You're missing the point. EC2 specifically didn't work for them because they were constantly spinning up new instances and killing old ones, so they were repeatedly incurring the minimum one hour cost. If your business model doesn't call for that sort of a workflow, the article is irrelevant to you.