Uber CEO Dara Khosrowshahi on turning around a verb

Executive overview

Uber burned more capital before its IPO than any company in history, then lost 85% of mobility volume overnight during the pandemic. Dara Khosrowshahi joined mid-crisis, with a mandate to reach profitability rather than a grand strategy.

The turnaround rested on a supply-led growth model borrowed from Booking.com, merging rides and eats into a single earner and marketplace platform, and eliminating high-vote shares to end the board power struggle. Compounding now makes the structural advantages visible.

The invisible hand really does set ride prices — Uber's job is to wire up every form of transport and reduce take rate as far as possible.

Early career: Barry Diller and the value of unvarnished truth

• Dara joined Allen & Company as an analyst straight from engineering school, chasing a woman to New York City
• Got assigned to Barry Diller's hostile bid for Paramount Pictures against Sumner Redstone
• When his supervisor fell ill, he presented his LBO model directly to Diller
• Diller's pattern: go to the source, skip edited versions of reality
• Herb Allen's philosophy: "bet on people, not companies" — the same loyalty Dara later modelled at Uber
• Advice to young people: don't over-plan, stay open to opportunities, and when you're in — go all in

Buying Expedia right after 9/11

• IAC had signed a term sheet to acquire Expedia before September 11, 2001; a MAC clause allowed them to walk
• Rich Barton called Barry Diller and gave him a clean out — without pressure
• The team debated, then Barry said: "If there isn't travel, there isn't life"
• Called Rich back: game on, no changes to the deal
• Lesson Dara carried forward: in the centre of the storm things look permanent; they revert

What Uber learned from Booking.com

• Expedia was demand-led (brand, air-first); Booking was supply-led (hotels, hotels, hotels)
• Booking targeted fragmented supply — more hotels meant more data, better conversion, cheaper Google traffic
• Uber has 5.6 million drivers and couriers — the most fragmented supply base in transport
• Post-pandemic growth came from rebuilding driver supply first, not demand
• Eats functions as a structural recruitment tool: couriers onboard faster (no car inspection), then get upsold to driving
• As supply improves, surge drops, ETAs fall, prices fall — demand follows

How the rides and eats flywheel compounded

• Once Eats reached scale, all technical and marketplace teams were merged into one
• Rides is the largest customer acquisition channel for Eats — at a quarter of the cost of paid channels
• Cross-promotion required experimentation: wrong surface or frequency hurts the rides experience
• New Eats customers are under 10% of volume in any year; the value is compounding repeat cohorts
• After three years of the machinery running, margins grow faster than competitors because of proprietary traffic
• On supply: Eats lets Uber recruit couriers who eventually transition to driving people

Why Uber prices rose — and why they're now falling

• Blue-collar wage expectations rose across the economy; driver earnings had to keep pace
• This was healthy catch-up: tech and capital wages had pulled far ahead
• Counter-cyclicality: in strong economies, supply is scarce and expensive; in soft economies, more drivers join, prices fall
• Q1 2023: trip growth accelerated to 24% vs 19% in Q4 — unusual at Uber's scale
• Uber's top economist argues the marketplace sets its own spot price; Uber does not control it
• Marketplace matching and pricing improve ~5% per year through ML — compounding on a $120–130B run rate

The CEO recruitment and shareholder base reset

• Daniel Ek gave the headhunter Dara's number and told him: "Since when is life about having fun? It's about impact"
• Dara told the board: the nanosecond my name appears in the press, I'm out — the secret held for two months
• Barry Diller reacted badly, then called back: "Speaking as chairman it's a mistake; speaking as a friend, how can I help?" — and edited Dara's board presentation
• After Dara joined, Travis Kalanick sold roughly 15% of the company at IPO lock-up expiration; stock fell, team panicked
• Over several years, Uber's entire shareholder base turned over — from growth and hedge funds to fundamental long-only holders like Fidelity, Capital, and Morgan Stanley

Eliminating high-vote shares: an 80-billion-dollar prisoner's dilemma

• SoftBank wanted to invest; Masa hinted he'd back Lyft instead if Uber didn't comply
• Founders and early investors held high-vote shares; selling would flip votes to low-vote — a game of chicken
• The solution: get every high-vote shareholder to agree simultaneously to convert to low votes
• If anyone defected, the whole structure collapsed — and SoftBank's $10–15B could have gone to Lyft
• Once done, no one could control the company; the power struggle ended
• Secondary benefit: focus shifted from "who controls Uber" to "how do we build a great company"

Competitive landscape: DoorDash, Lyft, and taxis

• DoorDash bet on suburbs and selection; Uber bet on urban, cheap, and fast — a bias that persisted through 2018–19
• Suburbs have larger families, higher food delivery demand; Uber under-indexed there
• Urban Eats share vs DoorDash is now "quite constructive"; suburbs remain DoorDash's profit pool
• Lyft: stronger than people give it credit for; new CEO is making moves; never count them out
• Taxis: 4.5 million globally — Uber is now wiring them up via blast dispatch (one-to-many offer) rather than one-to-one match
• Freight, grocery, alcohol, health transport, two- and three-wheelers: all segments in the "wire everything that moves" strategy

The case for lower take rates

• Expedia's take rate was 25%; Booking's was 15%; over 13 years Dara took Expedia from 25% to ~17%
• Those percentage points are pure margin — no value created, just extraction
• Uber's gross bookings grew ~22% last quarter; driver and courier earnings (including tips) grew 30%
• The goal: operate at the lowest sustainable take rate by eliminating fraud, automating, and improving efficiency
• High take rates are dangerous long-term — they cap ecosystem growth and invite competition
• "Pigs get fat, hogs get slaughtered"

Self-driving cars and the limits of prediction

• The last 2% of edge cases is unknowable — no one can say when full autonomy arrives
• The real question is what safety level society will accept: 10x safer still means ~4,000 US deaths per year
• After the fatal Phoenix incident, Uber redesigned its safety-first development process, then exited self-driving during the pandemic
• Uber's core skill is demand-supply matching at scale, not AV hardware
• Waymo and Aurora are now partners — a hybrid model where Uber's matching layer routes to either human or robot driver
• 98% fulfillment rate is the bar; any single AV provider will struggle to cover all origins and destinations alone

Building for earners — the cultural shift

• For most of Uber's history, the driver app was built like a consumer product: A/B tested, optimised for throughput
• A driver spending 4–6 hours a day on the platform hits a p95 experience multiple times a week — not once in a while
• Uber is the largest single source of work in the world: 5.6 million active earners, more than the largest employers
• The new framing: higher duty of care, slower development cycles, more humility, direct listening
• If Uber builds the best platform for earners, it wins structurally — lower churn, better service, better liquidity
• The goal Dara wants to be proud of: a tech company genuinely connected to the broad working population, not just elite consumers