Who is the big winner of the AI competition? Huang Renxun names Musk: he has three major advantages.

NVIDIA Corporation (NVDA.US) CEO Jensen Huangs latest interview video has recently gone viral on social media. In the video, Huang referred to Elon Musk, CEO of Tesla, Inc. and SpaceX, as a player in the AI race that is in a "phenomenal position," and he believes that this advantage comes not from personal style, but from a foundation built on computing infrastructure, real-world data, and three main AI business layouts. Huang stated that the cost of collecting real-world data is extremely high, while Tesla, Inc. (TSLA.US) has one of the largest vehicle fleets in the world, continuously generating massive amounts of driving data. Meanwhile, Tesla, Inc.s AI computing infrastructure is equipped with a large amount of NVIDIA Corporation hardware. Coupled with xAI, Tesla, Inc.s autonomous driving, and the Optimus humanoid robot, Musk has secured an important strategic advantage in the AI era. These remarks quickly spread across social platforms like X, sparking heated discussions within the AI industry and investment circles. As global tech giants continue to ramp up efforts in foundational models, autonomous driving, and humanoid robotics, Huang once again redirected market attention from merely competing on large models to a complete AI ecosystem composed of computing power, data, and end applications. Huang: Musks greatest advantage is infrastructure that others find difficult to replicate According to the circulated video, Huang was asked about the future development of the AI industry, with a focus on Musk's competitive advantages. He stated: The cost of collecting real-world data is incredibly high, and Elon has a huge advantage. In Huang's view, this advantage primarily stems from two aspects. First, it is the AI computing infrastructure. He pointed out that the AI factory used by Tesla, Inc. for autonomous driving training has a large number of NVIDIA Corporation GPUs, making it one of the most powerful AI training platforms in the world, providing ample computing power for the continuous iteration of the Full Self-Driving (FSD) model. Second, it is real-world data. Huang mentioned that Tesla, Inc. has one of the largest fleets of connected vehicles globally, constantly collecting data on driving environments, road conditions, and vehicle operation. This means that compared to AI companies that rely on public data to train their models, Tesla, Inc. can continuously obtain substantial amounts of new data from the real world, which is very costly to acquire and is also a critical foundation for the ongoing evolution of autonomous driving models. He therefore believes that Musk holds a natural advantage in the AI era, and this advantage, accumulated over years, cannot be replicated quickly. The three battlegrounds of AI: foundational models, autonomous driving, humanoid robotics Huang further stated that he is well aware of Musks judgments regarding the future of AI development, believing that Musk is laying out the three most important directions in AI, which Huang classifies as: xAI: responsible for foundational cognitive intelligence, i.e., foundational models and general AI capabilities; Tesla, Inc.: responsible for autonomous driving; Optimus: responsible for humanoid robotics. Huang expressed that these three directions are indeed the three most important battlegrounds in AI. This suggests that in Huang's view, future competition in the AI industry will not only occur between chat-based robots or large language models but will extend to robotics and the physical world. If foundational models are responsible for "thinking," and autonomous driving for "understanding the real world," then humanoid robotics is responsible for "entering the real world and executing tasks." Together, they form a complete closed-loop for the next stage of AI industry development. Shifting from "model competition" to "infrastructure competition" In recent years, the focus of competition in the AI industry has gradually shifted from the scale of model parameters to the more difficult-to-replicate capabilities of infrastructure. On one hand, large models continue to drive rapid growth in investments in GPUs and data centers; on the other, more and more tech companies are starting to recognize that high-quality real-world data is becoming a new scarce resource. As a result, autonomous driving has become an important data source for the AI industry. Tesla, Inc.'s millions of connected vehicles continuously generate driving data every day, which is not only used for FSD training but also allows the company to build a data moat that is difficult for other AI companies to replicate. Meanwhile, Musk has continued to expand the AI landscape in recent years: xAI is responsible for training the next generation of foundational models; Tesla, Inc. continues to advance FSD and Robotaxi; Optimus is positioned as a humanoid robot platform that can be deployed on a large scale in the future. Huangs classification of these three businesses as the most important future development directions in AI also indicates that he believes future AI competition will revolve more around computing power, data, and applications rather than simply who possesses the most powerful large model. This further highlights NVIDIA Corporation's core role in the AI ecosystem. It is worth noting that while evaluating Musk's advantages, Huang once again emphasized the importance of AI infrastructure. Whether for training foundational models, autonomous driving, or robotic systems, continuous investment in large-scale GPU clusters and data centers is required, and NVIDIA Corporation GPUs remain a crucial component of current global AI training and inference infrastructure. Huang has previously noted that in the future, every large enterprise will build its own "AI factory," and the core competitiveness in the AI era lies not only in algorithms but also in the capability to continuously transform data into intelligence through computing infrastructure. By using Tesla, Inc. as an example, he has reiterated this point: what is truly difficult to replicate is not a single model but the long-term accumulation of computing power, real-world data, and the reinforcing flywheel effect of application scenarios. This article is reprinted from "Wall Street Watch," edited by GMTEight: Li Fo.