Uber CEO on AI, Autonomous Vehicles, and the Future of Transportation

Summary of the video / interview

• AI is reshaping Uber’s entire model and the transportation timeline. The Uber CEO argues that the “challenge and opportunity” Uber now faces is the rise of AI—both as an internal tool (to build better systems faster) and as “physical AI” that will extend into autonomous vehicles and drones. He frames this as a likely multi-trillion-dollar marketplace that will change how society operates.

• Uber is fundamentally both digital and physical, with probabilistic outcomes. Unlike many software-first companies, Uber’s service spans the real world where things go wrong (traffic, driver cancellations, late deliveries). The CEO says Uber has already used AI for some time to operate in this probabilistic environment, but today’s generative AI and larger models enable more accurate prediction, better utilization, and more powerful automation across the company.

• The leadership challenge on arrival was “chaos,” and the response was structured simplification. When he took over Uber, he describes a period of internal and external turbulence: board conflict, stakeholder/regulator/public trust issues, and management instability. His method is to simplify what appears unassailable by breaking problems into components, then tackle each area with specific initiatives:

  • Board-level alignment (bringing in a new chairman, Ron Sugar, to unite the board around “fate of the company” rather than control).
  • Rebuilding trust via listening tours, acting on feedback, and improving internal/external communication.
  • Team and talent resets (moving away from “old world” leaders and building new management capacity).

• Stress management and an “engineering mindset.” He says he doesn’t stress much because he approaches situations like engineering problems: list problems, test/learn, avoid overthinking, and focus on solutions. He ties this attitude to his personal history as a childhood immigrant from Iran who lost everything and later rebuilt—experiences he says made him mentally resilient against chaos.

• Driving AI adoption, but not as top-down “mandates.” Uber’s culture is described as bottom-up. The CEO isn’t trying to be the single point of failure; instead, he pushes teams to rebuild processes from first principles using AI. A key framing: using AI to optimize parts of workflows (e.g., 20–30%) is helpful, but the real advantage comes from rethinking systems more deeply.

• Trade-offs: AI is powerful but expensive. Uber’s CEO says the company “blew through” its AI budget in a quarter, so they’re now metering costs, potentially slowing headcount growth while leveraging increased engineering throughput. He describes a strategy of:

  • Using frontier/expensive models for exploration
  • Switching to more efficient (including open-source) models for scaled production once patterns and use-cases are proven

• Uber’s core advantage for AVs is supply-side aggregation. He argues Uber must win the “demand aggregation” race for autonomous vehicles by first winning the supply side:

  • Recruiting more drivers and merchants broadly (including smaller/suburban markets, not only top cities).
  • For AVs specifically, building an ecosystem: partnerships, data collection, depots/charging, fleet and financing, and even autonomous insurance.
  • He claims AVs on Uber’s network can be significantly busier (30%+ more busy per vehicle) than non-network equivalents, improving ROI.

• AV is early in an S-curve, with many partners and no single guaranteed model winner. Uber reports many AV-related partnerships across compute, sensing, and software driver stacks. The CEO expects:

  • Multiple players in “foundation models” (analogous dynamics for AV)
  • “Amalgamation” of business models—coexistence of competition and collaboration with companies like AV OEM partners (and potentially competitors in parts of the stack)

• Premortem risks: public backlash and regulatory/societal readiness.

  • The biggest industry concern isn’t just technical safety; it’s how AI/automation impacts people (jobs, electricity costs, and everyday life).
  • For AVs: ensuring safe interaction with emergency services, achieving equitable access (not only for wealthy areas), and maintaining dialogue with regulators and drivers.
  • In the success case, winning supply access would enable Uber to unlock a larger transportation market and reduce transportation costs over time as hardware/software costs fall.

• Drones: progress is real but constrained by battery density and scaling realities. He explains why drone deliveries haven’t scaled as early hype predicted: battery density, payload/range, and recharging. He predicts meaningful scaling for food/grocery over 2–5 years, but not immediate “everyday at your house” normalization—more likely gradually over 5–10 years.

• Regional regulation and adoption vary: Middle East faster; Europe catching up; US slower regionally. He cites:

  • Middle East (e.g., Abu Dhabi/Dubai/Saudi): more entrepreneurial regulators moving faster.
  • Europe: starting commercial robo-taxi pilots and planning activity (e.g., London pilots).
  • US: moving forward unevenly by region (he points to California/Texas and slower places like New York/Boston).

• Uber Eats international lessons emphasize fundamentals and cross-platform advantage. He says success in international expansion comes from getting selection and reliability right (e.g., delivery within ~30 minutes, fewer mistakes). A differentiator is cross-platform upsell: mobility customers are a meaningful source of Eats bookings, plus Uber One loyalty benefits.

• Membership (Uber One): profitable later, expensive early—built like unit economics. He defends membership as trading off short-term margin for long-term customer lifetime value:

  • Membership can be ideal when service costs are relatively stable (Netflix-like).
  • Uber One initially makes customers less profitable but becomes profitable over years.
  • He compares the strategy to Amazon Prime’s early “valley of despair.”

• Marketing’s job evolves as Uber adds more services. He argues marketing isn’t just about getting people into the app; it’s about storytelling that helps users discover and emotionally understand new capabilities (e.g., Reserve and add-on convenience like coffee). He frames Uber as “more than rides”—about time-saving local services.

• Operational leadership principles: truth-from-sources and tolerating “troublemakers.”

  • He credits business mentors for insisting on ground truth (e.g., Barry Diller’s habit of going directly to model builders/data sources rather than filtered narratives).
  • As a leader, he practices transparency and “random interactions” to detect signals and seek out internal “troublemakers,” viewing them as necessary “mutations” for organizational evolution.

• Capital allocation priority: growth and innovation first, with buybacks as secondary. With large free cash flow, he says Uber prioritizes:

  1. organic investment and growth
  2. engineering and algorithm investment
  3. AV/EV ecosystem commitments (including financing partnerships)

Buybacks remain possible, but he frames the current moment as an opportunity-heavy environment for long-term compounding.

• Personal influences and closing theme. He reflects on learning from mentors (Barry Diller; Herbert Allen) about truth, investing in people, and staying open to learning. He closes by crediting his wife as a “kindness” that helped him become his authentic self.

Presenters / contributors

• Uber CEO (interviewee): Dara Khosrowshahi
• Host / interviewer: Sid (name not fully shown in the subtitles; referred to as “Sid” throughout)