VCs should choose their investments by the expected return, not the amount of money they require.
That server farm costs a lot, and therefore the required investment is a huge barrier to entry.
That business model is perfect for VCs, and that is where they should focus.
When they became the default search on Netscape, they had 30 computers.
Edit: Never mind, they switched over to Google in July 1999 (http://searchenginewatch.com/2167331), when it was quite possible that they still had 30 computers.
AdWords was not launched until 2000 so it wasn't possible they were turning a profit before then.
In fact, IIRC it was Google's success at doing the "Get users first and profit will come" that made it fashionable among Web 2.0 startups. The idea was pretty disreputable after hundreds of Web 1.0 startups tried it and flamed out spectacularly.
However all is not equal, the extraordinary returns in recent memory have come from the web, and so venture capital firms will likely still be putting their thumb in the pie, even if it means cashing out to founders, letting them be CEO, giving them smaller rounds, etc. for software start-ups.
It seems more popular to think of VC as working capital here: start the company on $20K, prove that it works, then take VC money to scale. Hell, that's the whole YCombinator model.
I think the VC should instead concentrate on businesses where their investment can make a real difference: Business plans that require a factory, or expensive custom hardware, thousands of servers, gigantic customer education, etc.
I know it's heresy on YC, but real VC's shouldn't be involved in companies where only $20k is required to start the business. Founders are taking the VC for a ride when they convince them to fund extravagant startups that should be started without big investment.
VC investors, if they still have cash, need to think bigger. They're not needed for software startups.
The amount of funding required to start a business is one of the best barriers to entry, and the VC should exploit that.