Whatever Happened to the Uber Bezzle?

How Uber outlived its own autopsy

In August 2021, the consensus was that Uber was finished.

Critics argued that the company was a ‘bezzle’. It looked like a company, but it was actually a long con. Its putative core business, selling rides, was dressed up with regulatory arbitrage and accounting gimmicks to suggest future profitability, in order to separate credulous investors from their money now. According to the critics, at some point soon the investors would realize they’d been taken, dump their shares, and the whole enterprise would collapse.

Four years later, Uber is still here… and not ‘burning investor cash while promising future profits’ still here, but ‘making real money’ still here. As of last month, late 2025, Uber was generating billions of dollars in trailing twelve-month free cash flow. It didn’t have a fluke quarter, and it didn’t pull an accounting trick, like taking gains on paper from asset revaluations. Uber has achieved sustained operating profitability across its core businesses for ten consecutive quarters.

The consensus was wrong. Why?

The 2021 critics were not cranks (well, the best of them weren’t). Their arguments were detailed, numerate, and grounded in Uber’s own filings. They had a strong and careful argument that ride-hailing was structurally incapable of producing durable profits in competitive markets. And yet the company did not collapse.

This is a puzzle. Let’s solve it.

The Analysis

The serious work began with Hubert Horan, a transport consultant with decades of experience analyzing airline and transit economics. Starting in 2016, Horan published a multi-part series titled Can Uber Ever Deliver? on Naked Capitalism, later expanded into a comprehensive essay for American Affairs in 2019.

Horan’s thesis was that Uber possessed no structural cost advantage over traditional taxis. The technology that Uber deployed—GPS dispatch, smartphone apps, cashless payment—was genuinely useful, but it was also replicable. Any taxi company could adopt the same tools, and many had. The app wasn’t a moat; it was a commodity. What Uber offered was lower prices, but those prices were kept artificially low through investor subsidies. Passengers were paying less than the true cost of their rides, with venture capital covering the difference.

And Horan claimed that what was true on the customer side was also true on the driver side. Horan documented how driver incentives further inflated costs beyond what the fare structure could support. Uber paid bonuses, guaranteed minimum hourly earnings, and offered surge pricing that flowed disproportionately to drivers rather than to the platform. These were necessary to attract and retain drivers in a competitive labour market, but they meant the company was hemorrhaging cash on both sides of the transaction.

So why were investors giving so much money for Uber to in turn give away to riders and drivers, without any hope of return? According to Horan, Uber was telling everyone it was playing a “winner-take-all” game. Once Uber got big enough, it would have network effects; a rider in Toronto would have the app on their phone and would naturally use it whenever they took a business trip to New York or Los Angeles, rather than researching and using a local competitor. The more riders used it, the more drivers would want to be on the system, and the more drivers there were on the system, the more riders would want to use it (because wait times would be lower than with a smaller competitor). By growing fast, Uber would build out a network, discouraging any rival from emerging. And once that had happened—rivals squeezed out and no new rivals able to appear—Uber would be able to raise fares on riders while cutting driver compensation, becoming very profitable forever.

Horan’s argument was that network effects didn’t apply here. In classic network-effect businesses, the unacceptably high cost of switching locks users in. But riders had both the Uber and Lyft apps on their phones, and drivers were registered on both platforms, and in the heady days of the 2010s, both were switching freely to whoever offered the better deal. There was no lock-in and no accumulating advantage. The subsidies weren’t buying a monopoly but only market share; and that would evaporate the moment the subsidies stopped.

Horan also emphasized that regulatory arbitrage was central to Uber’s economics, not incidental to them. The company’s cost structure depended on avoiding taxi medallion costs, dodging commercial insurance requirements, and classifying drivers as independent contractors rather than employees. Strip away the regulatory advantages and Uber’s costs would rise to meet or exceed those of traditional taxis, without any offsetting efficiency gains to compensate.

The numbers supported this reading. Uber lost $8.5 billion in 2019, the year of its IPO. It lost another $6.8 billion in 2020. By 2023, the company’s accumulated deficit exceeded $30 billion. When Uber reported nominal profits in 2021 and 2022, Horan and others pointed out that these were driven by one-off revaluations of minority equity stakes in other companies (like Didi, Aurora, and Grab); these were paper gains that had nothing to do with whether the core business of moving passengers from point A to point B could generate positive cash flow.

Let’s pause here to make the argument clear. Horan (often amplified by Yves Smith, editor of Naked Capitalism) was making several distinct claims:

1. Subsidy illusion: Uber’s value proposition to consumers, and its growth as a company, were propped up by investor cash, and absent subsidy, the company would only be able to cover its costs by charging fares no one would be willing to pay

2. Regulatory dependency: Uber’s unit economics depended on temporary regulatory arbitrage (classification, insurance, licensing) that would only last until regulators ceased to permit it

3. The impossibility of profit: Uber was operating in competitive markets, and as such competition would reduce their profit to zero

4. Inevitable collapse: Once subsidies ended and regulation tightened, Uber would fail rather than adapt into a durable business

Through 2022, Claims 1 and 2 fit the observable facts well: Uber’s initial growth had indeed been subsidy-driven, and regulatory arbitrage really was central to its operating model. And given what the numbers showed, Claim 3 was reasonable in the context of then-Uber, because the company seemed structurally unable to quit buying market share, and Claim 4 was a conclusion deduced from the preceding three.

Cory Doctorow—science‑fiction author, activist, and incurious take-slinger—took this argument further. Doctorow credited Horan’s work, calling him “the only person you need to read” on Uber’s economics, added one more step. Given Claims 1 through 4, Uber’s ultimate failure is assured, so why is the firm proceeding? Enter Claim 5:

5. Uber is a bezzle: Uber isn’t a firm, but a long con: it takes investor money (as well as a high stock-market valuation) now in return for a promise of future profits that are never coming, and which the firm’s leaders know are never coming

Doctorow’s language is explicit and blunt: pungent examples include “Uber was never going to be profitable. Never”, or Uber “is about to die.” And it included specific predictions.

• August 2021: “Uber’s time is up… Uber is going broke”

• August 2022: “Every bezzle ends. Uber’s days are, therefore, numbered”

• August 2023, after Uber had begun reporting operating profits: “Uber was, is, and always will be a bezzle”

All five claims, taken together, congealed into a meme: Uber is a bezzle.¹

What Changed

The simplest way to say what happened between 2022 and 2025 is not that Uber proved the critics wrong, but that Uber stopped being the same object the critique was describing. The company that achieved profitability in 2023 operates under different constraints, charges different prices, and extracts value differently than the company Horan analyzed in 2016 or Doctorow declared dead in 2021.

Several factors spurred the company to change. The obvious one was the pandemic, which delivered a massive shock to ride-hailing demand, temporarily collapsing trip volumes and forcing the company to confront how expensive its growth-at-any-price strategy was. Less obviously, when the pandemic went away, so too did the cheap capital that had been a feature of the markets from the 2008 global financial crisis onward. But those macro-environment shifts only provided an impetus; they weren’t the changes themselves.

So what were the changes that made Uber profitable?

It did exactly what its investors expected it to do: it charged riders more and paid drivers less.

Starting in late 2020 and accelerating through 2022, Uber enacted substantial fare increases across its markets. Riders in many cities saw typical fares jump by 20 to 50 percent from 2019 levels. By 2022, Uber rides were often more expensive than the traditional taxis they had disrupted. Simultaneously, Uber increased its ‘take rate’—the share of each fare it keeps after paying drivers—from roughly 22 percent before the pandemic to nearly 30 percent by 2024. This was a direct transfer of value from drivers to Uber’s bottom line.

Uber was able to make these changes simultaneously because it shifted from a transparent commission model to algorithmic allocation, where what a passenger pays and what a driver earns are determined separately (and can diverge dramatically). Uber now uses data on rider behaviour to charge each customer something closer to their maximum willingness to pay, while offering drivers the minimum they’ll accept for each trip. This is price discrimination on both sides of the market, enabled by information asymmetry and executed at scale.

The critics had predicted this move, or something like it. What they didn’t predict was that Uber could pull it off without triggering the expected consequences. Horan’s model assumed that raising prices or cutting driver pay would invite one of three responses: riders would defect to cheaper alternatives, drivers would quit for better opportunities, or competitors would undercut Uber’s new margins.

None of these materialized at the scale required to discipline Uber’s pricing.

Riders didn’t flee because they had nowhere to go. Traditional taxi systems, lazy monopolists hollowed out by a decade of stiff competition, couldn’t absorb returning demand even if riders wanted them to. Lyft, Uber’s only significant North American competitor, was in worse financial shape and couldn’t afford to restart the subsidy wars; instead, it mirrored Uber’s price increases. New entrants faced not only the difficulty of building a rival to Uber, hard in a capital-constrained environment, but also the credible threat of Uber retaliation; the market had watched Uber burn billions to crush competitors, and no investors were willing to test if the company was willing to do it again. The result was something closer to a duopoly than a competitive market.

Drivers didn’t quit en masse because they lacked alternatives. The gig model ensures a constant supply of new entrants: people between jobs, those needing supplemental income, those valuing schedule flexibility. Much as in long-haul trucking, driver turnover is high, but aggregate supply seems always to be sufficient to meet demand, as friend of Changing Lanes Abi Olvera has discussed recently. There were moments of strain—early 2021 saw acute driver shortages as demand rebounded faster than supply—but Uber navigated these with targeted, temporary incentives rather than structural pay increases. By 2022, the driver pool had stabilized at the new, lower compensation level.

The painful truth is that Uber’s early subsidies bought something durable after all: not the network effects the bulls had promised, but market power. By proving it would crush any rival, Uber created an effective barrier to entry. By the time it raised prices, there was no one left to undercut it.

The critics’ second major bet—that regulators would eventually force Uber to internalize its true costs—also failed to materialize as predicted, though not for lack of trying.

The crucial battle was California’s Proposition 22 in November 2020. California had passed AB5, a law that would likely require Uber to classify drivers as employees rather than independent contractors, with minimum wage obligations, benefits, and tax contributions. Uber, joined by Lyft and DoorDash, spent over $200 million on a ballot initiative to carve out an exemption, and won. The result preserved Uber’s contractor model in its largest market and sent a clear signal about the company’s political capacity.

That strategy didn’t succeed everywhere. The UK Supreme Court ruled in 2021 that drivers were ‘workers’ entitled to minimum wage and holiday pay. Uber absorbed the costs and raised fares to compensate. These were real expenses, but manageable rather than existential.

The pattern across jurisdictions was similar: regulators extracted concessions, Uber absorbed them, and the business continued. The scenario where labour reclassification would expose Uber’s fundamental unsoundness never arrived.

On the expense side, Uber made cuts that would have been politically difficult before the pandemic forced the company’s hand. In 2020, it laid off thousands of employees and shed over $1 billion in fixed costs. More importantly, it abandoned the moonshot projects that had consumed capital without generating revenue.

The autonomous vehicle unit, Uber ATG, had spent over $2.5 billion developing self-driving technology that never reached commercial viability. Uber sold it to Aurora in late 2020. The flying taxi project went similarly. E-bike and scooter operations were offloaded to Lime. Unprofitable international markets were exited or sold. What remained was a leaner company focused on two core businesses: rides and delivery.

Uber also quietly contracted the scope of its ride service. The pre-pandemic Uber had prided itself on ubiquity: coverage across suburbs and exurbs, service at all hours, fast pickup times everywhere. Much of this was subsidized; rides in low-density areas and off-peak times hemorrhaged money. Post-2020 Uber concentrated on dense urban cores and peak demand, where drivers could achieve higher utilization. The service became more like a traditional taxi fleet: profitable routes and times first, everything else second.

Evaluating the Claims

With the mechanisms traced, let’s return to the five claims that comprised the bezzle thesis and assess them against the evidence.

Claim 1: Subsidy Illusion. The claim was that Uber’s growth and value proposition were propped up by investor cash, and that absent subsidy, the company could only cover costs by charging fares no one would pay.

The subsidy was real: over $30 billion in accumulated losses funded below-cost rides for a decade. That investment did what it was supposed to do, namely digging a moat. Uber used the subsidy period to destroy the competitive alternatives that would have disciplined its pricing, and riders proved willing to pay the higher fares the company then decided to charge.

Let’s call this one correct in its prediction, but false in its diagnosis.

Claim 2: Regulatory Dependency. The claim was that Uber’s unit economics depended on temporary regulatory arbitrage—contractor classification, insurance avoidance, licensing circumvention—that would only last until regulators closed the loopholes.

Uber’s model does depend on the contractor classification; internalizing full employment costs would substantially erode the firm’s margins. But that classification does not appear to be temporary. Uber proved capable of reshaping the regulatory environment through lobbying, ballot sponsorship, and when all else fails just absorbing the costs.

Let’s call this one false as well. The arbitrage turned out to be more durable than critics expected, in part because Uber actively defended it.

Claim 3: Impossibility of Profit. The claim was that Uber, operating in competitive markets, would see profits competed away to zero because it possessed no structural cost advantage.

The key word is competitive. Uber secured market power sufficient to escape discipline. The critics were right that Uber had no cost advantage, but wrong that this precluded profit; in an uncompetitive market, pricing power is all you need, and Uber acquired it. So this claim was conditionally correct, but was predicated on an assumption that proved incorrect.

Claim 4: Inevitable Collapse. The claim was that once subsidies ended and regulation tightened, Uber would fail rather than adapt.

To the contrary, Uber adapted. The company that exists in 2025 is not the company the critics analyzed: it charges different prices, operates in a different market, and runs a different cost structure. The prediction of collapse treated the firm as static when it was dynamic. This was the core analytical error: assuming that because early Uber couldn’t be profitable, no future version of Uber could be either. This claim has proven to be false.

Claim 5: Uber Is a Bezzle. The claim was that given the foregoing, no honest person could believe that Uber would ever succeed, so the firm’s leaders must be confidence tricksters, extracting money now in the hopes of getting out before the inevitable collapse. And yet the collapse has proven evitable. So this claim is false as well.

That means that all five claims in the thesis are incorrect or false.

I’ll give them their due: Uber’s critics correctly identified that early Uber ran on subsidized growth disconnected from operational fundamentals, and correctly predicted that reaching profitability would require raising prices and squeezing drivers. They incorrectly predicted that Uber would prove unable to do this.

If Uber’s operating profits prove unsustainable over the next two to three years—if they depend on one-time cost cuts or a favourable competitive moment that passes—the bezzle thesis regains force. A recession that collapses demand, or a resurgent competitor willing to restart the subsidy wars, would test whether Uber’s profitability is structural or circumstantial.

Conversely, if Uber maintains profitability through an economic downturn or competitive challenge, the case for genuine transformation solidifies. The longer the profits persist, the harder it becomes to argue that they’re illusory.

Image: “Ponzi scheme” by Stock Catalog (CC BY 2.0, via Flickr)

[Yes, that’s really the image title…]

I highlighted Doctorow’s 2021 post because it represents the moment the bezzle meme reached peak influence: confident, widely shared, and, in retrospect, timed almost precisely as the underlying facts were beginning to shift.

The failure mode here is not bad analysis, but static thinking. We are good at analyzing snapshots and bad at tracking drift. We are good at conditional models and bad at noticing when their premises expire.

So what’s the lesson? Not that the critics were foolish. Horan’s analysis of Uber’s unit economics was rigorous and, on its own terms, correct. Uber really did have no cost advantage over incumbent taxis. It really was burning investor cash to subsidize below-cost rides. The critics predicted that profitability would require raising fares and squeezing drivers, and that’s exactly what happened.

The error was treating the competitive and regulatory environment as fixed, and thus that market discipline and legal constraints would eventually bite. They didn’t, because Uber spent the subsidy years making sure they wouldn’t.

I’m sanguine about this outcome. The taxi cartels with their medallions were extractive; they enriched speculators while limiting ride supply and delivering indifferent service. Uber modernized the sector, expanded access, improved reliability, and created flexible work that millions of drivers have chosen over alternatives. Against this, Uber and other ridehail firms have increased urban congestion and depressed transit ridership; but the solution to this isn’t demonizing Uber, it’s implementing a congestion charge, which we largely have not done. We’ve got no one to blame for that but ourselves.

Meanwhile, the sector is changing again. Waymo alone now provides over 450,000 robotaxi rides per week, and Zoox and Tesla are ramping up their service. Uber, having sold its own self-driving unit in 2020, has pivoted to partnerships: Waymo vehicles are now available through the Uber app in Phoenix, Austin, and Atlanta, with more cities coming. The economics aren’t there yet—Uber’s CFO conceded in August 2025 that “AVs today are not profitable”—but the trajectory is unmistakable. It may soon be the case that we think Uber walked so Waymo could run.

1. The economist J.K. Galbraith coined this term, defined as “the magic interval when a confidence trickster knows he has the money he has appropriated but the victim does not yet understand that he has lost it”.