This is my third post on cloud competition.  Today I want to address the question of how cloud computing changes competition at the infrastructure layer.

A part of the answer comes from yesterday’s post on the changing boundary between the application layer and the infrastructure layer.  As infrastructure pushes up ever higher, the lower portions of  infrastructure become further hidden and abstracted away.  AMD has not had a lot of luck in their effort to compete with Intel and make money at the same time.  In the traditional world of dedicated hosting some folks still cared about whether the server they were getting had Intel inside.  That’s the power of brand and also the perception that Intel might just be better.  But if you get your storage via Amazon S3 you stop caring.  As long as they deliver storage, who cares what they use?  Might as well be highly trained hamsters.

The same argument applies for other lower level infrastructure components, such as load balancers.  This means that the defacto number of customers for processors, load balancers and so forth will decline substantially over time and become limited to the far fewer cloud providers.  Usually when there is increased concentration in the customer base it also results in increased concentration in the suppliers.  For processors that has happened already, but I believe the same thing will play itself out for those infrastructure components that have remained somewhat independent but will be below the “surface” of the cloud.

Who will the cloud providers be? Google, Amazon and (probably) Salesforce are good starting bets because they are clearly pushing in that direction, have offerings in the marketplace and have cash generating businesses they can use to finance their cloud efforts.  But there are also a lot of startups that have been and continue to be financed with a wide array of approaches. These range from companies such as Etelos, that promise to take existing apps and put them in the cloud to companies such as Bungee Labs, that have created their own language (Salesforce has their own language too).

For developers there is a big risk going with anything proprietary since it implies lock in to a cloud provider.  Someone like Salesforce or Google has a shot at getting away with it to the extent that they can hold out the promise traffic / usage based on their existing user base.  But for startups like Bungee that seems like a tough way to go – what is the quid pro quo for using their proprietary approach?  The promise of increased productivity is bunk as I have recently argued in my post on language choice.

But the other approach for startups of embracing an existing framework as Heroku is doing with Ruby on Rails also has a big issue.  This time there is no lock and there is an existing potential base which is great for getting started.  But it raises the spectre of commoditization.  What if google offers scalable Ruby on Rails hosting (and why wouldn’t they?).  Now I can pull my app from Heroku, drop it into google and repoint my DNS and be done.

So cloud computing will be a tough area for startups to play in, which fortunately does not stop them from trying.  I believe the success here will come not from trying to outspend the big boys (unlikely) or trying to lock customers into something proprietary or offering a commodity service.  I have some ideas that will be the subject of future posts. 

This is my third post on cloud competition.  Today I want to address the question of how cloud computing changes competition at the infrastructure layer.

A part of the answer comes from yesterday’s post on the changing boundary between the application layer and the infrastructure layer.  As infrastructure pushes up ever higher, the lower portions of  infrastructure become further hidden and abstracted away.  AMD has not had a lot of luck in their effort to compete with Intel and make money at the same time.  In the traditional world of dedicated hosting some folks still cared about whether the server they were getting had Intel inside.  That’s the power of brand and also the perception that Intel might just be better.  But if you get your storage via Amazon S3 you stop caring.  As long as they deliver storage, who cares what they use?  Might as well be highly trained hamsters.

The same argument applies for other lower level infrastructure components, such as load balancers.  This means that the defacto number of customers for processors, load balancers and so forth will decline substantially over time and become limited to the far fewer cloud providers.  Usually when there is increased concentration in the customer base it also results in increased concentration in the suppliers.  For processors that has happened already, but I believe the same thing will play itself out for those infrastructure components that have remained somewhat independent but will be below the “surface” of the cloud.

Who will the cloud providers be? Google, Amazon and (probably) Salesforce are good starting bets because they are clearly pushing in that direction, have offerings in the marketplace and have cash generating businesses they can use to finance their cloud efforts.  But there are also a lot of startups that have been and continue to be financed with a wide array of approaches. These range from companies such as Etelos, that promise to take existing apps and put them in the cloud to companies such as Bungee Labs, that have created their own language (Salesforce has their own language too).

For developers there is a big risk going with anything proprietary since it implies lock in to a cloud provider.  Someone like Salesforce or Google has a shot at getting away with it to the extent that they can hold out the promise traffic / usage based on their existing user base.  But for startups like Bungee that seems like a tough way to go – what is the quid pro quo for using their proprietary approach?  The promise of increased productivity is bunk as I have recently argued in my post on language choice.

But the other approach for startups of embracing an existing framework as Heroku is doing with Ruby on Rails also has a big issue.  This time there is no lock and there is an existing potential base which is great for getting started.  But it raises the spectre of commoditization.  What if google offers scalable Ruby on Rails hosting (and why wouldn’t they?).  Now I can pull my app from Heroku, drop it into google and repoint my DNS and be done.

So cloud computing will be a tough area for startups to play in, which fortunately does not stop them from trying.  I believe the success here will come not from trying to outspend the big boys (unlikely) or trying to lock customers into something proprietary or offering a commodity service.  I have some ideas that will be the subject of future posts.