Any Hardware Company Not Investing Big in Cloud is Nuts
Posted by Bob Warfield on April 7, 2011
The problem these vendors face is “Cloud as disintermediator”. Companies that buy into the Cloud are letting the Cloud purveyor make the datacenter decisions. And Clouds aggregate a lot of those decisions under one roof to gain their economies of scale. It’s like waking up one morning as a hardware vendor and discovering your best customers had switched to another hardware vendor. The difference, is you can see it coming and you have a chance to do something about it.
In fact, hardware and infrastructure vendors have quite a lot they can do about it:
– Make it a point not to lose Cloud deals. When Cloud vendors come knocking, lock them up. There can’t be as many Clouds as there are traditional vendors, and each one will have a lot of inertia to stay with their original infrastructure choices. Make sure they choose you.
– Start your own Cloud. This one may be pretty risky if it locks you out of being the basic nuts and bolts of other popular Clouds, but you have to at least consider it. At the same time, Cloud vendors will need to decide how they feel about using say IBM hardware with IBM pushing their own Cloud. It muddies the waters.
– Invest in building products that are uniquely suited to the Cloud. There’ve been some fascinating glimpses of what large scale Cloud data centers need. There is a ton of intellectual property opportunity in the world of the Cloud. Now is the time to start staking your claims to it. Get out your blank sheets of paper, sign up with some big Clouds to work on their needs, and you will wind up with some good ideas. That’s only the start. In the commodity world of the Cloud, execution will be everything.
– While we’re on the subject of execution, think what the razor thin margins demanded by the Cloud mean to your business model of today. Think what they mean to the needs of the Cloud vendors. Skate as fast as you can to where that puck is going to be.
I liken the Cloud a bit to what CNC machinery did to the manufacturing world. We went from a situation where each and every machine required a skilled manual machinist to run it, to a world where computer-controlled machine tools can be run 4 or 5 to an operator and the operator can be far less skilled than a master machinist. The same could be said for servers. We’re moving from a world where every app had its own servers and lots of wasted capacity, to a world where we share that capacity internally via virtualization, and then finally to a world where we’re sharing capacity that is time sliced across many customers in huge data centers to wring every last drop of utilization. Yes, there are drivers like Big Data that offset some of that, but it is hard to escape the possibility that in the long run, the world may not need as many servers as it once did.
To get an idea just how different hardware purpose-built for the Cloud can be, check out Facebook’s servers. They’ve just open-sourced the designs for their server hardware, which is quite a bit more energy-efficient that their leased data center hardware. The difference in efficiency shows how much of an advantage can be had by specifically targeting the Cloud. Facebook was smart to open source, if only because it makes it more likely the design will be commoditized and hence easier for them to buy even more cheaply.