I'm not; if the experience of the last big entrant to try streaming games, OnLive, is any indication, that kind of service is an absolute furnace for cash. The Verge reported that at its height, OnLive was burning through $5 million per month to support a measly 1,600 concurrent users (http://www.theverge.com/2012/8/28/3274739/onlive-report).

It's possible that smarter architecture could bring that figure down; OnLive needed to have a physical machine for every concurrent player (!). And OnLive made plenty of other mistakes that a smart business like Valve might avoid. But even if you did everything right it still sounds like running a streaming games service at a profit with today's tech would be a serious challenge.

I agree that the business model probably doesn't make sense generally but Onlive isn't a useful example. Over $3000/user/month (5M/1600) sounds absolutely ridiculous. You could buy a new machine for each user every month, host it, license the software on it and still afford to pay the bandwidth.

It seems highly likely that the support costs for the 1,600 users were noise amongst the staff, marketing, licensing and other costs (including potentially wasteful overprovisioning based on expected sales).

Having said that if it cost even $10/month (and it may be more or less) to support a user the whole business model may break down.

There is no need for streaming like OnLive. On x86 platforms, I don't see why Steam couldn't emulate the Windows version, much like what Wine or Cedega does. This is one way that they could support cross-platform gaming on OS X, Windows, and Linux. But I guess instead, Valve is going after a common baseline to develop native code on those three OSes.