My prediction, the SteamBox will fail. Why? It is a PC that plays no third party triple A games on the market (can't play COD/Assassin's Creed/GTA V on Linux). I simply don't think enough people will buy a console to play just Linux games. The fact that it is an open architecture box will lead to poor customer satisfaction.
The geek market won't buy SteamBoxes, they build their own PCs. The console market desires simplicity and ease of use, they don't want their console to be obsolete because they bought the "cheap" model. Therefore, they won't buy it either.I also suspect that the price will be another factor that plays against the SteamBox.
I don't see an extremely profit-now-driven company like Valve deciding to sell SteamBoxes at a loss hoping to make money back later. The manufacturers of the SteamBox (FalconNorthWest, Alienware, etc) will charge a steep price for the machines because they have to turn a profit on every SteamBox sold and Valve will not knock the price down by subsidizing the manufacturers. So unlike Sony or Microsoft, Valve will be selling these boxes for a profit on a per-box basis. Consumers will experience sticker shock, much like the 3DO. Also, much like the 3DO, there are no third party triple A titles for the SteamBox. Heck, as of right now, there are no first party triple A titles either!
The only way Valve makes people buy a SteamBox is if they make SteamBox exclusives. So if HalfLife 3 is released on the SteamBox as a timed exclusive, people may buy it just to play HalfLife 3. So maybe that will drive 10 million steamboxes to be sold. But 10 million is not enough, you need 30million units of a console to be sold to turn enough of a profit to make the entire business worthwhile.