Scaling too early is related to over-engineering. Engineers worry about getting to the kind of scale that Facebook and Google have, so want to architect systems for that ultimate scale. This level of scalable engineering is not cheap, and will change the way you write your code, making it a slower process to develop it (and increasing costs and time to market)
If you are learning lessons along the way, the product you start building, and the final product version are likely to be quite different. Heavy duty engineering hinders adaptation, if you have a “throw-away” prototype, you will let it go easily. The second or third re-write is when you get it right.