There is little doubt that cloud computing constitutes a seismic shift in enterprise computing. Just as client/server computing shifted the balance of power from central mainframe data centres to distributed computing on servers and PCs, the cloud is shifting the centre of gravity of applications away from on-premise. Just as it is tough to spot a bubble in the financial markets until it is too late, such shifts in computing architecture take a long time to build up, and are often denied or ignored by those with vested interests in the status quo. Yet a recent large EMC study found that 46% of EMEA IT executives surveyed thought that cloud architectures would overtake traditional IT architectures within their organisation within three years.

The issues driving this change are easy enough to understand. A higher and higher proportion of IT budgets are taken up managing the complexity of the current enterprise software infrastructure. Adding new software or even upgrading existing software on-premise requires expensive testing, as the combination of infrastructure elements used these days generates many permutations of software versions that can affect applications, each vulnerable to some unexpected bug introduced by an upgrade in one layer or another of software. Enterprises do their best to lock down supported versions of their desktops and servers, but it is hard work, and is often resisted by their business customers who want answers today and are unaware of the complexity involved in their software plumbing. Moreover, unexpected spikes in demand may play havoc with in-house data centre capacity, meaning either idle expensive overcapacity to give redundancy or the risk of key applications slowing down or crashing under peak demand.

Cloud computing essentially outsources these problems to someone else: let the application software providers figure out how to manage a complex infrastructure upgrade, and let specialist data centre providers worry about capacity planning. The Amazon cloud service AWS, the largest such service, in 2012 stored 905 billion objects, representing probably an exabyte (1,000 petabytes) of data, and has grown since. Even the data centres of large multinationals are dwarfed by comparison. There is an analogy with leasing office space for a growing company: you either have to rent far more space than you need right now to allow for growth, or risk regular costly moves. It is far easier to lease office space on a flexible basis from a provider, though of course that provider will charge a premium for that flexibility.

Another key factor in the gradual move to the cloud is that the software licensing model for such applications has shifted. In the past enterprises negotiated multi-million dollar software licences with software vendors prior to installation, never sure whether they would be successful or what the level of take-up would be. In cloud computing it is normal to rent the software by the month based on take-up, so you know that if you are paying more then at least you are actually using the software more.

However the sheer inertia of enterprise computing means that this move is playing out in slow motion. A 2013 study by GigaOM of 855 respondents showed three quarters of companies now using cloud technology, with the addressable cloud market in 2014 likely to be more than double that of 2011. This still means that a quarter of companies have no cloud computing use whatever, and the footprint of cloud applications within enterprises may still be small in many cases. Nonetheless, the move continues inexorably, year by year, as the benefits are increasingly evident. Interestingly, the same survey showed that security was a lessening (though still significant) concern among cloud adopters.

Who will the winners and losers be as the cloud seeps into the enterprise architecture? Certainly end-user organisations should benefit from greater flexibility, and probably lower costs, though nailing down the true economic picture of on-premise versus cloud computing is a slippery subject. Cloud providers of both public and private clouds clearly win from rapidly increasing take-up.

Software vendors may be both winners and losers. The shift of licensing model to usage-based leasing may disrupt software sales organisations where compensation for hard-charging software sales staff has traditionally been driven by megadeals. It will be tough for many to adapt, and some are still in denial. It may be easier for newer vendors whose organisations are designed to operate in the new world from the ground up. Seemingly if you put a frog in a saucepan of hot water it will jump out, but if you place it in cold water and gradually heat the water, the frog does not notice the danger until too late and perishes. I wonder how many existing software companies are those frogs whose environment is gradually, almost imperceptibly, being changed by the onset of cloud computing. How many will react in time?