Case study — Travel · 10 min read · 5 questions

How Uber made its losses look like profits

By Yishan Lin · Aired January 2023 · Published November 2026

The thesis

Through 2022, Uber had raised over $25 billion and never turned a profit under standard accounting. On paper it came close more than once. In 2018 it reported net income of $997 million, but that was a $3 billion operating loss plus the paper value of stakes it received for leaving Russia and Southeast Asia: 38% of Yandex's taxi business and 23% of Grab. It did the same in 2016 with DiDi and in 2021 with Aurora and Zomato, whose shares later fell apart once they had to trade in public.

The operating business never closed the gap. Uber was built on subsidies: rides priced below cost and drivers paid above it, until local rivals ran out of money. The bet was that at enough scale the discounts could be switched off. But in 2019, with $49 billion of rides booked, Uber needed to keep 43.5% of every fare to break even, double what it took, while drivers were protesting for a 10% cap. Riders switch apps over a few dollars and drivers wait out the surge. Neither side is captive.

So Uber went looking for volume in food, groceries, alcohol and a super app. When pressure to show a profit arrived, it landed on the drivers. From 2016 to 2021 they kept 79 to 80% of each fare. In 2022 Uber's cut jumped from 19% to nearly 30%, and drivers earned 8% less while fares hit records. A business that needs subsidies to grow and tips to pay drivers a living wage has only moved its costs onto someone else. Value creation as a product of unsustainable subsidies and artificial discounting is not disruption. It's fraud.

How do you think about this? 5 strategy questions this case raises and does not answer.
Answer them below

Operating income, net income, adjusted EBITDA, gross bookings, revenue, take rates, price per trip and the value of stakes received in divestitures are Uber's figures as stated in the episode; share prices of DiDi, Aurora, Grab and Zomato are as of the episode, January 2023; per-trip earnings and the 2022 take rate are the episode's calculations from Uber's reported figures

Key takeaways

The model is a platform: software matching people who want an offline service with people who provide it, taking a cut of each transaction. With no buildings or cars to own, margins were supposed to look like software's. Airbnb, Instacart, DoorDash and Uber all run on it. From 2013 to 2018 startups copied it for flowers, laundry, valet parking, massages and cookies, and many of them are gone.

The growth strategy was subsidy. Raise a war chest, sell below cost, take customers from rivals who cannot match the price, and raise prices once they are gone. Uber's early drivers made thousands a week while passengers paid $20 for a 30-minute ride. Without billions in subsidies it would not have its market share, and the subsidies have shrunk since but never stopped.

That makes Uber a test of the whole idea. It raised over $25 billion and had no profit to show after 13 years. The only comparable case is WeWork, which raised $20 billion as a tech company and became a penny stock once it was seen as subsidized subleasing. If the best-funded startup of the biggest bull market cannot make money, the model behind DoorDash and the rest is in question too.

Uber's bottom line hides a recurring trick. In 2016 it lost a subsidy war in China and handed its business there to DiDi for 52 million DiDi shares, priced at $115 in a private round: about $6 billion on paper. The shares could not be sold before an IPO, yet Uber booked $2.8 billion as income, cutting a $3.2 billion operating loss to a $300 million net loss.

It did it again in 2018 after retreating from Russia and Southeast Asia. Merging with Yandex's taxi business left Uber with 38% of the combined company, worth $954 million at the time, and selling to Grab brought a 23% stake worth $2.2 billion. Together they turned a $3 billion operating loss into $997 million of net income, the only profitable year in Uber's history to that point.

In 2021 came $154 million of Zomato stock for leaving food delivery in India, $77 million in a German startup's shares for its European freight business, and a 26% stake in Aurora, worth $1.6 billion, for its self-driving unit. A $3.8 billion operating loss became a $496 million net loss. Public markets then repriced the stakes: Aurora fell from $10 to about $1, DiDi delisted within a year, and Grab traded near $3.

Uber's preferred yardstick is adjusted EBITDA, a measure it defines itself by leaving out 13 kinds of expense. A $5.2 billion quarterly net loss in 2019 became a $650 million adjusted loss; a $1.8 billion quarterly loss in 2020 became $837 million. In 2022 it presented a $9.7 billion net loss as $1.05 billion of positive adjusted EBITDA. That is legal, and misleading. Airbnb and DoorDash use the same device.

The underlying business is a middleman's: grow transactions and the commission on each. Uber is harder than most because it sets the fare itself, where restaurants price DoorDash orders and hosts price Airbnb stays. Charge too much and riders go back to taxis; keep too much and drivers leave. It reports rides and deliveries as one blended trip count, and keeps its driver count to itself.

Travis Kalanick ran Uber from 2010 to 2016 for growth at any cost, launching even where it was illegal. Dara Khosrowshahi, CEO from 2017, made peace with regulators, added safety tools and rebuilt trust with drivers, but kept the path to profit. By 2019 Uber ran 14 million trips daily across 700 cities, with pickups averaging 5 minutes, and still said its strategy was to build "the largest network in each market."

The plan assumed riders are not price-sensitive. Any airport says otherwise: crowds waiting for surge pricing to drop, flicking between Uber and Lyft for the cheaper ride. Experienced drivers do the same, logging on only when hotspots light up. And if low prices were needed to grow the network, it is hard to see why the network would hold once they end.

The 2019 numbers show the problem. Rides bookings had reached $49 billion, up from $18 billion in 2016. Uber kept $10 billion, a 21.5% take rate, up from under 20% in 2016, and paid $39 billion to drivers. With $21 billion of expenses, breaking even needed a 43.5% take rate, as drivers demanded a 10% cap. The alternative was doubling bookings to about $100 billion, by doubling either customers or prices.

Uber tried to buy loyalty on both sides. Uber Pro gave drivers gas discounts and free courses, and Hertz rented cars to those without one. Uber Rewards reached 27% of customers, 25 million members twice as likely to use both rides and Eats. Uber Comfort and a corporate sales team chased bigger fares, and six countries, Japan and Germany among them, were declared must-win.

The real bet was Uber Eats: bookings of $3 billion in 2017 had reached $14 billion by 2019, fed by subsidies and a lower take rate. In India and South Korea the take went negative, so Uber paid to deliver orders, then quit both months later. Jump scooters booked $75 million in 2019 against $160 million of costs and went to Lime at a loss, and self-driving took $1 billion from Toyota and SoftBank before being sold.

The pandemic cut rides bookings to $26 billion in 2020 while Eats jumped to $30 billion, and Uber dropped its Eats take to a record-low 13% to win share from DoorDash. Then came the super app: Postmates, Drizly, rickshaws and taxis in the app, and Uber Pass at $24.99 a month. In 2021 Eats hit $52 billion at a 16% take, and rides recovered to $36 billion only after their take fell below 20% to lure drivers back.

Before the IPO, fares quietly fell. Average price per trip dropped more than 10%, from $10.58 in 2016 to $9.18 in 2017, and growth exploded: bookings up 78%, trips more than doubled, monthly users up over 50%. Prices stayed near $9.50 through 2019. Drivers protesting predatory commissions in those years had the cause wrong. From 2016 to 2021 they kept 79 to 80% of every fare. The fares themselves had been cut.

Then the squeeze moved to drivers. Price per trip climbed to $11.52 in 2020, $14.20 in 2021 and $15.28 in 2022, yet in 2022 drivers earned 8% less. Uber's cut per trip rose 53%, from $2.74 to $4.20, and its take rate from 19% to nearly 30%. On one of the host's own rides, an hour in traffic to LAX, the fare was $80 and the driver made under $40.

Discussion

Question 1 of 5
Uber booked shares in DiDi, Grab and Aurora as income before it could sell them. Should private shares count toward profit at all, and what would you require before they did?
Question 2 of 5
In 2019 Uber needed a 43.5% take rate to break even. Which would have been hardest: doubling its cut, doubling prices, or doubling volume? Why?
Question 3 of 5
Is subsidizing rides different from a supermarket selling milk below cost? Where is the line between competitive pricing and buying a market?
Question 4 of 5
In 2022 drivers earned 8% less while fares hit records. How long can a platform squeeze one side before supply breaks, and what numbers would warn you first?
Question 5 of 5
The super app needs a rider to order food, groceries and alcohol through the same app. What would get you to use one app for all of it, and what would make you switch?

Common questions

Why did Uber report a profit in 2018?

Not because of its operations, which lost about $3 billion that year. Uber left Russia and Southeast Asia in 2018 in exchange for shares in its rivals: 38% of a merged business with Yandex's taxi arm, worth $954 million, and 23% of Grab, worth $2.2 billion. Counting those stakes as income turned the operating loss into $997 million of net income. Uber did the same in 2016 with DiDi and in 2021 with Aurora, Zomato and a freight startup.

What is Uber's take rate?

The share of each fare Uber keeps as the middleman; the rest goes to the driver, and tips pass through in full. It was below 20% in 2016 and 21.5% in 2019, when Uber kept $10 billion of $49 billion in rides bookings. To break even that year it would have needed 43.5%. It dipped below 20% again in 2021 to lure drivers back, then jumped to nearly 30% in 2022.

What is adjusted EBITDA, and why does Uber use it?

It is a measure Uber defines for itself, outside standard accounting rules, that starts from net income and leaves out 13 categories of expense. It makes losses look much smaller: a $5.2 billion quarterly net loss in 2019 became a $650 million adjusted loss, and in 2022 a $9.7 billion net loss became $1.05 billion of positive adjusted EBITDA. It is legal, and Airbnb and DoorDash use similar measures, but it is not profit.

Why did Uber leave China?

It lost a subsidy war with DiDi, in which both companies spent heavily on discounts to win riders and drivers. In 2016 Uber handed DiDi its Chinese business and agreed to stay out of the country for seven years, in exchange for 52 million DiDi shares valued at $115 each in a private round, about $6 billion. Uber booked $2.8 billion of that as income. DiDi went public in 2021, delisted within a year, and by early 2023 traded over the counter.

Do Uber drivers make less than they used to?

Yes, for two different reasons. From 2016 to 2021 drivers kept 79 to 80% of every fare, but the average fare fell more than 10% in 2017, most likely through subsidies to boost growth before the IPO, so drivers earned less even though the commission barely changed. In 2022 fares hit record highs, but Uber raised its take rate from 19% to nearly 30%, and drivers earned 8% less while Uber's cut per trip rose 53%.

Why did Uber get into food delivery?

Because rides alone could not produce the volume Uber needed to break even. Uber Eats booked $3 billion in 2017, $14 billion two years later, then $30 billion in 2020 and $52 billion in 2021. Uber kept that growth going by taking less: its Eats take rate fell to a record-low 13% in 2020 and only recovered to 16% in 2021, and in India and South Korea it went negative before Uber pulled out of both.

What is Uber's super app strategy?

The plan to make Uber the app for everything delivered or driven: rides, restaurant food, groceries, alcohol, convenience items and retail, with a $24.99 monthly membership, Uber Pass, tying them together. Each service is meant to feed the others, and drivers are meant to earn more by switching between rides and deliveries. Uber bought Postmates, Drizly and Cornershop to fill it out. By 2021, 25% of its drivers did both rides and deliveries.

Why doesn't Uber just raise prices?

Because its customers are more price-sensitive than its strategy assumed. At a 21.5% take rate in 2019, breaking even would have meant roughly doubling rides bookings from $49 billion to about $100 billion, by doubling either customers or prices. Riders already wait out surge pricing and compare Uber with Lyft, and if Uber becomes pricier than a taxi, demand falls. Raising its own cut instead risks losing drivers.

Watch on YouTube