, high throughput, low latency, low price, ubiquitous Internet access, which didn’t exist until a few years ago, and in fact, still doesn’t exist for the vast majority of people (e.g. almost all of Africa, much of Asia, parts of Eastern Europe and South America).

People have been renting time on remote computers for decades. In fact, “timesharing” was the original model for selling computing services back before computers were small enough and affordable enough that individual businesses could afford to own their own machines. The large information services of the 80’s (Compuserve, AOL, etc.) were another way to rent computing power/space. Next, as the Internet developed, people needed ways to maintain a 24/7 presence on the network and hosting companies popped up.

Cloud computing is just another version of the same idea. It took some time for data centers to become so developed, standardized, and scalable that cloud services could sell general purpose computing on virtual machines as a commodity and manage it all in a way that was both affordable and profitable, but it’s really just the latest generation of the same idea.

There are two answers. The first is that it didn’t really take off until high-speed internet access became ubiquitous. Cloud computing doesn’t work well unless you can be reasonably sure that you will always have high-speed access to your cloud resources.

The second answer is that it’s not really a new idea. Before PCs became affordable it was the norm to have many people connecting to one computer using dumb terminals. The machine you would be sitting in front of wouldn’t have any storage or processing abilities beyond what was necessary to send your input and display output.

I would say it depended on virtualization technology on commodity hardware. Time sharing and mainframe/client access has always existed, but required expensive special hardware to securely portion resources. Client/server access has always existed on commodity hardware since the internet. However, it required a dedicated server and you couldn’t just replicate that server with the push of a button. In order to maintain security, you had to maintain the security on that server yourself. Shared servers were susceptible to attack, unless they were locked down, and that limited options if you needed something custom.

With ubiquitous, cheap virtualization, you can create an entire server with full access, and that can be shared on a larger commodity platform. It can be copied, moved, replicated, and deleted on a whim. It required INTEL and AMD chips to support the virtual machines and time slicing that goes with servers, along with easy software to maintain multiple operating systems running at once.

What was the critical change that triggered massive adoption and massive marketing of cloud computing offers?

As other posters have mentioned, the one critical change was ubiquitous access to high speed internet.

The other critical change was the advancement of interactivity in websites. The modern-day richness in web user interfaces is what effectively allowed more and more traditionally desktop applications to be served via the cloud.

Apparently most people have a shaky grasp on ‘Cloud Computing’…

The short version:

A system whereby computing power has been abstracted away from physical infrastructure so that it can easily be bought, sold, and leveraged as a commodity.

The Long Version:

‘Cloud Computing’ is simply the next step in abstracting away the maintenance and infrastructure requirements involved in developing and supporting software platforms.

Cloud can be broken down by the types of services it represents…

SaaS (Software as a Service):

Geared more toward users. This can be anything from a website, CRM webapp, to a REST API. The point is, the data/interface is made accessible but the hardware details have been sufficiently abstracted away enough that they no longer matter.

Basically, you take software and make it publicly accessible. Requirements such as installation, resource usage (ie memory/cpu), updates, etc are no longer relevant. You connect and it works.

PaaS (Platform as a Service):

Geared for use by developers. These include anything that has ‘hosting’ after it. Including webservers, email servers, dns management, etc.

Basically, the platform options are limited to whatever the is provided by the hosting company but they can be leveraged by developers to build on.

IaaS (Infrastructure as a Service):**

The newest addition to the party and where the ‘Cloud’ name probably originated. It’s geared toward providing a full system architecture (ie complete OS) that can be built on without any need to maintain physical devices.

Basically, developers are given access to a virtual machine to develop and deploy. Since the virtual machine is decoupled from the hardware, it’s much easier to migrate and clone that machine to whatever physical locations are necessary.

Where 5 years ago, providing region-specific hosting would probably involve a lot of manual work to clone the platform to servers around the world, IaaS platforms automate the process.

It provides a LOT more flexibility compared to PaaS because the developer gets full access control over the VM. On top of that, the number of physical machines actually hosting the image can be easily increased/decreased to match demand (ex such as during peak access).

It’s no longer necessary for businesses to worry about downtime or scalability. IaaS costs more than PaaS hosting because it requires more resources but it’s still significantly cheaper than hiring dedicated systems administrators and providing the bare metal in-house.

There are probably hundreds of different types of _aaS platforms that exist in the wild but it all boils down to one concept. Computer hardware has been abstracted away to the point where systems have become a commodity that can be traded at whim.

Need a thousand clones up and running in 10 minutes for the Super Bowl, not a problem. Need them scaled back to 10 just as quickly, also not a problem. Need clones to do heavy number crunching? Yep, those exist. What about massive amounts of storage space to host media? Just as easy.

IT infrastructure in general is not revenue generating so the only gains to be made will come from minimizing cost. One way to do that is reduce/eliminate/automate the infrastructure as much as possible. At the end of the day, all the developers want and need is a platform to build their services on. Companies like Google/Amazon/Rackspace all specialize in massive scalability so doesn’t it make sense to tap into their infrastructure?

The disruptive change that ‘Cloud Computing’ represents is that it’s no longer necessary for anybody but designers, developers, and creative/media types to own computers that include a full OS. The web, games, documents, social applications, business applications, everything is being made accessible on the web.

In addition to Chloe’s excellent answer, I would say that the following factors have caused cloud computing to explode in popularity:

1. Growth of Internet use, and hence, Internet-related services (am including mobile services here, which mostly use Internet functionality)
2. Need for cheap, homogenous, easy-to-setup hardware for companies, startups, etc.
3. API-based (programming) control to set up new servers and scale them up or down

I personally think #3 is most important: if you are managing 100s or 1000s of servers, would you rather do through a command line or a GUI … or drive miles to get to your co-location services?

Of course, even if these had happened, it could have not been possible to do without the level of virtualization technology we have today, which directly helps point #3.

In short, I would say it was a perfect storm of factors that have enabled cloud computing to exist today in its current form, and to grow rapidly in popularity.

The defining characteristics of cloud computing are scalability and utility billing. Client/servers, thin clients, and ASPs mentioned by Jörg W Mittag are not cloud computing, unless they automatically scale up and down in real time with the customer being billed for the amount of resources (CPU, disk space, bandwidth) that they use. As Chloe correctly noted, this model only became possible with the advances in virtualization technology and high-speed Internet connections. These have been taking place within the last 5–10 years, hence this is when we started to hear about cloud computing.

Cloud computing is really time-sharing computing/shared hosting (very old models!) on modern hardware with virtualization to make things appear nicer (but again, that’s pretty old; IBM has been doing virtualization on their hardware for decades). It’s also what we do with that technology when we have the beginnings of ubiquitous networking. And it is the business model that goes with it; the ability to hire computing power or storage for very short amounts of time at minimal cost^[*] greatly changes how you go about planning and using those resources.

So… the major technological change was network ubiquity, but that’s not such a big thing really; the edges of that have been around for the whole of my professional career. No, it’s the business model innovation that was the real difference. A sane way of making it work financially for all concerned without complex multi-year account management was the missing piece. I’m not quite sure who invented it first either: the earliest I know of is Amazon AWS (who have been thoroughly copied) but I really don’t know if they were borrowing off of others.

So don’t knock the business-heads who are raving about this like it’s the greatest thing since sliced bread: I suspect they’re actually right and most of us techies are just too focused on the implementation aspects to see it (and those have huge precedents in this case). Innovations that enable new major classes of use are important, even if they’re not in areas that we know a vast amount about.

^{[* Not just monetary cost either, but also opportunity costs too. Being able to quicky respond to incidents is very valuable.]}

For cloud computing there have to be players in market that have the expertise and the infrastructure.

What Amazon is offering is basically an extension of what they already made for their own infrastructure.
In a “what works for ourself may work for others” they made it available.

Nice answers but this all started with the phone network was still a regulated monopoly. The structure of it was worldwide 99.99999 uptime and the ability to tolerate faults fault tolerant and highly available. System wide management of resources and early detection and preventative maintenance ensure that the bones of the system continue to work. Now with those concepts you begin to understand how critical system wide management of resources and the infrastructure of the network is essential now you build software communications layers on top of that tcpip was not the first then you can layer application protocols and build your applications to tolerate faults as well and make your redundancy geographically dispersed so that physical disruption does not cause outages and that’s one heavy cloud over your parade.