Don't place me in the "doesn't want to pay for content" box. I am OK with paying for content and I do pay for many things online. But there are two issues with ACM:
1. The research has been paid for with taxpayers' dollars (I wasn't the taxpayer, but still).
2. $15 is really expensive.
And of course, if it were just one article that would advance my work a lot, I'd gladly pay. But you don't know that ahead of time. And if you're building startups, you usually do a lot of wide-area research, so it isn't that one article, it's hundreds of articles that you need to skim through.
I also don't buy the argument that we need to pay so much just so that we get peer-reviewed content. JMLR (Journal of Machine-Learning Research) is a prime example that this need not be the case.