When you're small, you need a source of revenue quickly, or you'll die. You don't have a luxury to experiment too much, you have to build one thing that works.
When you're a huge company and have untold millions in cash, you can afford experimentation. In fact, you have to, because your revenue source is probably finite and you have to find another before the current source dries up.
This is why Google dived into mobile phones, self-driving cars, home entertainment, renewable energy, Internet providing, consumer cloud computing, goods delivery, etc. Some of the experiments became huge successes (see Android), some failed (see the long list of closed projects), some are too immature to judge (e.g. self-driving cars).
It just looks that experimentation is now locked at a thinner top exec level, while 'simple engineers' are not expected to do too much of it. Which is, of course, a pity.
When you're huge and cash-flush, you have the luxury of really innovating in truly hard areas.
This luxury is damn hard to obtain in a short-term gerbil-wheel economy like ours. Once you have it, throwing it away is like setting fire to a house as soon as you've paid off the mortgage or crashing your new luxury car as soon as you drive it off the lot. It is abysmally stupid and short-sighted. Investors should call for the heads of people who do this, literally. As in on a pike.
People think innovation comes from startups, but in reality it doesn't. Not because startups aren't smart and agile, but because they don't have the resources.
When I say innovation, I mean innovation. I don't mean application of existing innovations to new market areas or problem spaces. Startups excel at that.
But you'll never see a scrappy basement startup whip out a self-driving car, an artificial lung printed from a 3d printer, an orbit-capable reusable rocket, augmented reality goggles with a complete software stack, etc. Not unless the parts for those things already exist and can simply be combined in a novel way to yield a result in less than a year.
If you're big and cash flush, you can do what nobody else can do (except other monsters like you). You can do this.
That dent gets you a Ph.D, but if you do it for something of high economic value it gets you early entry into a market nobody else can enter because they don't know how.
If you do it, you've now created a piece of value that you can do one of many things with. You can feed it into your more short-term product-dev branches and do things nobody else can compete with, you can license it, or you can use it to pump up the prestige of your company in ways no advertising can.
"Holy crap! A self-driving car! And it really works. I mean, there it is, on the road, and it's driving better than I am and getting around faster than I am! I'm going to move all my business's hosting to Google Apps, cause they obviously have the smartest people on the planet..."
I do have the sense though -- and keep in mind I am an outsider -- that there might have been some "ADD" issues with Google's 20% policy as it was implemented. But I don't think the solution is to ditch it. The solution is to focus it, to try to get people to spend their 20% time pushing harder into deeper and more difficult areas instead of whipping out hacks. Maybe incentivize more people to work together more formally on 20% projects over longer spans of time, and incentivize them to tackle things that are very difficult... things only a big elephant can do instead of things a scrappy startup could do.
If an initiative doesn't seem far too wild, then it's not wild enough to be innovative. Quite a few lousy ideas seem too wild initially, but pretty much all of the actually valuable projects are as well.