Clouds are an immensely over-hyped area of tech right now.
Everyone claims to be a cloud, have a cloud, let you make a cloud, enable cloud in some way.
But not every tech company really has a cloud. And most of what passes for a cloud from most vendors isn't really cloud at all, just a reworking of existing enterprise product lines.
Real clouds are geared to saving money, enabling use of all resources on one task at a single burst, and supporting applications written in any operating system.
Google (GOOG) is this kind of cloud. It does virtualization, it does distributed processing, and its design is geared to saving money in every way possible. Google's MapReduce is the technology on which all other clouds are based. Google actually builds its own servers from parts and saves a bundle that way. They open windows on server farms because it saves money, move small server farms to remote locations because it saves money.
The problem with Google, if it is a problem, is that Google runs its cloud for Google, runs Google applications almost exclusively. Use of the cloud is tightly controlled. Nothing wrong with that, but it provides opportunity for others.
Amazon.com (AMZN) took that opportunity. Amazon's EC2 is the first general purpose cloud, a cloud you can rent for whatever purpose you may have. Amazon uses most of the same technologies Google does, but it's impossible to know exactly how much each spends for a given level of performance. We just know what Amazon charges for its cloud, and assumes it makes money at it.
Microsoft (MSFT) is partly cloudy. Yes, Azure is a cloud, its server farm that uses distributed computing techniques that can handle bursts and like Amazon you can rent its capacity. But how virtualized can a cloud be when it's devoted to one company's technologies? A proprietary cloud is something of a contradiction in terms. Microsoft, once considered a leader in many ways, is the “me-too” cloud, supporting a “me-too” ecosystem of products and services.
Apple (AAPL) iCloud is not a cloud at all. It's one step down from Microsoft's Azure, supporting only a narrowly defined set of Apple services. Is it virtualized? No. Distributed? Maybe. The company hypes a big data center it's building in North Carolina, but the whole idea of a cloud is that it's not supposed to be a single data center, but a structure under which data and programs can live in many different places.
IBM (IBM) claims it can make you a cloud. By that it means it will build out a data center that runs a cloud platform, supports a virtual environment, and distributes the load within that center efficiently. It has built clouds for both public and private clients, but these are isolated systems. They're more efficient server farms, they run cloud software, but are they really clouds?
RackSpace (RAX) both supports development of an open source cloud stack called OpenStack and runs its hosting service as a cloud. They eat their own dog food, and it's delicious. OpenStack is based on work originally done at NASA, it has a lot of contributors, meaning what you can do with it keeps growing and changing. It's both cloud technology and a cloud.
Dell (DELL) and HP (HPQ) claim they're going to build clouds, that they're going to become like Amazon or Rackspace real soon now, and that they understand or have implemented cloud technologies. Both have large consulting arms that are struggling to get their arms around how cloud requirements change the nature of enterprise computing. No, they're not cloud now. They promise cloud.
VMware (VMW) and Red Hat (RHT) are among the cloud software “arms merchants.” Both offer stacks of cloud software you can use to build something that might run on your own hardware or might run on a public cloud – your choice. Both want to be the Microsoft of the cloud, and both are publicly-traded. Unlike other arms merchants such as Eucalyptus Systems, Cloudera and a host of others. Not to mention all the enterprise companies trying to build security, privacy and other enterprise functionality into clouds.
Notice I didn't mention scalability. The whole idea of a cloud is that this problem goes away. That is, you can scale to the limits of your hardware, and problems or functions can grab all that capacity in bursts if needed. That doesn't mean limits don't still exist. But they're mostly hardware limits.
One point that isn't made about cloud now, but will be made about cloud in the future, is cost. We assume that clouds cost less than enterprise systems, based on their architecture. But what's the relative cost of running public vs. private clouds, or running on Amazon vs. Rackspace vs. Google? There remain many investment questions concerning clouds that remain quite foggy.
As an investor, just know this. Not everyone who says they have cloud has cloud. Amazon and Google have cloud, everyone else wants to claim cloud, and the race will go to whoever can get costs down fastest and move enterprise system capacities up into the cloud.
Right now the sweet spot is in cloud software, in RAX and RHT and VMW. That will shift, over time, as hard questions are asked about the cost of actually running a cloud. Here GOOG and AMZN and RAX (again) have an advantage, because they have the best visibility on those costs, given their experience. While RAX may indeed be overvalued based on results, hype getting ahead of reality, action is in time going to move from software to management and operations.
Which is why in the end I like my 20 shares of GOOG. If they went head-to-head with Amazon's EC2 as a general purpose cloud host, I would like my chances. The question there remains whether Larry Page can monetize the advantages of his superior infrastructure. Your answer is as good as mine.