At USV we have been thinking a lot about the nature of competition as the cloud computing era takes hold.  The three closely related questions we are interested in are:

1. How does cloud computing change competition at the “application” layer?
   

2. How does cloud computing change the boundary between “infrastructure” and “applications”?
   

3. How does cloud computing change competition in the “infrastructure” layer?

Application and infrastructure are in quotation marks because they are convenient shorthands but cloud computing is changing their nature.  When all is said and done, we may have new terms or at least new meaning for those terms. I will lay out some thoughts on each of these over the next three days starting with the first question.

Cloud computing will further reduce the cost of getting started with an application but it does not stop there.  If done right, it will also reduce the currently still big cost of scaling an application.  And, as importantly, it will further shift the balance from “how?” to “what?” in application development. This is a continuation of trends that have already been at work over the last decade and have been central to our investment thesis.  In this view, defensability has shifted from technology to data leverage and network effects.

Cloud computing, however, has the potential to even erode those.  For instance, it will be harder to accumulate a truly proprietary data asset.  With data in the cloud, if a startup can find a “small cracK” that gives enough of a toe-hold that the origin points of data will want to send a copy to the startup (at essentially zero additional cost), then the startup can start to build up their own data asset.  Similarly, network effects may be weakened as the cloud makes it easier to connect to existing networks.  Yes, network owners will try to fight it, but the “small crack” argument applies there as well.  With dramatically reduced cost, a new network can emerge at the fringes and afford to take its time for the network effect to really kick in (arguably, that’s the strategy pursued by FriendFeed).

I believe this will have a number of interesting effects. First, questions of data ownership and API access restrictions will become more contentious as incumbents try to protect their data assets and networks.  Second, startups that do succeed in building their own data assets or networks will likely command high exit valuations as incumbents will pay up to sustain their defensibility. Third, venture investing economics in the application layer will be further compressed as even less capital is required, potentially over a longer period of time and with little visibility into whether the “small crack” can be leveraged until the data or network effect actually emerges.