Even NVIDIA Lacks Enough GPUs: An Executive Reveals Internal 'Fights' Over Chips

NVIDIA also has to deal with the scarcity of GPUs that characterizes the entire technology sector. This was revealed by Xinzhou Wu, head of the automotive division, in an interview for The Verge's "Decoder" podcast.

According to Wu, even within NVIDIA, the computing resources allocated to training and testing AI models are limited, to the point that different departments are competing internally to access the necessary computing capacity.

The executive explained that there is an internal prioritization system, managed through almost weekly comparisons with colleagues from various teams, to establish how to allocate computing resources between training and testing models. In some cases, Wu added, direct intervention from CEO Jensen Huang is necessary to resolve the more complex issues related to resource allocation.

The statements provide a unique insight into the internal dynamics of a company whose GPUs are now the backbone of the generative AI boom. The demand for NVIDIA chips continues to exceed the available supply, with companies like OpenAI, Microsoft, Meta, xAI, and Amazon engaged in a tight race to build increasingly larger and more complex AI models.

According to Wu, NVIDIA balances immediate business needs with broader strategic opportunities, including what Huang himself describes as the "zero trillion dollar business": entirely new markets that could be worth trillions of dollars in the future.

Among these long-term bets is autonomous driving. Wu stated that NVIDIA believes "everything that moves will become autonomous," and for this reason, the company is consistently investing in providing chips, software, AI models, simulation tools, and security systems for autonomous vehicles. While remaining a much smaller division compared to the data center business, the automotive sector continues to be considered a priority by Huang.

Wu reaffirmed the company's confidence, personally shared by Huang, in the future of autonomous vehicles, specifying that investments concern not only the allocation of external computing capacity but also the production capacity of chip manufacturing plants. This latter aspect, according to the executive, is becoming an additional area of internal competition as the demand for NVIDIA semiconductors continues to grow.