Autobrains launches Munich robotaxi pilot with Uber, Nvidia

Autobrains, an Israeli autonomous-driving company, announced a three-way collaboration with Uber and Nvidia to launch a robotaxi programme in Munich, according to a joint press release posted on Uber's investor site and contemporaneous reporting by Automotive World, Electrive, Euronews, The Times of Israel and The Jerusalem Post. The partners describe the deployment as OEM-agnostic and built on NVIDIA DRIVE Hyperion hardware; Electrive reports the stack targets SAE Level 4 capability. Multiple outlets note that the programme is conditional on German regulatory approval and that details such as fleet size and vehicle OEM(s) have not been disclosed.

Autobrains' core contribution is described in coverage as an "Agentic AI" architecture that decomposes driving into multiple specialised agents rather than a single monolithic model. The Jerusalem Post published a direct quote from Autobrains founder and CEO Igal Raichelgauz: "We created a unique system for autonomous driving through agentic artificial intelligence, which breaks down the reasoning into small tasks." Automotive World and Electrive report the approach is designed to work with standard sensor suites and to be portable across vehicle platforms that support Nvidia's Hyperion compute.

• •Software: Autobrains' Agentic AI for perception, planning and decision-making (reported by Automotive World and The Jerusalem Post)
• •Platform: Uber's ride-hailing network to provide booking, routing and marketplace integration (noted in the partners' announcement)
• •Hardware/compute: NVIDIA DRIVE Hyperion as the vehicle compute and sensor reference architecture (reported by Electrive and the press release)

Industry-pattern observations: decomposing driving into specialist agents is a growing architectural choice among autonomy vendors aiming to separate long-horizon reasoning, local planning, and safety monitors. This design can simplify testing and validation of discrete behaviors and can help map responsibilities to redundant safety layers, but it also increases integration and orchestration complexity compared with monolithic end-to-end models. The reported OEM-agnostic objective aligns with recent industry efforts to avoid vendor lock-in and reduce per-vehicle integration costs when scaling robotaxi fleets.

Europe has lagged the U.S. in large-scale robotaxi deployments, in part due to stricter regulatory regimes and differing urban mobility patterns. Coverage from Euronews and The Times of Israel places the Munich choice within a broader EU effort to harmonise autonomous-vehicle testing across member states. For Uber, the Automotive World reporting frames the tie-up as a way to access autonomy without depending on a single automaker. For automakers, the OEM-agnostic, Hyperion-compatible approach provides a possible route to participate in ride-hailing programmes without building a full-stack self-driving system in-house.

• •Regulatory outcome: whether German authorities grant the required permissions for commercial robotaxi operations in Munich, as noted by Euronews and the partners' announcement.
• •OEM partnerships and vehicle selections: media reporting says the partners will announce participating manufacturers later; the choice will affect sensor suites and integration timelines.
• •Safety data and operational design domains (ODDs): outlets report Autobrains intends human-supervised test drives initially and to compile safety statistics; published operational limits and safety metrics will determine how broadly the service can scale.
• •Interoperability with Hyperion: real-world integration with NVIDIA DRIVE Hyperion across different OEM platforms will be a practical test of the OEM-agnostic claim.

For practitioners: tracking published safety metrics, ODD definitions, and the software-to-hardware integration patterns from this programme will provide useful signals about realistic timelines, validation strategies, and fleet-operational approaches for scaling robotaxi services.