When he puts it like that, it makes me think Uber has a much better business model than I thought. Taxi efficiency is very heavily dependent on information - if you know exactly where the customers are, where the drivers are, and then can run shortest-path algorithms in real-time to match them up, you can save huge amounts of driver time waiting and finding customers. Figure you save an hour a day * 200 workdays/year * 200K taxi drivers in the U.S. and that's 40 million hours saved in the U.S. a year. At a rate of $10/hour, that's $400M/year. Then add in economies of scale from knowing where all the drivers and customers are, and a better brand and better customer experience from reducing wait times and transaction costs to get a taxi.

Is that worth $80B? I don't know, but I'd guess probably not. But it'd easily pay for a SF HQ and several thousand engineers.