All podcasts / Invest Like the Best / Summary

Dara Khosrowshahi - Uber's Bet on AVs, AI, and Building a Super-App

2026-06-03 - 68 min - source - Read full transcript

Patrick O'Shaughnessy (host)Dara Khosrowshahi

Key insights

Uber's AV strategy is to win as the demand aggregator through supply access, not by building its own autonomous driving technology.

Dara frames Uber as a 'supply-led company': the scarce resource is depot access, charging infrastructure, financing, insurance, and fleet operations, not the driving stack itself. Uber has over 30 AV partnerships (Waymo, Nuro/Lucid, Wayve, Zoox, Pony.ai and others) and positions itself as the go-to-market layer that gets any given AV developer's cars onto the road and generating trips.

autonomous-vehicles

Dara expects many AV winners rather than a single dominant player, mirroring the foundation-model market.

He compares the AV landscape directly to LLMs: multiple frontier players plus open-source entrants will coexist. Some partners (e.g. Waymo) will want their own consumer brand and app while still using Uber's network for incremental utilization, similar to how hotels and airlines use OTAs alongside their direct channels.

autonomous-vehicles

AVs riding on Uber's network run roughly 30% more trips per vehicle per day than AVs operating off-network.

That utilization delta materially changes the ROI math on expensive AV hardware, which is why Dara argues AV manufacturers need Uber's demand aggregation even if they eventually build their own consumer-facing apps.

autonomous-vehicles

AI adoption inside Uber is intentionally bottoms-up rather than mandated from the top.

Dara says he avoids driving specific AI mandates and instead wants 'people inventing all over the company,' citing an example of developers in India driving 10x code-commit throughput using autonomous agents in ways leadership didn't predict or direct. His role is to find and promote those unpredictable pockets of adoption.

ai-adoption

Intelligence is genuinely expensive at scale, and companies have to actively manage the tradeoff between exploration and efficiency.

Uber blew through its full-year AI budget in a single quarter as usage ramped. Dara's resolution: use expensive frontier models to explore new interactions, then migrate proven use cases to cheaper or open-source models once they're validated, since Uber's core business is high-volume and low-margin.

ai-adoption

Dara's approach to leading through organizational chaos is to decompose a seemingly unsolvable problem into independent, tractable components.

He uses a 'vector mathematics' analogy: a complex multi-dimensional problem can be broken into its component dimensions, solved independently, and recombined. Applied to Uber in 2017, this meant treating board dynamics, stakeholder trust, and team composition as separate, addressable problems rather than one undifferentiated crisis.

leadership-and-resilience

Dara's family losing everything after fleeing Iran, and watching it break his father, shaped a deliberate emotional separation between professional outcomes and personal identity.

He describes his father as a formerly commanding figure who never fully rebuilt after immigrating, and says witnessing that taught him to give maximal effort at work while refusing to let business outcomes define who he is - summarized as 'I'm not going to let the chaos of the world affect me mentally.'

leadership-and-resilience

Barry Diller's most important lesson to Dara was to always get the truth directly from the primary source, not from filtered layers of the organization - and Dara now applies that by deliberately cultivating internal 'troublemakers.'

Dara recounts Diller, during the Paramount/Viacom hostile-tender fight, insisting on talking to the junior analyst who actually built the financial model rather than the MD or VP, since seniority tends to filter what a leader hears and the outlier 'P95' data points carry the real signal. Dara now applies that lesson organizationally: as companies scale, incentives push people toward conformity, so he deliberately creates random, non-hierarchical interactions and seeks out internal dissenters, likening companies to organisms that need mutation to avoid stagnation.

leadership-and-resilience

Uber prioritizes organic growth investment and AV capital commitments over buybacks, despite generating over $10 billion in free cash flow.

Dara frames capital allocation as more art than science: get core unit economics right (costs growing slower than revenue) first, then deploy excess capital into algorithms, engineering headcount, and AV fleet commitments (e.g. the Santander financing line), treating buybacks as a lower priority than compounding growth.

capital-allocation

Uber One's membership economics deliberately mirror Amazon Prime's early unprofitability, betting that lifetime value outweighs first-transaction losses.

Dara says acquiring a Uber One member is initially unprofitable because of the value given back (discounts, free delivery, waived fees), but the program is now solidly profitable at 50 million members growing 50% year-over-year, following the same 'valley of despair' pattern Amazon walked through with Prime before public markets understood the unit economics.

capital-allocation

Uber's structural advantage in food delivery and other verticals is cross-platform upsell from its mobility base, not price competition.

About 13% of Uber Eats bookings originate from users who opened the mobility app first. Dara argues this compounding, multi-service loyalty ('more content means more retention,' as with Netflix) is what lets Uber win market share while maintaining higher margins than single-vertical competitors.

super-app-strategy

Uber's endgame for its 'super-app' bet is pre-emptively integrating travel data (flights, hotels) so the app anticipates rides before the user requests them.

Dara describes a future state where Uber uses email/calendar data to pre-book airport transfers and hotel-to-destination rides automatically, and potentially lets the Uber app function as a hotel room key - extending the brand from purely on-demand transactions into planned, multi-day travel experiences via products like Uber Reserve (already a $5B+ run-rate business).

super-app-strategy

Companies

Uber - Subject of the episode; Dara Khosrowshahi is CEO
Expedia - Dara ran Expedia for 13 years before joining Uber; source of his 'demand-first' contrast with Uber's 'supply-first' model
Waymo - One of Uber's ~30 AV partners; case study on partnering with a company that also wants its own consumer brand
Neuro - AV partner building robots/vehicles with Uber; co-developing a Lucid-based AV with Uber
Lucid - Building a mid-size AV (targeted $60-70k) with Nuro for Uber's network
Nvidia - AV partner providing compute and sensors for autonomous driving stacks
Wayve - AV partner building an end-to-end driving model licensed to OEMs, with Uber handling demand
Zoox - Named as one of the multiple AV players Uber expects to coexist with rather than a single winner
Pony.ai - Named as one of the multiple robotaxi/AV players in Uber's partner ecosystem
Santander - Uber's financing partner for a $1B+ line to fund EV and AV fleets
Netflix - Referenced twice: as a membership-economics analogy (fixed content cost) and via Reed Hastings as a leader Dara admires
Amazon - Prime cited as the model for variable-cost membership economics that Uber One followed through its early unprofitable years
Costco - Named alongside Amazon as an example of a great membership-driven business
McDonald's - Uber Eats restaurant partner example used to describe coexistence between platform and merchant brand
Starbucks - Uber Eats restaurant partner example
Chipotle - Uber Eats restaurant partner example
Spotify - Referenced for its internal 'bets board' framework for capital allocation decisions
Allen & Company - Where Dara started in investment banking and met Barry Diller and Herbert Allen; formative mentorship
Zipline - Drone delivery company Dara says he has watched closely as a bellwether for the category
Colossus - Positive Sum's quarterly publication, promoted by the host at the open and close
Positive Sum - Patrick O'Shaughnessy's firm, disclosed as producer of the podcast

Techniques and frameworks

Vector-mathematics problem decomposition - Dara's mental model for leading through chaos: break a complex, multi-dimensional problem into independent components, solve each, then recombine
Supply-led growth - Uber's playbook for new/sparse markets: recruit drivers, restaurants, and couriers first; demand follows once supply is secured
Premortem - O'Shaughnessy's recurring question format: assume failure five years out and work backward to the most likely cause
Membership 'valley of despair' - Amazon Prime-style pattern where early members are unprofitable but lifetime value turns positive after repeat usage; Uber One followed the same curve
Troublemakers-as-mutation - Dara's org-design belief that companies evolve like organisms; deliberately seeking out internal dissenters and unstructured, random interactions surfaces the signals structured processes miss

Summary

Patrick O'Shaughnessy interviews Uber CEO Dara Khosrowshahi across two very different registers: the personal story of how he took the job and manages pressure, and a deep operating discussion of Uber's bets on autonomous vehicles, AI, and its expanding "super-app" strategy. Khosrowshahi opens with the origin story of the 2017 hire - a headhunter call he initially dismissed until Spotify's Daniel Ek told him at a Sun Valley conference that "life is not about happiness, it's about impact." He then describes walking into a chaotic company (a fighting board, a distrustful public, a demoralized workforce) and using a "vector mathematics" mental model to decompose the crisis into separately solvable dimensions: board control, external trust, and internal talent. He connects his composure under pressure to watching his father lose everything and never fully recover after the family fled Iran, which taught him to separate professional outcomes from personal identity.

The AV discussion is the episode's technical core. Khosrowshahi is explicit that Uber does not need to win the autonomous-driving technology race; it needs to win as the demand aggregator across however many AV players end up competing, much as multiple foundation-model providers coexist rather than a single winner emerging. Uber has more than 30 AV partnerships (Waymo, Nuro paired with Lucid, Nvidia, Wayve, Zoox, Pony.ai, and others) and is building the surrounding infrastructure - depots, charging, financing (including a Santander line for EV/AV fleets), and insurance - so that AV developers can focus purely on the driving stack while Uber supplies utilization. He cites AVs on Uber's network running about 30% more trips per vehicle per day than off-network AVs as the concrete proof point, and names supply access, not consumer demand or even regulation, as the biggest risk to Uber's position in the category.

On AI, Khosrowshahi describes a deliberately bottoms-up adoption strategy: rather than mandating specific AI use cases, he wants unpredictable pockets of adoption (like Indian engineering teams driving 10x code-commit throughput with autonomous agents) to surface organically, with his job being to find and promote them. He's candid that intelligence is expensive - Uber blew through its full annual AI budget in a single quarter - and describes a two-speed approach of using frontier models to explore new interactions before migrating proven use cases to cheaper models at scale.

The conversation's second half turns to Uber's super-app ambitions: the Uber One membership program (50 million members, growing 50% year-over-year), which followed the same early-unprofitable, later-compounding economics as Amazon Prime; the cross-platform upsell effect where mobility users convert into Eats, grocery, and now hotel and train bookings; and a vision of using travel data to proactively pre-book rides around flights and hotel stays, potentially turning the Uber app into a hotel room key. Khosrowshahi closes with leadership reflections drawn from two mentors: Barry Diller, who taught him to always seek the primary source of truth rather than filtered reporting layers, and Herbert Allen, who taught him to bet on people rather than companies. He describes intentionally cultivating internal "troublemakers" as sources of organizational adaptation, framing companies as organisms that die without mutation.

Notable Quotes

"Since when has life about happiness? It's about impact." - Daniel Ek, as recounted by Dara Khosrowshahi

"The magic happens when you learn." - Dara Khosrowshahi

"Companies that don't mutate, that just sit with a single process, a single information flow - those are the companies that die. So I'm looking for those mutations. I'm looking for those troublemakers constantly." - Dara Khosrowshahi

"How quickly magic turns to normal... what's magical now is going to seem normal to all of us ten years from now." - Dara Khosrowshahi, on the early experience of riding in an AV