← Back to list
Colossus (Invest Like the Best / Business Breakdowns)Podcast3 Jun 2026Source: colossus.comHost: Patrick O'Shaughnessy

Dara Khosrowshahi - Uber's Bet on AVs, AI, and Building a Super-App - [Invest Like the Best, EP.476]

In plain words

Uber CEO Dara Khosrowshahi explains that Uber won't build its own self-driving cars but will act as a service provider for all autonomous vehicle (AV) companies—handling charging stations, insurance, and financing. He believes multiple AV players will win, and Uber wins by aggregating supply and demand. Key mentions: Waymo (partnership boosts trips per vehicle by 30%), Lucid (co-building a $60k-70k car), and Expedia (hotel cashback for Uber One members). AI has made engineers 10x more productive, so Uber will slow hiring.

AI SummaryAI-generated · may contain errors · verify against the original

Uber CEO Dara Khosrowshahi discussed the company's strategy in a podcast, with the core view that Uber will succeed as a "demand aggregator" in the era of autonomous vehicles (AV) and physical AI. He views Uber as a supply-driven company, believes there will be multiple winners in the AV space rathe

~11 min full read · 8 sections
Deep Analysis

At a Glance

Dara Khosrowshahi (Uber CEO, former Expedia CEO for 13 years) systematically articulated Uber's strategic positioning in the era of autonomous vehicles (AV) and physical AI during a podcast. Core thesis: Uber is a supply-driven company; the AV space will have multiple winners rather than a single dominant player. Uber wins by becoming a "demand aggregator" and "AV ecosystem service provider," not by building its own vehicles.


1. Uber Is Fundamentally a Supply-Driven Company

Dara Khosrowshahi believes Uber is the polar opposite of Expedia — supply takes precedence over demand.

  • Mechanism breakdown: At Expedia, consumer demand is acquired first, then matched with hotel/flight inventory; at Uber, drivers must be recruited and restaurants/retailers/grocery stores signed first, with demand following naturally. "Our biggest opportunity and challenge is ensuring we lock down every piece of supply — every car that can take you from A to B, every restaurant, every retailer, every grocery store. If we do that, demand will take care of itself."
  • Data support: When Dara joined, Uber Eats had annual gross bookings of less than $1 billion; it now exceeds $100 billion — requiring massive organic investment to build the supply side.
  • Current penetration: In cities where Uber operates, approximately 40-50% of signable restaurants/merchants are already onboarded, leaving substantial room for growth.

Inference: Uber's competitive moat in the AV era is not technology, but the already established supply network and demand aggregation capability — "We are signing partnerships with over 30 AV partners, including Waymo, Neuro, Lucid, NVIDIA, and others."


2. The AV Era: Uber Doesn’t Drive, It Acts as an “Ecosystem Service Provider”

Dara Khosrowshawi has made it clear that Uber will not build its own AVs but will instead offer a “go-to-market solution” for all AV companies.

  • Historical analogy: This is similar to the online travel agency (OTA) industry—Expedia both competes with Marriott and cooperates with it (by bringing incremental customers). The same dynamic applies to Uber’s relationship with Waymo: there is competition (Waymo has its own brand and channels) as well as cooperation (Uber brings Waymo a more than 30% improvement in vehicle utilization).
  • Data chain: AVs on the Uber network generate 30% more trips and revenue per vehicle per day compared to independently operated AVs—“that 30% has a massive impact on return on investment.”
  • Services Uber provides: Site selection for charging stations/parking lots, collaboration with fleet partners, financing (announced a $1 billion financing facility with Santander for EV and AV fleets), autonomous driving insurance, and real-time traffic data feedback to models.

Extrapolation: Dara believes the AV space will not be dominated by a single player, but will resemble the foundation model space—multiple winners plus open-source solutions coexisting. Uber’s goal is to become the “largest AV supply aggregator,” just as it is already the world’s largest mobility service aggregator.

Falsification condition: If Uber fails to secure sufficient AV supply (i.e., partner with enough AV companies), it may fail.


3. AI Adoption: A Two-Phase Strategy from "Exploration" to "Efficiency"

Dara Khosrowshahi describes Uber's AI strategy as "first encourage exploration, then pursue efficiency."

  • Current Status: Uber is already an AI/machine learning-native company (long using AI to optimize matching, pricing, and routing), but generative AI has brought a qualitative change—model scale is "about 10,000 times" larger than older models.
  • Specific Applications:
  • Predicting user destinations: correct 75% of the time, enabling "one-tap interaction"
  • Universal search: no longer distinguishing between "rides/food/groceries," a single search displays all services
  • Engineer efficiency: code submissions from Indian developers increased 10x, using autonomous agents
  • Cost Reality: Dara revealed that "the AI budget for the entire year was used up in one quarter," forcing him to "control headcount growth"—after improving engineer efficiency, there is no longer a need for new hires at the same scale.
  • Two-Phase Strategy:
  • Exploration Phase: Use frontier models (e.g., OpenAI, Claude) to experiment with new interactions
  • Scaling Phase: Shift to more efficient models (open-source or cheaper token solutions)

Inference: Dara believes AI will fundamentally change Uber's operations—"I push the team to start from first principles and rebuild systems and processes from scratch with AI, rather than just optimizing by 20-30%."


4. Super App Strategy: From Mobility to Full Travel Chain

Dara Khosrowshahi positions Uber as a "local living + travel" super app, with the core lever being the Uber One membership system.

  • Membership data: Uber One has 50 million members, up 50% year-over-year. The membership model draws on Amazon Prime — "In the first year, we lose money on you, but after 2-4 years, you become more profitable."
  • Cross-platform advantage: Approximately 13% of Uber Eats orders come from mobility users (who see the Eats entry point within the mobility app). Uber One members enjoy mobility discounts, free delivery, zero surcharges on groceries, and 10% cashback on hotels.
  • Hotel business logic: Uber completes 1.5 billion trips annually outside users' home cities, with 15% of trips being airport transfers — "The first thing you often do after landing is open Uber." A partnership with Expedia returns most of the economic value to Uber One members.
  • Ideal end state: After a user books a hotel, Uber automatically obtains flight information via email, pre-arranges pickup and drop-off, notifies the front desk upon arrival, or allows the Uber App to serve as a room key — "I want to bring the magic I have in the city to your travel experience."
  • From on-demand to planned: Uber Reserve (scheduled rides) has achieved a $5 billion annual run rate (which did not exist five years ago), proving the brand can extend into "planned" scenarios.

Falsification condition: If users cannot accept Uber extending its brand from "on-demand" to "planned" (e.g., booking a vacation hotel months in advance), the hotel business may fail.


5. Leadership: Core Principles Learned from Barry Diller and Allen & Co

Dara Khosrowshahi shared three key leadership principles.

1. Seek the truth from the source (Barry Diller): Diller insisted on direct conversations with the lowest-level sources of information—"Filtering wears down the edges of a story, and those edges are often your advantage." Dara applies this at Uber—"I look for 'troublemakers' in the company. They are the mutations, the source of the company's evolution."

2. Bet on people, not companies (Herbert Allen): Allen told him, "Companies are sometimes good and sometimes bad, but great people are always great." When Dara left Allen & Co to join Barry Diller's startup, his compensation dropped significantly, but "I bet on a person."

3. Maintain random interactions: Dara deliberately avoids overly structured schedules—"I don't just want to meet with my direct reports. I create random interactions and look for those 'troublemakers.' The larger a company gets, the more people tend to get along, but 'troublemakers' are often pushed out. I need them."

Implication: Dara believes AI will accelerate internal company change by five times—"Companies that adapt to change will win; those that cling to 'we've always done it this way' will struggle."


Mentioned Positions

Position Guest Stance Key Data
Waymo Partner/Coexistence AV utilization rate on Uber's network increased by 30%+
Neuro Partner Collaborating on manufacturing Lucid mid-size vehicles, estimated at $60,000-70,000 per unit
Lucid Partner Same as above
NVIDIA Partner Providing computing, sensors, and software drivers
Wabi Partner Not detailed
Wave Partner Focused on licensing end-to-end models to OEMs
Pony AI Partner Not detailed
Joby Mentioned (drone passenger transport) Under construction
Zip Mentioned (drone delivery) Closely monitored
Santander Financing Partner $1 billion credit facility for EV/AV fleets
Expedia Partner (hotels) Most economic value returned to Uber One members
McDonald's/Starbucks/Chipotle Partner (Uber Eats) Also in a competitive relationship
Reed Hastings (Netflix) CEO to learn from Admired for structured thinking and occasional "gambler" traits

Judgments Worth Remembering

1. "Uber is a supply-driven company, not a demand-driven one" (Dara Khosrowshahi) — The complete opposite of Expedia: first lock in drivers/restaurants/merchants, and demand naturally follows. Currently, only 40-50% of signable merchants have been penetrated.

2. "The AV space will not have a single winner; there will be multiple winners" (Dara Khosrowshahi) — Drawing an analogy to the foundation model space (multiple players + open source). Uber does not build drivers but acts as an "ecosystem service provider": charging stations, financing, insurance, data, and demand.

3. "AVs on the Uber network generate 30% more trips and revenue per vehicle per day than independently operated AVs" (Dara Khosrowshahi) — This 30% has a decisive impact on AV investment returns and is Uber's core bargaining chip in partnering with AV companies.

4. "We used up the entire year's AI budget in one quarter" (Dara Khosrowshahi) — Uber adopts a two-phase strategy: frontier model exploration → efficient model scaling. Meanwhile, headcount growth is controlled because engineer efficiency has already improved tenfold.

5. "Uber One loses money on you in the first year, but you become more profitable after 2-4 years" (Dara Khosrowshahi) — Drawing on the Amazon Prime model, the membership system is the core adhesive of Uber's super app strategy. Currently, 50 million members, up 50% year-over-year.

6. "I look for 'troublemakers' in the company — they are mutations, the source of the company's evolution" (Dara Khosrowshahi) — The larger a company grows, the more it tends toward harmony, but "troublemakers" are often driven away. Dara deliberately creates random interactions to find them.

7. "Barry Diller insists on getting the truth from the source — filtering wears away the edges of a story, and the edges are your advantage" (Dara Khosrowshahi) — Dara applies this at Uber: stay transparent, tell the team the truth, and obtain unfiltered information from the bottom up.

8. "Herbert Allen said: Bet on people, not on companies" (Dara Khosrowshahi) — Dara took a significant pay cut when leaving Allen & Co to join Barry Diller, but "I bet on a person. The best decisions in my life have all been bets on people."