The theory when it comes to startups is to take a lot of shots with the hope that one of them will go in. Build a startup during a hack-a-thon. Ride a bus to Austin and by the time you arrive an LLC will have been formed. Its easy: just pursue absolutely anything and make sure you call it a startup.
In the fall of 2008 I had a new idea: an email service for cloud based applications. At that time if you hosted an application on Amazon's servers and needed to send an email there weren't many options. Amazon's cloud was seen by the rest of the internet as a giant blacklisted entity. I presume many abused the ability to dynamically spawn servers with new IP addresses and soon entire octects were banned. If you wanted to send email you'd have to hit up SoftLayer, ServerBeach or The Planet and run your own dedicated hardware. This would not be a cheap option if all you needed it for was just to send email.
I saw an opportunity to offer an email API that would be priced on a usage basis. Companies would use it to avoid the minimum dedicated cost of $72/month and keep their own deployment simple. This service would eliminate the need to maintain an email system and be vastly cheaper than the status quo.
My first version was actually an HTTP based API until I went to use it myself as a developer and found that it sucked. Then I realized that the API for email was really SMTP. I built an SMTP server with Java and OSGI. It would accept SMTP calls, increase the meter by one for the authenticating account and deliver the message.
The more I thought about it the more I realized that Amazon would eventually fill this hole and offer it as a service. This pain couldn't continue for much longer and they would beat the shit out of me on price.
In the summer of 2009 a new startup called SendGrid came about with the same idea. They went on to raise $27.4 million dollars. They are a big company now. A former Oracle executive is now their CEO. They still fucking deliver email.