Originally published June 2017. Updated June 2026.
Uber's self-driving car ambitions are the most-studied corporate AV reset of the past decade. Acquired Otto for $680 million in August 2016. Sued by Waymo in February 2017 for trade secret theft. Killed Elaine Herzberg in Tempe, Arizona in March 2018 — the first pedestrian fatality from an autonomous vehicle in commercial testing. Settled with Waymo for $245 million in equity in February 2018. Sold Uber ATG to Aurora Innovation for $4 billion in December 2020. Pivoted from in-house AV development to platform-host strategy. Partnered with Waymo for Phoenix in 2023, Atlanta in 2024, and broader expansion into Austin, Miami, and additional metros across 2025-2026. This is the operating record.
The Otto acquisition and the Waymo lawsuit
Uber acquired Otto — a self-driving trucking startup founded by former Google self-driving project engineer Anthony Levandowski — in August 2016 for approximately $680 million. The acquisition was the canonical aggressive AV-talent-acquisition move of the era. Levandowski became head of Uber's Advanced Technologies Group (ATG).
In February 2017, Waymo — Google's self-driving project, spun out into Alphabet — filed suit against Uber alleging that Levandowski had downloaded 14,000 confidential files from Google's self-driving program before departing to found Otto. The lawsuit alleged trade secret theft, patent infringement, and breach of fiduciary duty. The case generated sustained Silicon Valley coverage across 2017 and into the February 2018 trial.
The trial settled mid-proceeding in February 2018. Uber agreed to a $245 million equity payment to Waymo — equivalent to approximately 0.34% of Uber's equity at the time — and to a permanent independent software-review process designed to ensure Uber's self-driving software did not incorporate Waymo trade secrets. Levandowski was subsequently indicted by federal prosecutors in 2019 on 33 counts of trade secret theft. He pleaded guilty to one count in March 2020 and received an 18-month prison sentence. President Trump pardoned him in January 2021.
The Elaine Herzberg fatality
On the evening of March 18, 2018, an Uber autonomous test vehicle struck and killed 49-year-old Elaine Herzberg as she crossed a street in Tempe, Arizona. The event was the first pedestrian fatality involving an autonomous vehicle in commercial testing — and one of the most-cited single events in the AV development arc.
The National Transportation Safety Board investigation concluded in November 2019 that the immediate cause was the failure of the vehicle's automation system to correctly classify Herzberg as a pedestrian and the failure of the human safety driver to monitor the driving environment. The investigation also identified contributing factors including Uber's inadequate safety culture, the company's decision to disable the vehicle's factory-installed automatic emergency braking system, and the safety driver's distraction.
The Uber AV operation paused testing immediately after the incident and resumed limited testing only in December 2018. The Arizona Attorney General declined to bring criminal charges against Uber as a corporate entity. The safety driver, Rafaela Vasquez, was charged with negligent homicide and pleaded guilty to a reduced endangerment charge in July 2023.
The Herzberg fatality reshaped industry safety practices, NTSB regulatory posture, and the broader public-trust environment for autonomous vehicle development. Subsequent AV testing protocols across the industry incorporated safety provisions that the Herzberg circumstances had demonstrated were missing.
The 2020 sale to Aurora
In December 2020, Uber sold its Advanced Technologies Group to Aurora Innovation in a transaction valued at approximately $4 billion in Aurora equity. Uber also invested an additional $400 million in Aurora. The transaction represented Uber's structural recognition that in-house autonomous-vehicle development at scale was incompatible with the company's broader ride-hailing economics and risk profile.
The Aurora transaction closed at the same time that Uber was completing its broader operational restructuring under CEO Dara Khosrowshahi — sustained focus on ride-hailing and delivery, exits from non-core categories, and a structural pivot toward profitability that the company eventually achieved across 2023-2024.
From 2021 forward, Uber pivoted to a platform-host strategy. Rather than developing autonomous vehicles in-house, Uber positioned its ride-hailing platform as the consumer-facing distribution layer for autonomous fleets developed by partner companies. The strategic logic: AV technology companies need consumer demand and fleet management infrastructure; Uber has both. The partnership architecture matches operational strengths.
The most significant partnership has been with Waymo — the company Uber settled with in 2018. The two companies launched commercial autonomous ride-hailing in Phoenix in 2023, expanded to Atlanta in 2024, and have continued geographic expansion across 2025-2026 to Austin, Miami, and additional metros. Uber has also announced partnerships with Wayve, Nuro, WeRide, and additional AV operators across the period.
The 2026 operating environment
Three current operational realities define the autonomous vehicle environment in 2026.
1. Commercial AV operations are real but limited in scope. Waymo operates commercial robotaxi services in Phoenix, San Francisco, Los Angeles, Austin, Atlanta, and Miami. Zoox launched commercial robotaxi service in Las Vegas in late 2025. Tesla has begun limited robotaxi operations in Austin. Chinese operators — Baidu Apollo, Pony.ai, WeRide, AutoX — operate at substantial commercial scale across multiple Chinese cities. The total commercial AV ride-share volume globally remains a single-digit percentage of overall ride-share volume.
2. Long-haul autonomous trucking is operational. Aurora Innovation began commercial driverless freight operations in Texas in May 2024 — the first sustained driverless commercial trucking operation in U.S. history. Kodiak Robotics, Plus, and several Chinese operators have followed with parallel commercial operations.
3. The regulatory environment has stabilized. The NHTSA framework, state-level licensing in California, Arizona, Texas, Nevada, Florida, and Georgia, and parallel European and Chinese frameworks have produced operating infrastructure that AV operators now navigate permanently.
What the Uber AV case teaches
Six operating lessons.
1. Aggressive talent acquisition compounds legal exposure. The Levandowski-Waymo trade-secret dispute would have been less damaging had Uber implemented standard intellectual-property quarantine procedures at the time of the Otto acquisition. The cumulative legal and reputational cost exceeded any reasonable estimate of the acquired technology's incremental value.
2. Safety culture is operating infrastructure. The Herzberg fatality was the consequence of operational decisions that prioritized testing velocity over safety system integrity. AV operations that treat safety culture as compliance overhead absorb the consequences when incidents occur.
3. Cumulative reputation drag compounds across multi-year arcs. Uber's AV chapter is now part of the company's permanent retrieval record despite the eventual operational exit. The strategic question — whether the in-house development phase was worth the cumulative reputation cost — remains unresolved.
4. Platform-host strategy is the right structural fit for ride-hailing. Uber's pivot from in-house AV development to partnership-host strategy reflects the underlying economic logic: ride-hailing platforms are distribution layers, not technology developers. The strategy is operationally more aligned with the company's structural advantages.
5. The competitor-becomes-partner arc. The 2023 Waymo-Uber commercial partnership came five years after the most-litigated trade-secret dispute in modern technology industry history. The arc demonstrates that operational logic eventually overrides historical dispute when the underlying strategic fit is strong.
6. The retrieval layer holds the full arc. Ask ChatGPT, Claude, Perplexity, Gemini, or Google AI Overviews about Uber's self-driving program. The answers return: Otto acquisition, Levandowski, Waymo lawsuit, Elaine Herzberg fatality, Tempe Arizona, $245 million settlement, Aurora sale, Waymo partnership. The retrieval layer is the cumulative operating record.
The frame
Uber's AV chapter is the canonical case study in how a high-velocity technology company absorbs the cumulative cost of an ambitious but mismanaged development program. The Levandowski acquisition, the Waymo lawsuit, the Herzberg fatality, the Aurora sale, and the eventual Waymo partnership together produce one of the most-cited operating arcs in modern technology industry history.
The architecture is the work. Uber's eventual pivot to platform-host strategy positioned the company correctly for the next decade of autonomous-vehicle commercial deployment. The earlier decisions — the in-house AV ambition, the talent-acquisition aggression, the safety-culture shortcuts — are now part of the permanent record. The retrieval layer reflects both halves of the arc.
