Because the number one thing openness generates is chaos and multiple competing claims about reality. Say, many Linux distributions, each claiming to be great, and in fact, many variants of Linux distributions often with many versions and many wrinkles, and many varations of packages, libraries, and so on.
If you want to build or customize things, openness is great. If you just one to pick something up, use it, and move on, a huge amount of confusion, overhead, and pain is involved in trying to pick the "right" version (particularly if you're unfamiliar with openness and wrongheadedly looking for the "real" version, as many early Linux dabblers were) and get it to work quickly and easily.
There is thus a huge amount of value added by anyone that quells the chaos—even in a tiny sphere or product—and that can quickly, clearly, and succinctly explain to users just what their version does, without ambiguity either within itself as an instance or over time. The nature of the beast—this value is the result of "closing the openness," if you will, means that it can't be opened, or the value will be lost.
End users want operating systems and devices that are not open systems with unclear edges that bleed into the ecosystem, but rather a single, coherent, object or product that they can acquire, use in predictable and stable ways, and then lay down once again. They want systems and devices about which books can be written (and bought, and referred to months down the road) without quickly becoming obsolete, and with the minimal risk that this book or that add-on that they purchase will fail to work becuase they'd misconstrued the incredibly subtle differences and variations in product naming, versioning, and so on.
In short, massive openness is incredibly generative and creative, but leaves in place a systems/software/hardware version of the "last mile problem" for computing. Having a fabulous network is one thing, but consumers just want one wire coming into the house, they want it to work, they want it to be predictable and compatible with what they have, and they want to know just where it is and what its properties, limits, and costs are. They are not interested in becoming engineers, the technology they use is only useful to them as a single, tiny, and managable facet of the larger ecosystem that is their life.
This "last mile problem" cannot be solved with openness in hardware or software any more than the last mile problem for wired providers can be solved by opening up all of urban geography to any comers and saying "lay all the cable you want, anywhere you want, to and from any building or system!" First off, it would result in a mess of wires (not un-analagous to what we see across much of free software's development space) and next because most consumers wouldn't be able to make heads or tails of it, much less make a choice, and they'd probably resent the complexity in their backyard and try to do away with it.
Openness leads to closedness because to the extent that openness dominates in the development and engineering space, closedness increases as critical need for carrying whatever is developed to the average consumer space, in precisely the same measure.