I am frankly bemused that anyone seriously believes Microsoft or anyone else is going to fund their on-demand applications from advertising revenues. The idea is complete bull, on two counts.
First up, ads in applications don't work. All the evidence from the past ten years of online services is that the more engaged the user is in pursuing an activity, the less likely their attention will be diverted by an ad. The sole exception, which of course is why Google AdWords has been so spectacularly successful, is search, for the simple reason that clicking on relevant ads is a natural extension of the search activity. That's why advertisers pay more for ads in search results than they do for contextual ads in content pages— and it's why the terms of service for partner web sites that carry Google ads explicitly forbid placing the ads on application pages.
What about Gmail — isn't that an application? Well, Google hasn't published any metrics for Gmail, but it's fairly obvious given the amount of storage you get with a Gmail account that Google aims to monetize Gmail from users clicking on ads when they search their archived mail, not when they're writing emails. I mean, it stands to reason doesn't it? Are you going to click on an ad right in the middle of crafting an email? Only if you've hit a patch of writer's block. Whereas monetizing ads in search results is a tried-and-tested model at which Google already excels.
The second factor is that the world does not revolve around advertising, it revolves around trade. Businesses need to be able to make and sell stuff before they have any money left over to spend on advertising. By definition, therefore, there is far more money to be made from directly supporting businesses in the making and selling of products and services than there is from vying for the small change they have left over for advertising. The majority of on-demand revenues will come not from advertising but either from taking a slice of people's sales revenues (the Amazon.com and eBay model) or by providing services that help them operate more efficiently (the salesforce.com model). Google is not the model to follow, not until it manages to wean itself off its total dependence on advertising and expands into other services (which is why GoogleBase is so interesting, by the way).