And that's only half of it. Users that are likely to request an invite are more likely to be technically savvy and provide better feedback than "maps is b0rked. plz fix!!11". If you are going to have to deal with feedback, it's much better to not have to sift through low quality feedback during the early days of launching a new product.
Spoken like someone who has never tried to buy tickets to Google IO!
But on a serious note, there are some products at Google that clearly scale well beyond what most other groups are capable of, but there are some products at Google that... don't.
Like any big company, things aren't completely uniform across the board.
I believe the "drip-feed" invite idea suffocated Google Wave. When it first came out I was begging my inbox & skype for invites. Once I finally got in it was ... a wasteland. I dished out invites like there was no tomorrow but in all honesty, if people could JUST sign up, I might have had someone to talk to.
Who knows, maybe it might still be running. I'm not sure if there were other factors but personally it never made sense to me.
I believe if Google really wanted to, they could go from 0 to 1 million users in minutes.
You mean instead of handing out N invites, just pick N users at random? Why would that be better? You confuse and potentially upset the randomly-chosen users, and you disappoint the eager people who actually want to try the new thing.
When google does first-come first served (Google I/O last year) they get dinged. When they do randomized (Nexus4 initial sale and this year's I/O) they get dinged.
Fact is, they can't satisfy everyone. The algorithm used to differentiate the few from the many will obviously upset some groups.
Would you rather they charge for it? That'd be the Microsoft thing to do [1].
The problem with Google I/O is that it was not actually first come first served, but random. As in: the registration site gets pound so hard that you need to be lucky to be able to load the page and proceed with your registration.
Also, for Google I/O it would have served to have people apply, and review applications to pick real developers and not dumbass unable to read a line of code but hoping for a cool gadget to flip on ebay.
Anyway, as much as I want to try the new map right away I would do the invite system like them: you get to ramp up your userbase progressively, and people trying it actually want to try it (not grandma opening Maps and being completely confused by the new UI and possible bugs).
I suppose neither are ideal, really. With an invite only model you run the risk of biasing your feedback. The advantage of a random rollout would be the marketing sleight of hand: "gradual rollout" might be more defensible than "by invitation".
Gmail and Google Groups have both had random rollouts for changes, with optional rollbacks. That seems more fitting (to me at least).
IO is the place to launch new products. It's just slightly more impressive than showing a message "And btw. we've lunched this new version of maps. Check it out (with the probability of 0.00053846)".
Sure Google can handle scale better than anyone, but a stampede is still a stampede. Soft launches are way safer and easier to manage.