If all Google did with Android was to make another closed, proprietary stack, it's unlikely it would have gone all that far. In fact, everything Google did was designed to grow Android as far and as fast as possible. The most fundamental of these was to make Android itself open source.
This strategy has been a raving success, with Android clearly the 800 pound gorilla of the handset market worldwide, but it hasn't been without cost; the huge variety of Android versions, devices, and screen geometries, plus the fact that every phone has code from both the manufacturer and carrier makes the configuration matrix huge; all this makes development and testing a challenge. Compare this to Apple, which has a few phones, on which all the code is theirs, and nearly all users run the current version of the OS (well maybe not today, but soon everyone will go up to iOS7).
But anyone can take the Android source code, modify it and make their own distribution. Perhaps more than any other open source project, this has been a popular feature.
The biggest group creating custom Android versions, or mods, is Cyanogenmod,. They've got venture capital and they're already working on things that they didn't have time for when it was just a hobby.
This is very cool in many ways. Right now, nearly everyone (in the US anyway) gets their copy of Android from one of the big carriers. These companies 1) install crapware on the phone that you can't remove without rooting it, and 2) usually don't provide upgrades of Android, even when they would run just fine on the phone.
Of course they do this because they're in a position to push you around and they want to maximize the amount of money they can get out of you. The lack of Android updates is a tactic to make you want to get a new phone when the new phones come out with a new, better Android. They lose money on that new phone, but they make a ton on the 2 year plan you buy in order to get the phone at an affordable price. And when a problem develops in your 20 month-old phone with an old Android version on it, nobody involved has much of an incentive to do much to fix it.
Cyanogenmod and the other modders aim to give you OS options for your phone. You can get innovative features that Google and the carrier won't provide, you get no crapware, you get updates very quickly, and it's all based on the same code that the carriers use to make their lame versions.
But there's essentially no chance that modders can take significant market share from the carriers: the main reason for this is that the throwing arbitrary distributions into the mix limits the ability of carriers to support phones. It may be that if you call up or go into a store, the carrier will do all they can to help you, but the odds that you'll get competent support for a modded phone are low.
Very few ordinary users will dare going this way, and it would be a bad decision for them to do so. Many of the support calls the carriers field amount to "How do I this Android thing?" If the customer could have any old distribution, support could not function, at least not with affordable employees working off of scripts. In this, the carriers are actually being reasonable.
A better market for modders, I think, is device resale. There are already many places you can buy used/refurbished phones and tablets, but perhaps those organizations could mod the phone up to the current Android level. That would make it much more appealing. It's not going to have support from the handset maker or carrier anyway, at least not for Android.
But all this ends up further fragmenting Android. Even if the versions all tend to get more modern (as they definitely are), the sheer variety of target platforms has to increase the number of potential problems.
So I'm happy for Cyanogenmod. Read their story, it's hard not to like them. And I wish them well, but I'm under no illusions that it will help with the fragmentation problem. The world of Android just keeps getting more and more complicated.
[This story was edited to correct typos.]