Even the head of Nvidia's automotive division has to fight for computing resources inside the company. Xinzhou Wu, who leads the chipmaker's automotive business, says he competes with the company's booming AI data center unit for GPU capacity on a weekly basis. The internal debates involve revenue projections, strategic priorities, and sometimes require CEO Jensen Huang to step in. It's a revealing glimpse into how Nvidia balances its dominant AI chip business with longer-term bets like autonomous driving.

Wu says the auto industry is finally embracing the "software-defined vehicle" concept, replacing dozens of electronic control units with one or two powerful computers. He points to Mercedes' current generation of EVs using Nvidia's platform as proof. But the transition has been slower than expected, with legacy automakers struggling to adapt their supply chains and engineering teams. Chinese carmakers, starting from scratch with EV architectures, have moved faster. Wu, who previously worked at Chinese OEM XPeng, says the pace of change in China forced even global players there to accelerate.

On autonomy, Wu believes Level 4 self-driving will reach mainstream consumers within five years. He argues lidar is necessary for safety at that level, pushing back against Tesla's camera-only approach. Nvidia's strategy involves an open platform where automakers can pick and choose components. The company also shares anonymized driving data across its partner ecosystem to train better models. Wu describes a future where cars run two parallel stacks: an end-to-end AI model for driving decisions and a classical safety guardrail that validates every trajectory in real time.

Source: The Verge