There's a million people who could use this $100k to mine bitcoins, run personal servers, do scientific calculations, or more. Google isn't interested in those people: they want to lock in companies and keep them there as they grow quickly. If you worked at Google, what would you propose to catch these long-term valuable customers, without giving away a ton of free infrastructure?
By screening for "you are in an accelerator" you not only have ensured some due diligence has been done, but you also know someone else's money is on the line, so they will keep tabs on them.
Seems a little late to lure them into the Google Platform.
But if some startups have planned for portability and would rather stretch their OpEx runway instead of paying Amazon, switching could make sense.
Hope this helps --
head of marketing, google cloud platform