RobotWorld

How inDrive Plans to Bring Robotaxis to Central Asia by 2027 — and Why the Tech Stack Matters

By RobotWorld·10/24/2026

A New Frontier for Autonomous Ride-Hailing

When most people think of robotaxis, they picture San Francisco or Shanghai. Kazakhstan's capital, Astana, probably doesn't come to mind. That's about to change. Ride-hailing platform inDrive — one of the fastest-growing Uber rivals in emerging markets — has announced plans to begin testing autonomous vehicles in Astana in the first half of 2027, in partnership with Chinese autonomous-driving company Pony.ai and backed by Nvidia's compute platform.

It's an ambitious bet, and the technical story behind it is worth unpacking.

Who Are the Players?

inDrive operates across Latin America, Africa, the Middle East, and Central Asia, competing on a peer-to-peer pricing model that lets drivers and passengers negotiate fares directly. Its stronghold in markets like Kazakhstan gives it a natural testing ground that's less saturated — and less regulated — than Western AV corridors.

Pony.ai is one of China's most established autonomous-driving companies, with a track record of robotaxi operations in Beijing, Guangzhou, and select U.S. locations. The company has built a full-stack AV system — meaning it develops its own sensors, perception software, mapping, and planning algorithms rather than integrating off-the-shelf components.

Nvidia provides the underlying compute muscle. Autonomous vehicles are, at their core, real-time AI inference machines. They must process streams from cameras, LiDAR, and radar simultaneously, run object detection, predict the behavior of other road users, and issue safe driving commands — all within milliseconds. Doing that reliably requires serious silicon.

The Hardware Foundation: Why Compute Is Everything

Modern robotaxis don't just need a powerful processor — they need an automotive-grade one, validated for the safety standards and environmental conditions of real roads. Nvidia's DRIVE platform is specifically engineered for this. The NVIDIA DRIVE AGX Orin, for example, delivers up to 254 INT8 TOPS and is designed to handle the full compute stack for L2+ through L4 autonomy — covering sensor fusion, path planning, and even in-cabin AI in a single system. For fleets targeting higher autonomy levels, the newer NVIDIA DRIVE AGX Thor, built on Blackwell GPU architecture, pushes that figure dramatically higher while consolidating autonomous driving and cockpit functions onto one ASIL-D–certified SoC.

Pony.ai's integration with Nvidia's platform means the robotaxis destined for Astana will likely leverage this kind of centralized, high-throughput compute to handle the complexity of real urban driving.

Complementing compute is sensing. Robotaxis typically rely on a layered sensor suite: cameras for texture and color, radar for velocity and all-weather reliability, and LiDAR for precise 3D mapping of the environment. For production-scale deployments, automotive-grade LiDAR sensors like the Hesai AT128 — a 128-channel hybrid solid-state unit certified to ISO 26262 ASIL-B — represent the kind of hardware that goes into mass-deployed robotaxi fleets. Similarly, the Luminar Iris offers 1550 nm laser sensing capable of detecting low-reflectivity objects at range, even in low-light conditions. And the RoboSense RS-LiDAR-M1 brings solid-state MEMS design to ADAS and AV applications with a 200 m detection range. These sensors, combined with AI compute platforms, form the perceptual backbone that lets a vehicle "see" reliably enough to carry passengers.

Why Astana? Why Now?

Choosing a Central Asian city for early testing isn't arbitrary. Astana offers a relatively modern, grid-based urban layout, lower traffic density than megacities, and a government that has signaled openness to technology partnerships. For a company like inDrive — already embedded in the local ride-hailing market — the regulatory relationships and user base are already in place.

There's also a commercial logic to emerging-market AV deployment that's easy to overlook. Labor costs for human drivers are often the primary expense in ride-hailing economics. In markets where fare prices are kept low by competitive pressure, the margin benefit of removing the driver is proportionally larger than in premium Western markets. If robotaxis can operate safely and efficiently, the unit economics could be compelling.

The 2027 target is a testing phase, not a commercial launch — an important distinction. Real-world pilots in controlled corridors allow companies to gather local data on road behavior, weather conditions (Astana experiences harsh winters), and edge cases specific to the region before scaling.

What This Signals for the Broader Industry

The inDrive–Pony.ai announcement reflects a broader pattern: AV technology, long associated with Silicon Valley and Chinese tech hubs, is spreading outward into markets that don't yet have dominant incumbents. Companies that arrive early with tested hardware and local partnerships can define the standard before regulators and competitors catch up.

It also underscores how much the AV supply chain has matured. Purpose-built automotive AI chips, production-ready LiDAR sensors, and full-stack software platforms that once existed only in prototype form are now available at the scale needed for fleet deployment.

For those interested in how autonomous systems are developed at a more accessible level — from edge AI inference to robotics perception — platforms like the NVIDIA Jetson AGX Orin 64GB (up to 275 TOPS in a compact module) or the NVIDIA Jetson Orin Nano Super Developer Kit offer a hands-on way to explore the same underlying AI architecture that powers production autonomous vehicles, without the price tag of a full automotive compute platform.

The Road Ahead

Robotaxis in Central Asia by 2027 would have seemed far-fetched a decade ago. Today, it's a plausible near-term milestone — one that combines mature AI compute, proven autonomous-driving software, and a ride-hailing platform with genuine local reach. Whether the timeline holds will depend on regulatory approvals, winter performance data, and the thousand real-world edge cases that only actual road testing can reveal.

What's clear is that autonomous mobility is no longer a story confined to a handful of global cities. The frontier is expanding — and Astana may be one of the next places it arrives.

References

This article was drafted with AI assistance and reviewed before publishing.

Related reading