Jinju Woo - Head of Automotive, NVIDIA. Tracked across 24 mentions in podcasts and expert conversations analyzed by Sonic.

Jinju Woo

Head of Automotive at NVIDIA

Claims

As speaker

Mentioned in

Key positions and views

Tesla is the current leader in L2++ autonomous driving, but the Level 4 competitive landscape is more open, where Waymo is the only player to have proven a city-wide, driverless deployment.

Widespread and safe deployment of Level 4 autonomy across all operational design domains is significantly enhanced by, and likely requires, LiDAR, in addition to cameras and radar.

A dual-stack software architecture, which runs an end-to-end neural network in parallel with a classical, deterministic stack acting as a safety guardrail, is the responsible approach for verifying autonomous system outputs.

The future business model for autonomous driving will shift from one-time hardware sales to capturing a percentage of the recurring revenue generated from every autonomous mile driven.

The rapid industry-wide shift to a centralized compute architecture, as witnessed in China from 2018-2023, is a fundamental and necessary evolution for any OEM wanting to remain competitive.

Timeline

2018-2023

Woo observed a 'massive change' in the Chinese auto industry, where both new and legacy OEMs rapidly adopted a single central computer electrical architecture to remain competitive.

Current

Woo describes NVIDIA's current L2++ autonomous driving system as 'mapless' and states that the company is rolling out its ADAS technology across all Mercedes-Benz vehicles by the end of the year.

Current

Acknowledges that Tesla is ahead of competitors in L2++ technology, while Waymo is the only company proven to deploy driverless L4 at a city-wide scale.

Near Future (Next few years)

Woo states that NVIDIA is working with Uber to roll out a Level 4 autonomous service within this timeframe.

Next Generation

Woo reveals that NVIDIA's next-generation autonomous driving model, for future vehicles, will be trained with embedded language, allowing it to reason and interact via chat.

Podcast consensus on Woo

Points of consensus

▶Jinju Woo consistently presents NVIDIA's automotive strategy as a comprehensive, full-stack solution encompassing hardware (chips, Hyperion platform), software (dual-stack architecture), and data (simulation, data sharing programs).Jul 2026

▶Woo emphasizes the critical role of a centralized compute architecture in modern vehicles, citing its rapid adoption in China between 2018-2023 and its implementation by partners like Mercedes-Benz as essential for competition.Jul 2026

▶A recurring point is the necessity of robust sensor suites for higher levels of autonomy; Woo specifies that while L2++ can be 'mapless' and LiDAR-less, deploying Level 4 across all domains is 'much better' with LiDAR for redundancy and safety.Jul 2026

▶Woo consistently frames NVIDIA's business model as a long-term play, aiming to capture a percentage of revenue from every autonomous mile driven, and supports this by building an ecosystem with partners like Mercedes-Benz, Uber, and even competitors like Tesla via cloud services.Jul 2026

Points of debate

▶There is an internal tension between NVIDIA's goal of dominating the automotive sector and its own limited supply of GPUs, which requires weekly prioritization meetings, sometimes involving the CEO, to allocate compute resources.Jul 2026

▶Woo presents a nuanced competitive view, acknowledging Tesla's leadership in L2++ systems while simultaneously asserting that Waymo is the only company to have successfully deployed L4 at a city-wide scale, positioning NVIDIA as a key enabler for the rest of the industry.

▶A strategic contrast exists between NVIDIA's approach for L2++ systems, which is 'mapless' and does not require LiDAR, and its L4 Hyperion High platform, which explicitly includes LiDAR and other sensors for redundancy, indicating different technological philosophies for different levels of autonomy.Jul 2026

▶Woo highlights NVIDIA's global ambition, with partnerships spanning OEMs worldwide, but also notes significant operational constraints due to regional data regulations in China and Europe that prevent cross-border data aggregation for model training.Jul 2026

Key themes

▶NVIDIA's Full-Stack Automotive EcosystemJul 2026

Jinju Woo details NVIDIA's strategy to provide an end-to-end solution for autonomous driving. This includes the Hyperion hardware platform with its sensor suites, the dual-stack software architecture for safety, cloud-based training and simulation infrastructure, and a data-sharing program among its DRIVE partners.

NVIDIA is not just a chip supplier but is positioning itself as the central nervous system for the software-defined vehicle, creating a sticky ecosystem that makes it difficult for automotive OEMs to switch providers.

▶Pragmatic Approach to Autonomous LevelsJul 2026

Woo articulates a clear distinction in NVIDIA's technological approach for different levels of autonomy. The company pursues a 'mapless', vision-centric L2++ system for mass-market ADAS, while insisting that robust, multi-ODD Level 4 deployment is 'much better' with the inclusion of LiDAR for enhanced safety and redundancy.

This bifurcated strategy allows NVIDIA to compete with vision-only players like Tesla in the L2++ market while simultaneously building a more technologically conservative and safety-focused platform for the future of fully autonomous L4 services.

▶The Business of Autonomous MilesJul 2026

Woo outlines a forward-looking business model focused on recurring revenue. Instead of solely selling hardware, NVIDIA aims to 'get a percentage of the revenues... from every mileage that driven autonomy in the future,' a model supported by partnerships with OEMs like Mercedes-Benz and ride-sharing companies like Uber.

This signals a strategic shift from a transactional hardware sales model to a service-based, recurring revenue model, which could provide a more predictable and scalable income stream if autonomous driving becomes widespread.

▶Navigating Resource and Geopolitical ConstraintsJul 2026

Woo is candid about the challenges NVIDIA faces, including fierce internal competition for its own limited GPU supply, which requires executive-level prioritization. Externally, the company must navigate U.S. export controls on high-performance chips to China and adhere to strict regional data localization laws in Europe and China, which complicates global data aggregation for AI training.

Even a dominant player like NVIDIA is not immune to supply chain constraints and geopolitical fragmentation, which act as significant headwinds to its global automotive ambitions and development pace.

Source episodes

Suggested prompts

How does NVIDIA's strategy of supplying cloud infrastructure to competitors like Tesla, who build their own inference chips, impact the long-term competitive dynamics in the autonomous vehicle market? &nearr;Given the internal competition for GPU resources at NVIDIA, what are the potential risks to the automotive division's development timeline and its ability to compete with rivals who may have more dedicated compute? &nearr;How does NVIDIA's dual-stack software architecture, which uses a classical stack as a 'safety guardrail' for its end-to-end model, address regulatory and public trust concerns about AI in safety-critical systems? &nearr;What are the strategic implications of Woo's assertion that LiDAR is essential for widespread L4 deployment, and how does this position NVIDIA against companies pursuing a vision-only approach? &nearr;

Key concepts

Autonomous Driving Levels (L2++/L4) 1 ep NVIDIA Hyperion Platform 1 ep Sensor Technology (LiDAR, Cameras, Radar) 1 ep Software Architecture (End-to-End, Dual-Stack) 1 ep AI & Simulation (Neurac, Data Sharing) 1 ep Business Strategy & Partnerships 1 ep Compute & Hardware (GPU, Inference Chips) 1 ep Geopolitics & Regulation 1 ep Automotive Market Trends (China) 1 ep

Notable quotes

“even NVIDIA, basically we do have a limited supply of GPU for compute. So we have an internal priority. And, you know, I'm working with my colleagues basically almost on a weekly basis to decide how to set aside these, you know, different compute... And sometimes we need Jensen to help.”

Jinju Woo · Even Nvidia's head of automotive is fighting for compute | Decoder

“we hope that we can get a percentage, you know, of the revenues that, you know, the ecosystem can get from every mileage that driven autonomy in the future.”

Jinju Woo · Even Nvidia's head of automotive is fighting for compute | Decoder

“Other than the end-to-end model, which is basically you have pixel in, you have trajectory out. We also have a classical stack... And when you have two stacks basically kind of run in parallel, this— the classical stack is acting like a— sometimes we call it the big brother, but essentially it's a safety guardrail.”

Jinju Woo · Even Nvidia's head of automotive is fighting for compute | Decoder

“I had the, let's say, opportunity to witness what happened in China, you know, from 2018 to 2023. And, you know, the whole industry, you know, went through this massive change. Just in 5 years... everybody is adapting to a basically a single central computer kind of, you know, electrical architecture because that's how you compete.”

Jinju Woo · Even Nvidia's head of automotive is fighting for compute | Decoder

Report last updated: Jul 28, 2026

Create a free account to see Jinju Woo's full intelligence report - every claim, the relationship network, and AI Q&A across all sources. No card needed.

Get started free

Back to Entities Intelligence Report