One of the reason businesses have trouble really innovating is that it's hard in a business to work on long-term things when markets are very short sighted. Only mega-corps, monopolies, and governments can usually do that... or hobbyists / lifestyle businesses who are more casual about hard business demands.
That being said, MS is surely cash-rich enough to think long term. So this doesn't apply as much here.
I've also found that of all things optimization almost gets you looked down upon in most teams -- even young ones. "Premature optimization is the root of all evil," and all that, which is usually misinterpreted as "optimization is naive and a waste of time." It's seen as indicative of an amateur or someone who isn't goal-focused. If you comment "optimized X" in a commit, you're likely to get mocked or reprimanded.
In reality, "premature optimization is the root of all evil" is advice given to new programmers so they don't waste time dinking around with micro-optimizations instead of thinking about algorithms, data structures, and higher order reasoning. (Or worse, muddying their code up to make it "fast.") Good optimization is actually a high-skill thing. It requires deep knowledge of internals, ability to really comprehend profiling, and precisely the kind of higher-order algorithmic reasoning you want in good developers. Most good optimizations are algorithmic improvements, not micro-optimizations. Even good micro-optimization requires deep knowledge-- like understanding how pipelines and branch prediction and caches work. To micro-optimize well you've got to understand soup-to-nuts everything that happens when your code is compiled and run.
Personally I think speed is really important. As a customer I know that slow sites, slow apps, and slow server code can be a reason for me to stop using a product. Even if the speed difference doesn't impact things much, a faster "smoother" piece of code will convey a sense of quality. Slow code that kerchunks around "feels" inferior, like I can see the awful mess it must be inside. It's sort of like how luxury car engines are expected to "purr."
An example: before I learned it and realized what an innovative paradigm shift it was, speed is what sold me on git. The first time I did a git merge on a huge project I was like "whoa, it's done already?" SVN would have been kerchunking forever. It wasn't that the speed mattered that much. It was that the speed communicated to me "this thing is the product of a very good programmer who took their craft very seriously as they wrote it." It told me to expect quality.
Another example: I tried Google Drive, but uninstalled it after a day. It used too much CPU. In this case it actually mattered -- on a laptop this shortens battery life and my battery life noticeably declined. This was a while ago, but I have not been motivated to try it again. The slowness told me "this was a quick hack, not a priority." I use DropBox because their client barely uses the CPU at all, even when I modify a lot of files. Google Drive gives me more storage, but I'm not content to sacrifice an hour of battery life for that.
(Side note: on mobile devices, CPU efficiency has a much more rigid cost function. Each cycle costs battery.)
Speed is a stealth attribute too. Customers will almost never bring it up in a survey or a focus group unless it impacts their business. So it never becomes a business priority.
Edit: relevant: http://ubiquity.acm.org/article.cfm?id=1513451