In the client-server world the name on the box meant everything. You bought HP (NYSE:HPQ) boxes or you bought Dell (NASDAQ:DELL) boxes or you bought IBM (NYSE:IBM) boxes. The data center was a collection of servers.
With the cloud that's all changing. Google (NASDAQ:GOOG) is now one of the largest server OEMs there are, not because they are selling servers but because they're assembling parts in their own server rooms, building their own.
So the box no longer defines a data center server. Instead the chip does.
Thus, chip companies like Nvidia (NASDAQ:NVDA) are now producing modules that are specifically for cloud-based datacenters, essentially plug-in cards that cloud service providers can do to accelerate graphics performance. And ARM Holdings (NASDAQ:ARMH) is publishing, through a start-up called Calexda, a "road map" so that its chips can be "virtualized" to the cloud.
This last story points to a second big data center trend, the idea that hardware eventually becomes software. Make your architecture the one data centers choose and you will always be getting business, goes the thinking.
As an investor, just remember the two rules in considering investments in the chip sector. Boxes are becoming chips, and chips are becoming software, as the future grows cloudier for everyone and the data center becomes more virtual.
In the end it's all software. This is something ARM figured out fairly early on. The strategy limits revenue opportunities, but it also limits costs, allowing the company to focus on the software design that can make its offerings unique enough to sell but generic enough to be worth tweaking.
Until hardware companies like Intel (NASDAQ:INTC) figure out the new game, they will remain in trouble.
Why does all this support ARMH over INTC? Because ARM has always been nothing more than a design -- it's a software company that other people use for chip designs they then see produced, to order, in third-party foundries. INTC, by contrast, makes chips, controlling the whole process soup-to-nuts.
For data center customers, it's much more convenient to turn ARMH software into custom hardware, on your schedule, than be stuck with whatever capabilities a corresponding INTC chip may have.
As they said with the New York Yankees in the 1950s, so it may be true today. Break up Intel.