Nvidia’s DRIVE Hyperion Platform Just Got a Lot Bigger—Here’s Where New Robotaxi Fleets Are Launching

The expansion directly builds on an October 28, 2025, partnership where Nvidia, Uber, Stellantis, and Foxconn agreed to deploy 5,000 Level 4-ready autonomous vehicles by 2028 . While that earlier announcement laid the groundwork—introducing the DRIVE AGX Hyperion 10 architecture, confirming a 100,000-vehicle Uber target starting in 2027, and launching the Halos Certified safety program—the 2026 update fills in the map . It names specific cities, clarifies manufacturing roles, and broadens the partner roster to include automakers and local specialists who can navigate regional roads and regulations.

Where the New Robotaxi Fleets Are Going

Each partner brings a different geography and operational timeline to the table:

• Foxconn (Taiwan and Asia) — The manufacturing giant is deepening its collaboration with Nvidia to develop and deploy Level 4-ready robotaxi fleets. The initial rollout will start in Kaohsiung, Taiwan, with a planned service launch in 2028 that begins with airport-to-city routes and expands along Taiwan’s high-speed rail network . Foxconn’s contract design and manufacturing muscle is central to scaling the physical vehicle production.
• VinFast + Autobrains (Southeast Asia) — VinFast is working with autonomous-driving software company Autobrains to bring Level 4 vehicles built on DRIVE Hyperion to the Southeast Asian market. This pairing combines a regional automaker with specialized AV software to tailor the platform for local conditions .
• Uber + Autobrains (Munich, Europe) — Uber will integrate several DRIVE Hyperion-powered autonomous vehicle fleets into its global ride-hailing network. Among the most immediately visible projects is a collaboration with Autobrains to launch a robotaxi program in Munich later this year . This puts the European deployment on a faster track than the broader 2028 fleet targets.
• HUMAIN (Saudi Arabia, Middle East) — HUMAIN is working to bring DRIVE Hyperion-powered robotaxis to Saudi Arabia, expanding the platform’s footprint into the Middle East for the first time .

The distribution of partners reveals Nvidia’s strategy: pair its standardized hardware and operating system with local manufacturers and software integrators who understand regional driving environments, regulations, and consumer markets.

What’s Inside the DRIVE Hyperion Platform

All of these deployments run on the same underlying architecture: Nvidia Halos, a full-stack safety system designed for physical AI workloads—meaning it’s built to handle the real-time, safety-critical demands of Level 4 autonomous driving .

The platform consists of four tightly integrated layers:

1. DRIVE AGX in-vehicle compute — The current generation uses two Nvidia DRIVE AGX Thor systems-on-a-chip (SoCs) on a single board, delivering up to 1,000 INT8 TOPS plus 2,000 FP4 TFLOPS of AI performance per SoC . The dual-chip design provides the performance and redundancy needed for autonomous operation without a human fallback.
2. Nvidia Halos OS (built on DriveOS) — The software foundation of the Halos safety system is a safety-certified operating system that handles real-time processing, security monitoring, and functional safety compliance (ISO 26262 ASIL-D capable) .
3. Compatible multimodal sensor suite — A standardized array of 14 high-definition cameras, 9 radars, 1 lidar, 12 ultrasonic sensors, 4 interior cameras, and an exterior microphone array gives vehicles 360-degree perception . Standardizing the sensor set makes it easier for partners to validate their perception stacks.
4. Nvidia DRIVE AV software — Purpose-built software for highly automated and autonomous driving, delivering the end-to-end stack from perception to planning to control .

This hardware-software bundle is what Nvidia is positioning as the universal foundation for robotaxi fleets, regardless of the vehicle manufacturer or ride-hailing network operating on top.

How This Extends the October 2025 Partnership

To understand the significance of June 2026, it’s worth looking back at what was announced in October 2025 :

• Stellantis, Foxconn, and Uber committed to 5,000 Level 4 vehicles by 2028. That number provided the first concrete fleet-sizing target .
• Uber announced plans for 100,000 autonomous vehicles globally, starting in 2027, signaling the ride-hailing giant’s intention to shift a meaningful portion of its network to self-driving .
• The Halos Certified Program launched as a framework for safety evaluation of autonomous vehicles, giving regulators and insurers a standardized assessment method .

What makes the June 2026 announcement an expansion rather than a repeat is three things:

• New partners filling new regions: VinFast (Southeast Asia), Autobrains (software layer for both VinFast and the Munich Uber fleet), and HUMAIN (Middle East) were all absent from the 2025 lineup.
• Concrete deployment cities: The 2025 announcement was broad—the 2026 press release names specific launch cities (Kaohsiung, Munich) and regions (Southeast Asia, Saudi Arabia) with clearer timing.
• Shift toward manufacturing and local integration: Foxconn’s role is now explicitly about vehicle production, not just collaboration, while VinFast and HUMAIN emphasize adapting the platform to their home markets.

By building on the same technical foundation while expanding the list of operators and geographies, Nvidia is following a familiar playbook: become the indispensable infrastructure provider, then let partners compete on services and market execution.