Xinzhou Wu, head of automotive at Nvidia, sat down for an in‑depth interview to discuss how the company’s GPU expertise is reshaping the future of cars. While Nvidia dominates headlines for its AI‑driven data‑center growth, its automotive division has quietly supplied chips to major manufacturers for years, now aiming to deliver a complete autonomous driving stack that can be deployed as a plug‑and‑play solution.

From Software‑Defined to AI‑Defined Vehicles

Wu explained that the industry’s buzzword “software‑defined vehicle” referred to consolidating dozens of electronic control units (ECUs) into one or two powerful computers. “Today we’re entering the era of the ‘AI‑defined vehicle,’ where generative AI models continuously rewrite and upgrade the car’s software over‑the‑air,” he said. This shift accelerates development cycles while redefining what a vehicle fundamentally is.

China’s Early Lead

Having spent five years leading an autonomous team at a Chinese OEM, Wu noted that China bypassed the legacy gasoline‑car transition by building EV platforms from the ground up. This head start allowed Chinese manufacturers to integrate Nvidia’s AI chips without the burden of retrofitting legacy ECUs, giving them a competitive edge over Western OEMs still grappling with hybrid‑to‑EV conversions.

Resource Competition Within Nvidia

Wu candidly described the internal tug‑of‑war for compute capacity between Nvidia’s booming AI business and its automotive projects. “When the AI division’s demand spikes, we have to argue for bandwidth and silicon allocation,” he said, emphasizing that automotive customers are often cost‑averse and slower to adopt, making the case for resources a constant negotiation.

AI‑Defined Autonomy and the Tesla Question

The interview also dove into technical specifics: Nvidia’s autonomy stack blends a “classical” control layer with reasoning models that let an AI system effectively talk to itself to decide how to drive. When asked about Tesla’s Full Self‑Driving (FSD) claims without lidar, Wu remained cautious, noting that Nvidia’s approach aims to achieve comparable safety levels through pure AI‑defined architectures.

Overall, Wu’s perspective paints a picture of an industry at a crossroads—balancing rapid AI innovation, cost pressures, and regulatory scrutiny. Nvidia’s automotive division appears poised to lead the next wave of vehicle intelligence, provided it can secure the necessary compute resources and convince traditional OEMs to adopt the AI‑defined paradigm.