NVIDIA–Uber Expand to 28-Market Robotaxi Network on DRIVE Hyperion by 2028, Opening in LA and SF in H1 2027

Headline Finding

NVIDIA and Uber announced on 16 March 2026 an expanded partnership to deploy a full-stack NVIDIA DRIVE AV-powered robotaxi fleet across 28 cities on four continents by 2028, beginning with Los Angeles and the San Francisco Bay Area in the first half of 2027. The announcement — made alongside BYD, Geely, Nissan (via Wayve software), Bolt, Grab, and Lyft all adopting the same DRIVE Hyperion platform — signals a coordinated, multi-operator push toward commercial L4 ride-hailing at a geographic scale that no single city’s regulatory apparatus has yet been designed to handle. The deployment clock has moved materially ahead of the governance clock.

Detail

At NVIDIA’s GTC conference on 16 March 2026, NVIDIA announced that Uber is building what it described as ‘one of the world’s most expansive autonomous ride-hailing networks’ powered by NVIDIA DRIVE Hyperion. The expanded NVIDIA–Uber partnership targets a fleet of autonomous vehicles running the full-stack NVIDIA DRIVE AV software across 28 cities and four continents by 2028. The rollout begins with Los Angeles and the San Francisco Bay Area in the first half of 2027.

The same DRIVE Hyperion platform is simultaneously being adopted by global automakers BYD, Geely, and Nissan (the latter powered by Wayve software) for next-generation L4 AV programs, and by mobility providers Bolt, Grab, and Lyft for autonomous mobility initiatives. Isuzu and TIER IV are using the NVIDIA DRIVE AGX Thor system-on-a-chip — part of DRIVE Hyperion — for L4 autonomous bus development. The breadth of adopters across OEMs, ride-hailing operators, and bus developers on a single reference architecture implies a convergence of the supply side around one hardware-software stack, accelerating the pace at which L4 vehicles could enter urban service across multiple jurisdictions simultaneously.

Governance Implication

1. The technology/deployment clock moved significantly: a named 28-city, four-continent commercial robotaxi rollout with a 2027 start date is now on the public record, giving cities roughly 12–18 months before the first wave of full-stack L4 ride-hailing vehicles arrives in Los Angeles and San Francisco — jurisdictions that do not yet have comprehensive L4 commercial frameworks in place.
2. The spatial-allocation lever is under pressure: a multi-operator, multi-OEM convergence on one platform means curb access, pick-up/drop-off zones, and lane allocation decisions in 28 cities will need to be made at speed and in coordination, yet no multi-city governance body exists to set consistent spatial rules.
3. The legitimacy lever is at risk: when multiple operators (Uber, Bolt, Grab, Lyft) deploy on the same underlying stack across dozens of cities, public accountability becomes diffuse — it is unclear which entity (platform, OEM, software provider) bears liability, complicating the consent and trust architecture cities need.
4. The fiscal lever is latent but real: a fleet of this scale displacing human-driven ride-hailing generates per-trip fee and congestion-pricing revenue opportunities for cities, but only if permit and fee instruments are in place before deployment begins — cities that wait risk foregoing value capture entirely.
5. The governance clock implication: cities in the 28-market footprint should treat H1 2027 as a hard deadline for establishing av-permitting frameworks, curb-management rules, and per-trip fiscal instruments; the platform convergence means that whatever rules are set will be tested simultaneously by multiple operators at scale.

Sources

https://nvidianews.nvidia.com/news/drive-hyperion-level-4 https://www.reuters.com/technology/nvidia-uber-expand-self-driving-car-partnership-2026-03-16/ https://techcrunch.com/2026/03/16/nvidia-uber-autonomous-vehicle-partnership-28-cities/