Inside every autonomous vehicle is a compute platform interpreting data and making decisions in real time. But unlike a data center, where space, cooling, and power are planned around servers, putting an advanced computer into a car is far more constrained. The system must process data from multiple sensors, run AI models, and manage the vehicle’s operation — yet it must fit into a compact footprint, draw limited power from the vehicle’s battery, and operate reliably in a closed environment with little airflow. Balancing these demands is an ongoing engineering challenge. In this blog, we share how Avride — a company building autonomous vehicles — approaches this task.

CPU’s Role Inside the Vehicle

Avride’s vehicles are powered by the main compute unit (MCU) which performs the core computations behind perception, motion planning, decision-making, and much more. In practice, it’s responsible for what the vehicle sees, interprets, and responds to on the road. A CPU inside the main compute unit coordinates the overall compute pipeline, managing the flow of tasks and communication, while also handling several autonomy functions directly, like localization and vehicle control. It is also continuously monitoring the health and reliability of onboard systems, detecting anomalies early and maintaining safe operation under all conditions — whether by reducing speed or coming to a complete stop when needed.

Avride’s engineers were looking for a CPU platform capable of supporting these critical workloads without drawing excessive power or generating excess heat.

“In such a complex system as an autonomous vehicle every watt matters. Every watt saved can be used to extend range, improve performance, or increase reliability. In addition to that, maintaining a stable operating temperature is especially important for overall vehicle safety,” said Vitality Podkolzin, Head of Embedded Systems Development at Avride.

Finding the Right Compute Foundation

Ampere®’s Arm-based architecture appeared to be a practical match for these requirements. Ampere processors paired with ADLINK motherboards provided a notably well-balanced mix of efficiency, reliability, and integration simplicity. The configuration delivered real-time autonomy capabilities with reduced power and thermal demands. MCU’s powered by Ampere’s processors could fit into the current vehicles' footprint without any architectural changes.

“Energy efficiency and predictability are just as important to us as raw compute power,” said Vitality Podkolzin, Head of Embedded Systems Development at Avride. “Ampere’s architecture offers a good balance for our needs — stable performance for the CPU-side workloads and low power consumption, which together support reliable and predictable system behavior.”

“Avride’s use case is a perfect example of what Ampere CPUs were designed to do,” said Jeff Wittich, Chief Product Officer at Ampere. “They needed efficient compute that could support demanding real-time workloads in a power and thermally constrained environment. This is exactly where Ampere delivers.”

Designed to Scale in the Real World

Today, the majority of Avride’s autonomous vehicle fleet operates in Austin and Dallas in Texas — one of the most demanding urban testbeds in North America. High temperatures and dense traffic push both thermal and performance margins to their limits.

Avride already had its own proven cooling system for the onboard computer, and the new compute platform integrated into it cleanly. The lower system load meant the existing design required no modifications for operation in hotter climates — an important consideration for maintaining overall system scalability.

By lowering power consumption and optimizing system efficiency, Ampere enabled additional capacity for future growth. With more available power and thermal margin, new capabilities can be added without altering the underlying vehicle hardware.

Looking Ahead

Months of testing the Ampere-based compute platform has confirmed its ability to support an efficient and integrated environment across the fleet. This foundation provides the performance, consistency, and system-level efficiency required for large-scale autonomous operations, while maintaining flexibility for hardware updates.

By rethinking how compute is implemented inside the vehicle, the company has created room to explore additional capabilities and future improvements.