Nvidia's auto group won supply contracts with Mercedes, Stellantis, and BYD this year. Inside, the team still out argues the data center business for the GPUs that make the chips work.
Nvidia's head of automotive has spent three years telling carmakers that the future belongs to a few powerful on-board computers instead of dozens of small ones. The same executives who buy that pitch from him are now hearing it from Mercedes-Benz, Stellantis, Lucid, and BYD. Inside Nvidia, Xinzhou Wu has a harder audience: his own colleagues on the data-center side, who control the GPU supply that goes into the chips he sells to those carmakers. In a long Decoder interview this month, Wu says the arguments are won on the strength of the use case, not the size of the business.
That detail lands in a quarter that made the internal fight numerically visible for the first time. Nvidia's Q1 FY27 earnings release, covering the period ended April 26, 2026, reported total revenue of $81.6 billion, with Data Center revenue at a record $75.2 billion and Data Center networking at a record $14.8 billion. The same release folds the old Automotive line into a new "Edge Computing" segment that also includes gaming, robotics, and AI-RAN base stations. For scale, Nvidia's last separately reported full-year automotive figure, in fiscal 2026, was about $1.7 billion, roughly 1.3% of total revenue at the time. The structural demotion is the story, not the design wins.
The product story is a transition Nvidia calls the "AI-defined vehicle," which in practice means one or two central computers replacing dozens of small electronic control units, or ECUs, that have run cars since the 1980s. Wu's pitch to Mercedes, Stellantis, and the Chinese EV brands is that those central computers can be upgraded over the air, that the same hardware can run better driver-assistance software the longer a customer owns the car, and that the architecture is finally L4-ready, meaning able to drive itself without a human in defined conditions, on production silicon. Mercedes shipped the first of those cars in January. Nvidia's January 7, 2026 blog post announced DRIVE AV software in the all-new Mercedes-Benz CLA, and a follow-up post on January 30 confirmed a future S-Class built on the full DRIVE Hyperion stack.
The design-win list has grown faster than the internal allocation has shifted. Axios reported on May 13, 2026 that Nvidia's autonomous-driving roster now stretches from Mercedes and Stellantis to Lucid, Volkswagen, Zoox, Hyundai, Nissan, BYD, and Geely. MotorTrend added in March that the latest Hyperion partners are building toward Level 4 autonomy, the next step beyond the driver-assistance systems most new cars ship with today. Wu's team is winning more external sockets, on paper, than at any point in Nvidia's history. They are still arguing for the GPUs that fill those sockets against the data-center group, which sells to OpenAI, Microsoft, and the rest of the AI-cloud customer base.
Wu offered an explanation for why China, specifically, has moved faster on the centralized architecture. In the interview he credits the EV-native platforms of Chinese carmakers, which started life without the legacy wiring of a combustion-era car, for forcing an earlier move to the software-defined vehicle, or SDV, design. That pace is now what global auto partners expect Nvidia to match, he said, and it is part of what makes the internal GPU argument a strategic one rather than a procurement fight.
Asked directly about Tesla's claim that its Full Self-Driving system can do what Elon Musk says it can without lidar, a laser sensor that maps the road in 3D, Wu declined to put a quote on the line. The host characterized Wu's answer as a hedged but pointed response, and readers should treat his stance as on-the-record commentary through that host characterization rather than a direct quote.
For a buyer in 2026, the practical version of this story is a short list of questions. Which generation of Nvidia Drive, if any, is on the car? Is the central computer an over-the-air-upgradable unit, or is it locked at the factory? What is the brand's actual Level 3 plan, the step below L4 that lets a human take their eyes off the road in defined conditions, and on what timeline? Mercedes and BYD can answer those questions today. Most other brands cannot, and the reason has less to do with the car than with how Nvidia splits its GPUs between the data center and the driveway.