+ Vendor/plat-form lock-in.
+ Data mining startups' traffic.
+ Access to startups' code and data.
+ Ability to pull the plug.
For Vendor/Platform lock-in this could be virtual, ie the burden of switching the code base for administrative tasks to Amazon or Azure or some other provider could be prohibitive. Likewise for architectural layout.
In the bigger picture, is this a first foray into the space where MicroSoft Bizspark lives?
But I think your Point 1 is highly likely. How many startups are going to build on Google's Cloud Platform, then after a year port everything to AWS, particularly if Google can prove that they're cheaper or at least cost-competitive with similar feature set. (Assuming that's true, of course, and that you don't need something specific to AWS)
Google's internal data, which is mined (and audited), and the Cloud datacenters/data, are hosted entirely separately (though that's as much to protect us from you as you from us).
We do not do anything with data stored by cloud platform customers.
Absent these, what is the business case for providing this service at no cost to the qualified companies?
As I indicated, I don't see a first order monetization strategy and the second order strategies I have proposed seem consistent with the second order strategies general employed by the company.
Regarding the business case, Google has a LOT of compute power. The cost of that compute power on the cloud platform is not the cost in hardware and electricity to Google, it also pays for the engineering. Whether or not people take up Google on the 100k offer, the engineering has been paid for, so really Google is only giving away the electricity and maintenance. I don't know (and if I did know I wouldn't say) how much of that 100k is actual realized cost for Google.
But however much it is, that is at most 100k of REALLY EXTREMELY WELL targeted advertising for the Google Cloud Platform. A customer that goes through 100k of Google compute resources in a year is likely one that would rather continue paying Google's competitive rates for the same services in the future.
It's hardly lock-in if the reason for not leaving is "we're cheaper than the competition". That's just having a better offer.
(disclaimer: Xoogler here)
But startups could easily use AppScale to move off Google and on to AWS/Digital Ocean etc. when they start to get real traction or when the free credits run out.
From the general terms and services:
The rights you grant in this license are for the limited
purpose of operating, promoting, and improving our
Services, and to develop new ones. 
In addition, the cloud services have this extension:
Google may change these terms from time to time and post
any modified terms at http://developers.google.com/site-
terms. You understand and agree that if you use the
Service after the date on which these terms have
changed, Google will treat your use as acceptance of the
updated terms. 
I don't think it does. Until recently, they seem to have been using the paid apps for Business e-mails to build advertising profiles with no repercussions.
I also don't think there's anything in the T&Cs for Google Drive(both free and business) that says that they won't use the contents of the docs for advertising purposes.