Tldr: Looking purely at valuation there is no evidence Uber created a valuable “new use case”. Instead, it has captured most of the value of the taxi market that it disrupted.

In 2014 I wrote a post in response to discussions of the valuation and potential for Uber. At the time Uber was valued at around $15-20B. Today it is worth $200B (not bad, though you would have done better, and very likely slept better, buying Microsoft, never mind Nvidia, which is extraordinary, but that’s another topic). A post challenged an NYU Stern professor for even considering the taxi market as a proxy for Uber’s opportunity, partly referencing the idea that analyzing the market that way was like trying to size the market for cars by looking at peak horse.

Spoiler alert- my post did just that, and concluded that in fact the analyzing the market for cars by looking at 'peak horse' would have been extraordinarily accurate- the real change was in all that the car enabled – the suburb and, to a greater or lesser extent, the American way of life…

Going back to the original debate it’s interesting to think about the loss in value of US taxi market from its 2013 pre-Uber peak. The NYC Yellow cab market was worth $17B in 2013, based on the price of medallions, and there was a significant limo business, perhaps 20% of the overall, so close to $20B as a value of the NYC taxi market. Headline GDP has doubled since then, so without Uber the market might expect that to have grown to $40B today. If we say NYC is as much as 33% of the US market, which seems high but not crazy, that puts expected cab market value at $120B today. But with Uber the value of these medallions has fallen by ~90% in the markets where they are traded (including NYC), so that’s an implied loss of $108B. Uber’s US and Canda business is about half of its revenue, so if we assume its value ($200B today) is approximately aligned with its revenue, its US business is worth ~$100B (and we’ll ignore Eats/Freight etc for simplicity). Lyft is valued at $8B. So Uber and Lyft’s businesses have, more or less exactly, captured the loss in value ($108B) of the US taxi business. The ride hailing model has not created *any* new market value.

Let’s look at this another way. Uber was positioned as a revolution: “ The Game Changer: Uber as a Car-Ownership Alternative.” If that was true, we would reasonably expect evidence of changes by now, perhaps the obvious one would be a fall in personal car ownership. Auto insurers peg 2022 US car registrations as about 10% above 2014 (283M up from 253M). Yahoo Finance  puts the value of the Auto and Truck Dealership sector as $156B. Not the cars themselves, but the people taking a margin selling them, collectively are worth almost what Uber is worth.

It’s hard to look at car companies themselves because of the meme stock nature of Tesla, but Toyota, the largest company by most measures, is worth $307B today (up from ~$180B in 2014), comfortably more than Uber, with a much lower overall market share.

All this is not to disparage Uber. And not to say that from a consumer point of view things are not better now (but when was that valued by the market?)*. It is to point out that Uber never really created a valuable “new use case,” which was the gist of my question/argument in 2014. Instead it has captured ~all the value of the taxi market. In many apparent market disruptions the incumbents ultimately gain the benefits (shipping and containerization might be one example, telcos and VoIP another), in others the new entrant can win. Uber, a disrupter, has clearly won in its market entry, but the size of the prize was not changed, and winning has (not yet) created any meaningful new value in adjacent areas.

Sept 2025

References and notes

Aaron Levie’s Peak Horse tweet

How to miss by a mile: Bill Gurley. Two key planks of the argument here have not proved out: The price advantage that Gurley quotes here is now often upside down- Uber has become the high price offering in some markets, higher than traditional taxis (e.g. this summer SeaTac to downtown Seattle, LAX to Koreatown and downtown), and, as mentioned above, even with the pandemic, there is no sign that car ownership is falling

A Disruptive Cab Ride to Riches: The Uber Payoff: Professor Aswath Damodaran’s original analysis
In this piece Professor Damodaran does make one assumption I would call out as questionable, by a lot. He asserts the TAM for the global taxi and limo market to be $100B, when at the time the market value of the NYC medallions was $17B, to think that the Yellow Cab side of the NYC market represented 15% of the world market seems aggressive (especially when he asserts Tokyo is the world’s largest taxi market)

* The question of “what about the improved user experience” needs to be set against the recent increases in prices and also the unresolved question of whether being an Uber driver is a good economic decision (when you factor in all the costs of running a car as an Uber, the question: is the effective net driver pay above minimum wage? would be one way of looking at this)

Those of a certain age will probably remember being asked, when pitching a new technology idea, some variation of “What’s your business model [son]?” ChatGPT and its ilk (“Conversational AI” or “AI”) are already seen as existential threats to the existing search players from a user point of view. While history has shown that, in general, with large audiences comes some reasonable monetization, it is worth considering how well-suited Conversational AI is to delivering search-style and levels of monetization. Two issues are discussed here, the type of questions/searches conversational AI is well suited for, and the specificity and display of the results.

 One of the lessons learned in the early mobile search world (where the author worked at Microsoft) was that not all searches are created equal. What we found launching Microsoft’s first mobile search product was that, while we saw respectable headline advertising rates for searches, our overall <revenue per search> was relatively poor. It took some time to work out that revenue per search was a poor metric to analyze. What we needed to do was to look at <revenue per monetizable search> and separately <percentage of monetizable searches>. Many of the searches we were doing a good (to be fair your mileage would vary) job of answering were not ones that anyone wanted to advertise against.

 It's worth thinking about the most directly impacted internet institution, not Google (or Bing) or Reddit, or TikTok, but Wikipedia. Surely nothing on the internet is more likely to be replaced by fluent bullshit conversational AI than that a moderately well sourced “book of everything” written by committee. The purpose of this is not directly to consider that question, but rather to look at the monetization, or lack thereof, around Wikipedia. There isn’t any. Search for <typical English history question that apparently everyone is going to get ChatGPT to answer to destroy history homework>, for example “what was the impact of the Magna Carta” and you don’t see ads, either on the Google page or on the resulting Wikipedia page. There is no money in English history. ChatGPT also does a poor job of delivering on the typical search query of “Best plumber in Des Moines” or “Best sushi restaurant open now in Seattle.” At best it points you to sites to check to look for the actual answer or gives an out-of-date list with no ability to assess the credibility of the ranking. These are not “AI” problems, they are pure data problems, which require access to trusted survey mechanisms and accurate opening hours and locations.

The conversational style of most AI solutions, which again is typically seen as a positive – Look Ma! It produces real sentences! – is also problematic from a search advertising point of view. The accidental genius of search ads is that they both look like search results, but can be clearly labelled, and are not offensive or intrusive. The ad for the hotel in Cancun is next to or above the search result and looks similar. How are you going to deliver that in a conversational result about Cancun as a spring break destination? The dreaded advertorial looms, or worse, a question about bias in the results. Also ChatGPT does not (yet) deliver the very specific results we have come to expect from Google, and as such its general pointers to where and what to buy (e.g. “you can buy that from www.amazon.com” which is a typical answer to a question about buying a very specific item) won’t support very specific search ads. In addition, worth noting that one of the things Google “did wrong” according to conventional, eyeball/portal wisdom early on was that it encouraged people to click away from its site- again this was critical for delivering the targeted audience that advertisers wanted to the right place, the advertiser’s website. To make money advertising, you need to keep your eyeballs, or charge a lot of money as they leave. Can AI do the latter?

Can ChatGPT create sources of revenue? No doubt, but as a consumer site it is unclear how it could deliver the type of targeted, below the line, ROI based advertising that advertisers have come to expect, and that consumers are comfortable with. Could AI enhance search advertising? Of course, and it is already being applied to optimize search advertising by many companies.

To conclude- an “answer” web site can clearly be useful, but the ecosystem of making money around search has relied on a specific set of use cases and interaction modes (in particular clicking away) which will need to change significantly before conversational AI can stand on its own. It seems just as likely that AI will be used to enhance and improve the existing solutions, rather than replacing them.

This is a quick test to see how the site looks

Apple is tying a new (but old) payment method to a new phone and a new OS. While obviously mass market in intent, the rollout means it’s likely to be used initially by tech-savvy early adopters.  These are the same people who have been saving their credit cards with e- and m- commerce merchants for years now (isn’t the magic of Uber the fact you don’t pay?). A quick analysis of my credit card bill shows more than half the charges are stored card-on-file transactions. If my supermarket had its own Starbucks-like app, that would take out another 10-15% of my transactions

The protests about Apple Pay being turned off in drugstores are instructive- there are increasingly fewer places where people actually whip out a card to pay. Opentable’s experiments with eaters paying for meals booked through the system via the Opentable app and a stored card will take another chunk of the “payment opportunities” off the table.

If Apple Pay data is readily available, will one effect be for high frequency retailers to develop their own apps, as Starbucks has done? While it’s hard to imagine an app for every store, it’s not as hard to see a consumer using a combined shopping, loyalty, and payment app for their top-5 most frequent stores (on in Opentable’s case, top scenario). Apple Pay might be the infrastructure behind that, but Braintree/Balanced/Stripe are surely more likely to be the multi-platform plays.

One relatively common activity which does tend to drive purchases at unusual/less frequent places is travel. That’s also a situation where the payment problems can be harder to deal with- you simply don’t know what you’ll get and you can’t just run home. People won’t be heading to the airport relying on their phone as a wallet, in the way Starbucks has meant you can leave your wallet at home when you are running out for coffee.

Cloud commerce, often triggered by a local device, is here to stay; whether it needs to mimic a payment mechanic developed in the 1970s is less clear to me.

Looking at Facebook’s reported data since Q4 2012 it is clear that mobile really is driving growth, and digging in a little further it seems that as users transition to mobile, they tend to become primarily-mobile users, there are very few casually-mobile users of Facebook.

First a look at the PC-only users, people who in a given month only touch Facebook from a PC; these are falling both in absolute and, even more, in relative terms.

Facebook has grown strongly in this period: as these PC-only users have fallen- what has taken up the slack? Mobile-only users have more than replaced the lost PC-only users. On a daily basis, mobile-only users (DAUs or Daily Active Users) overtook other users (i.e. PC-only and people logging in on both platforms) at the end of 2013. 

The trend is the same, but the picture is quite different, on a monthly basis (below). It is very clear that ALL the Facebook growth in the past 7 quarters has come from mobile-only users (MAUs, or Mobile Active Users), but there are apparently a much larger number of other (PC or multi-platform) users.

This relationship looks even stranger if you take one of Facebook’s own important metrics for engagement, the relationship between Monthly and Daily users, i.e. what proportion of monthly users are regular/daily users? As can be seen in the data table below, there are more mobile-only users on a daily basis than there are in a month. How can that be?

It seems clear that many users who are mobile-only on a particular day check in occasionally on a PC in a given month, while still using mobile on that day. But there is very little evidence of predominantly PC-based users occasionally checking in on mobile. If they did the absolute number of PC-only DAUs might be the same or higher than the PC-only MAUs, what we see is that those proportions are almost the same (so the absolute numbers are lower). The charts below show this clearly. Note how PC-only is the same proportion of daily and monthly use, while the mobile users are mostly purely-mobile viewed on a daily basis, but mostly multi-platform on a monthly basis.

This all seems broadly positive for Facebook- mobile monetization is improving, and as people shift away from PC, they are shifting towards mobile-only or mobile-first use.  It does raise the question of the benefits of splitting messaging from the core mobile app experience.  If messaging is a gateway drug for mobile Facebook usage, then you want people to be in the core experience, rather than letting them think that they can get what they want from a messaging app, while sticking with the core wall and status experience on the PC.

Notes-

All data from FB annual and quarterly reports.

One number (mobile only DAUs for Q3 2013) is estimated, as it has not been reported by Facebook.

Tags: fb