Sam Altman argues in How to Grow Huge that the only way for a startup to grow really large is to create products that people love and promote. As the user base grows, users attract ever larger numbers of users to the product, producing compounding growth.
The point is a terrific one and I think it can be generalized. To grow really large, startups have to create proprietary distribution channels. The one Sam champions, word of mouth marketing intrinsic to a strong brand, is one example of a proprietary distribution channel.
In each of these cases, the startup owns the distribution channel. Sometimes the channel is integrated in the product (Dropbox), sometimes the channel is adjacent (HubSpot, Cyanogen). In either case, no competitor can touch it. Challengers may try to replicate the success of the distribution channel by copying it, but cannot compete within the channel, which means it's an incredibly cost-effective channel for its owner.
Contrast these proprietary distribution channels with hyper-efficient and competitive online ad markets both SEM and display, SEO rankings, and open platforms like the iOS and Android and Salesforce and Intuit and other online app stores. These are open markets with lots of competition which enforce efficient pricing and reduce margins. It's a boxing match with every entrant taking their fair share of punches.
Sometimes startups begin in the open market, and having achieved some scale, foster their own channels. Zynga and Kabam grew to some scale on Facebook's traffic hose before both deciding to move to their own platforms. AirBnB drafted on Craigslist's listing velocity to engender enough liquidity of their own. YouTube and MySpace is the same story. Ultimately, each of these success stories built their own user acquisition channels.
Sam is absolutely right. The only way to grow really large is to build and control proprietary distribution channels. It's a huge strategic imperative for startups reaching scale.