inDrive, Pony.ai and Nvidia Plan Robotaxi Tests in Kazakhstan by 2027
The ride-hailing app that lets passengers bargain over fares wants to bring self-driving cars to Central Asia, starting in a city known for brutal winters.
Key points
- inDrive, Pony.ai and Nvidia signed a memorandum of understanding (a non-binding letter of intent to work together) on Friday to test robotaxis in Astana, Kazakhstan, in the first half of 2027.
- Nvidia will supply its Hyperion platform, a ready-made bundle of sensors and computing hardware for autonomous vehicles, and its Halos safety system, which monitors whether a self-driving system is behaving safely.
- Pony.ai's Gen-7 robotaxis will launch with a safety operator in the driver's seat, with fully driverless rides as the eventual goal.
- Astana, a city of roughly 1.7 million people, endures months of snow and ice each winter, making it an unusual but potentially valuable test environment for autonomous driving software.
- inDrive is already exploring a separate robotaxi deal in Riyadh with Saudi firm AI Driver and Chinese company WeRide, suggesting a deliberate push into markets that Uber has not yet locked up.
Three companies signed a deal Friday that could put self-driving cars on the streets of Kazakhstan's capital within 18 months. Ride-hailing app inDrive, Chinese autonomous driving firm Pony.ai, and chip and AI platform company Nvidia announced a memorandum of understanding to begin robotaxi testing in Astana in the first half of 2027, as first reported by TechCrunch.
The tests will start with a human safety operator on board. Fully driverless rides are the long-term ambition, not an immediate promise.
What does each company bring to this?
Each partner covers a distinct gap. inDrive provides the app and the local operator relationship in Kazakhstan. Pony.ai supplies its Gen-7 robotaxis and the autonomous driving software inside them. Nvidia contributes the Hyperion platform, a pre-integrated stack of cameras, radar, lidar (sensors that measure distance using laser pulses) and computing hardware, plus Halos, its safety monitoring system.
Nvidia's argument is that Hyperion cuts development costs by giving companies a common foundation rather than forcing each one to wire together sensors and chips from scratch. Ali Kani, Nvidia's vice president of automotive, told TechCrunch those savings matter most in emerging markets where ride-hailing is already cheap and autonomous services must compete on price from day one.
Nvidia CEO Jensen Huang was visiting Kazakhstan in person on the day the deal was announced. That kind of executive-level presence matters for government relationships, and Kazakhstan had already announced its own autonomous vehicle rules and 2027 pilot plans before this partnership was revealed.
Why Astana specifically?
Astana's winters are punishing. The city spends months under snow and ice, conditions that trip up cameras and sensors far more than a dry California highway does. inDrive CEO Arsen Tomsky framed that directly: success there would show whether the technology can handle similarly harsh cities elsewhere.
The cold-weather angle is genuinely useful to the wider industry. Most high-profile robotaxi programs, including Waymo's fleet in Phoenix and San Francisco, have built their track records in mild climates. A dataset from Astana would be different.
Where does this fit in the wider robotaxi race?
Pony.ai said it aims to have more than 3,500 robotaxis running in over 20 cities worldwide by the end of 2026, up from 1,975 vehicles at the end of June 2026. Kazakhstan is part of a deliberate push beyond China.
That push increasingly crosses paths with Uber. Earlier this month, Uber announced plans to test Pony.ai's robotaxis in London. Uber is also working with British startup Wayve and multiple partners across the Middle East. inDrive's CTO Yuri Misnik described his company's approach as "capital-light by design," which is a pointed contrast to Uber's heavier investment posture.
This week Nvidia also upgraded its Cosmos platform with synthetic data tools aimed at robots and self-driving cars, and published new retrieval models designed to cut the cost of running AI agents. The Kazakhstan announcement fits a pattern of Nvidia moving simultaneously on the hardware, software and partnership fronts in automotive.
A memorandum of understanding is not a contract and carries no binding commitments. Whether the 2027 timeline holds will depend on Kazakhstan's regulatory process and how quickly the three companies can align on operations. inDrive's separate Riyadh agreement, signed in December 2025 with a target pilot date in the first half of 2026, has produced no confirmed public update on whether testing has started. That precedent is worth watching.
