Pairs with the Flywheel Designer Canvas — a ready-to-use strategy tool. Included with a subscription, or $1.99.
Buried in the registration statement Uber filed to go public is a small diagram its own lawyers signed off on: more drivers, more rides per hour, higher earnings, more riders, lower wait times, lower fares - and back to more drivers.1 It is a flywheel, drawn in the cleanest possible terms by the company that lives inside it. Each turn is meant to make the next turn cheaper and faster, until the loop spins so fast no competitor can catch the wheel. It is one of the most elegant growth diagrams in modern business. It is also, in the very same filing, hedged into oblivion.
The official story is that Uber's flywheel built a winner-take-all global moat. The real story is stranger and more honest: the loop is real, it genuinely spins, and Uber wrote down in black and white that it might never matter. The flywheel works. It just doesn't travel.
“Those network effects may not result in competitive advantages or may be overcome by smaller competitors.”1
The loop is real - and it lives inside one city
Start with what's true, because it's genuinely impressive. The liquidity flywheel works exactly as drawn: pack a city with drivers and your wait time drops; drop the wait time and more riders open the app; more riders mean more fares per driver-hour, which pulls in more drivers.1 At the global level the numbers are staggering - $162.773 billion in gross bookings and 11.273 billion trips in 2024.2 But here is the catch the diagram hides: every one of those turns happens inside a single metro. A dense, fast-spinning wheel in Chicago does nothing for a rider standing on a curb in Jakarta. The network effect is a local phenomenon wearing a global costume. MIT Sloan put it plainly: 'ride-hailing will always be an intensely competitive business in large, local markets,' and Uber's resilience rests on 'a relentless aggressiveness rather than a structural tendency toward a global winner-take-all equilibrium.'5
This is the whole game, and it's why the flywheel metaphor misleads. A real flywheel stores energy - each push leaves the wheel spinning a little longer on its own. Uber's flywheel resets at every city limit. The energy you spent achieving liquidity in one market is non-transferable; you have to spend it again, from zero, in the next. The proof is in the obituaries: Uber sold its Southeast Asia operation to Grab in 2018, a rival that had launched in 2012 with forty drivers in Malaysia.6 Uber arrived with more scale, more capital, and a head start, and lost anyway - because none of that scale was local where it needed to be.
| The S-1 diagram suggests | What the record shows | |
|---|---|---|
| Scope of the network effect | Global, compounding | City-local, resets per market |
| Switching costs | High - users lock in | Low - drivers and riders multi-home |
| Energy stored per turn | Compounds across markets | Spent again from zero in each city |
| Outcome in contested markets | Winner-take-all | Sold Southeast Asia to a local rival |
The wheel leaks at both ends
Even inside a single city, the flywheel has a slow puncture. A moat made of network effects only holds if the two sides get stuck to you - if leaving is painful. On Uber, leaving is a tap. Drivers run Uber and Lyft on the same dashboard, switching to whichever app is paying more that hour; riders price-check two apps before they ride. The cost of multi-homing, as one platform analysis documents, is 'relatively low' on both sides, so neither drivers nor riders 'allow any single platform to develop a long-term competitive advantage.'7 Picture two phones suction-cupped to the same windshield, both pinging - that is what the moat actually looks like from the driver's seat. The liquidity Uber spends to win never fully sets, because the moment a rival offers a better fare or a fatter bonus, both sides drift over and drift back.
And then there's the financial tell. If the flywheel were truly compounding into an unassailable moat, you'd expect it to throw off cash long before it became a household word. It didn't. Uber posted a GAAP net loss every year through 2022 - $9.141 billion in that final losing year alone - and recorded its first full-year profit only in 2023, at $1.887 billion.23 Even that number deserves an asterisk: Uber's own earnings release notes the result included roughly a $1 billion pre-tax tailwind from unrealized gains on revaluing its equity investments - paper gains, not rides.4 The 2024 figure looks heroic at $9.856 billion2, but Adjusted EBITDA - the cleaner read on the actual operating engine - was $6.484 billion in 2024 versus $4.052 billion in 2023.23 The flywheel didn't suddenly start minting money. The investment portfolio got re-marked.
Uber maximizes the first term - density - brilliantly. But the second term is near zero because drivers and riders multi-home7, and the third is near zero because liquidity won't cross a city line.5 So the product the moat math depends on collapses, while the cost of conquering each new market stays high. That's the structural reason it took until 2023 to reach a GAAP profit2 - and why even that leaned on non-operating gains.4
The fair objection: it won anyway, didn't it?
The honest counter is hard to dismiss: Uber is enormous, it's now profitable, and it dominates its home market. If the flywheel is so weak, how did it get here? The answer is that Uber didn't win on the flywheel - it won on the bundle. A second loop has quietly become the real thesis. A rider who downloads Uber for a ride sees Uber Eats in the same app; an Eats user starts taking rides because the app is already there; an Uber One subscriber uses both more to justify the fee - and Uber One reportedly carries around 36 million subscribers who lose tangible benefits if they leave.8 This second wheel does what the first one couldn't: it raises switching costs and spans products, not cities. The durable advantage was never the liquidity network effect Uber drew in its S-1. It's the bundling of Mobility, Delivery, Ads, and a subscription into one app you'd actually feel the pain of deleting. The flywheel was a great story; the bundle is the business.
A flywheel only earns the name if each turn leaves the wheel spinning on its own - if effort accumulates. Before you believe a platform's loop diagram, ask one question: does the energy from winning one market carry into the next, or does it reset at the border? For software with global network effects, it carries. For anything whose product is physical - rides, food, rooms - liquidity is local and the wheel resets in every city, so you pay full freight to win each one. The tell is the cash: a truly compounding flywheel funds its own next turn. A resetting one burns capital for a decade and calls the survival a moat. When the loop won't travel, stop selling the flywheel and start building switching costs the loop can't - which, for Uber, meant a bundle, not a network.
Uber's flywheel is one of the best-drawn diagrams in the history of the S-1, and most of what it claims is true. It spins, it lowers wait times, it pulls in drivers - city by city by city. What it never did was store energy across the borders it crossed, which is why China went to Didi, Southeast Asia went to Grab, and profitability waited fourteen years and arrived leaning on a portfolio re-mark. The lesson isn't that flywheels don't work. It's that a flywheel which resets at every city limit isn't a moat - it's a treadmill that happens to be shaped like a wheel. Uber's real moat, when it finally arrived, wasn't the loop everyone admired. It was the one app you'd hate to delete.
Flywheel Designer Canvas
A one-page canvas for mapping a business's flywheel: the reinforcing loop, how it was started, the second-order loops it spins off, the moat it creates, and how it could spin backward. Use it to diagnose whether you have a real flywheel or a funnel drawn in a circle — and to design one of your own.
Included with any subscription, or unlock this tool for $1.99. Get it → · See plans →
Sources
Where this comes from — the filings, records, and reporting behind it.
- 1Uber's S-1 explicitly describes its 'Liquidity Network Effect' flywheel diagram: More Drivers → More Rides Per Hour and Higher Earnings → More Riders → Lower Wait Times and Fares → More Liquidity. The filing also warns: 'those network effects may not result in competitive advantages or may be overcome by smaller competitors.'
- 2Uber FY2023 10-K: Revenue $37.281B, GAAP net income $1.887B (vs. net loss $9.141B in 2022), Gross Bookings $137.865B, Trips 9.448B, MAPCs 150M. FY2024 10-K: Revenue $43.978B, GAAP net income $9.856B, Gross Bookings $162.773B, Trips 11.273B.
- 3Uber FY2023 10-K (cross-check): Gross Bookings $137.865B (+19% YoY), Revenue $37.281B (+17% YoY), Adjusted EBITDA $4.052B (+137% YoY), net income $1.887B vs. net loss $9.141B in 2022. Mobility Gross Bookings grew 32% YoY constant currency.
- 4Uber's Q4 2023 earnings press release states FY2023 net income of $1.4B in Q4 included 'a $1.0 billion net tailwind (pre-tax) primarily due to net unrealized gains related to the revaluation of Uber's equity investments'—documenting that reported GAAP profit was materially boosted by non-cash investment gains.
- 5Uber's network effects are city-local, not global. MIT Sloan Management Review states: 'in many international markets, Uber is the insurgent, and the network effects it enjoys in the United States provide limited advantage' and 'ride-hailing will always be an intensely competitive business in large, local markets.' The resilience of Uber's position depends on 'a relentless aggressiveness rather than a structural tendency toward a global winner-take-all equilibrium.'
- 6HBR documents Uber's exit from Southeast Asia: 'On March 25 mighty Uber bowed out of Southeast Asia by selling its operation in several countries to local rival Grab.' Grab launched in 2012 with 40 drivers in Malaysia and defeated Uber despite Uber's prior U.S. scale, demonstrating that platforms whose product is physical must localize and cannot transfer global network effects.
- 7Multi-homing on both sides of Uber's marketplace undermines the flywheel's moat: 'multihoming costs for drivers to co-exist on Uber and Lyft are relatively low. Many drivers participate on both platforms. Given the ease of booking rides, multi-homing costs are very low for travelers/riders on these platforms as well.' Producers can 'easily switch between platforms without allowing any single platform to develop a long-term competitive advantage.'
- 8Uber's cross-platform bundling strategy (Mobility + Delivery + Uber One subscription) creates a secondary flywheel: 'A rider who downloads Uber for transportation sees Uber Eats in the same interface. An Uber Eats user starts taking rides because the app is already there. An Uber One member uses both more often to get value from the subscription fee.' Uber One had 36 million subscribers as of reporting, with members losing tangible benefits if they switch platforms.