Nilay Patel interviews Jinju Wu, head of automotive at NVIDIA, about the state of the auto industry's transition to autonomous and software-defined vehicles. They discuss NVIDIA's open platform strategy, the role of synthetic data and AI models in overcoming the data gap, and the internal resource battles at NVIDIA where the automotive division fights for compute and fab capacity against the booming data center business.
Summarized by Podsumo
NVIDIA's open platform strategy allows automakers of any capability to adopt their technology, from full turnkey solutions (Mercedes) to just using NVIDIA cloud services (Tesla).
To overcome the data gap in autonomous driving, NVIDIA is building an ecosystem where partner OEMs share data, leveraging synthetic data and neural reconstruction for exponential efficiency.
NVIDIA's autonomous driving safety approach involves two parallel stacks: a high-performance AI model and a classical safety guardrail that verifies trajectories in real-time.
Despite being a strategic division, the automotive team must compete for limited GPU compute, fab capacity, and talent against NVIDIA's booming data center business.
"Believe it or not, even NVIDIA... we do have a limited supply of GPU for compute. So we have internal priority... And sometimes we need Jensen to help."
— Jinju Wu
"We are not picking winners per se. We try to help OEMs based on their capability at different levels."
— Jinju Wu
"The short answer is yes. [Level 4 autonomy] requires LIDAR. We believe that LIDAR is an important sensor to provide the safety and the redundancy required..."
— Jinju Wu