Why Amazon’s AWS and Google AdWords Fuel the Startup Boom

A current YCombinator company founder who also participated in the summer 2009 session has compared and contrasted his experiences. The most sweeping change: valuations are far higher than they used to be: ” It’s weird to see entrepreneurs compare convertible note caps of 7, 8, 10 million dollars from this lens of low, medium and high. I see those numbers and I just can’t believe that any seed company is receiving a 7 million dollar cap. None of us could get over 3 in 2009.”

In three years, a larger population of companies ends up with double the valuation. It’s hard to explain this through market sentiment alone: even if investors are twice as optimistic about the future, or about technology in particular, that kind of optimism should be reflected in much higher levels for other market indices. It’s startups in particular that have benefited, not technology stocks in general.

Why are startup valuations, in particular, so much higher? Because capital costs are steadily approaching zero, and marginal costs are easier than ever to predict.

YCombinator’s original thesis focused on the first point. Paul Graham notoriously pointed out that, for many consumer web startups, starting a company had about as much overhead as being unemployed. In the early days of YCombinator, starting was cheap and scaling was comparatively hard. But now it’s possible to arbitrarily scale up infrastructure through Amazon’s cloud computing offerings. And it’s easy to arbitrarily scale up marketing through AdWords. So once a startup has gotten to the point that “customer lifetime value” and “customer acquisition cost” are meaningful concepts, it’s entirely plausible to scale up by an order of magnitude or more.

And that means there’s more embedded optionality in the first round of funding. Early stage investors aren’t buying x% of a real business: they’re buying x% of a y% chance at a really amazing business. All else being equal, a higher y% and a higher value for “really amazing” will mean that x can drop without blowing up the deal.

Investors tend not to couch their decisions in terms of explicit optionality, but the standard practice among lots of VCs is to “date” aggressively in the early stages, and then marry cautiously later on. Increasing the information content of dates encourages more dating, and higher demand for dates raises the cost of admission. Even if they’re not explicitly running these numbers, it’s hard for trends to work out any differently: investors often chase “the next [startup of the moment].” If AirBNB had started in 2004, “the next AirBNB” would have been orders of magnitude less exciting than AirBNB is today. And AirBNB is exciting not just because it’s an obviously good product, but because it’s easier to model than previous iterations on that business would have been.

In a sense, this is a strong argument against a bubble. If there’s one positive thing you can say about today’s crop of startups, it’s that investors have better information—in terms of aggregate data and in terms of fast access to company-specific numbers—than they did in previous years. If better information leads to higher prices, that’s a sign that startups were excessively undervalued years ago.