In the era of physical AI, a "visual data competition" will unfold
Morgan Stanley believes that just as chatbots require text data to train large language models (LLM), physical robots need data to train their vision-language-action models (VLA). It is expected that as computing power continues to expand and efficiency improves, AI companies will need a large amount of visual data to create a "digital twin" of the physical world, and visual data will become the competitive focus for AI giants
As artificial intelligence transitions from the digital world to the physical world, the focus of competition will shift to visual data.
In a research report released on the 21st, Morgan Stanley analysts Adam Jonas, Daniela M Haigian, and others pointed out that just as chatbots require text data to train large language models (LLM), physical robots need data to train their vision-language-action models (VLA). Morgan Stanley believes that AI is expanding from purely digital domains into the physical world, such as in areas like autonomous driving, humanoid robots, and electric vertical takeoff and landing vehicles (eVTOLs).
The report cites Elon Musk's earlier statement this year that "within three to four years, AI will be able to solve any cognitive task that does not involve the physical world." Morgan Stanley expects that as computational power continues to scale and efficiency improves, AI companies will need vast amounts of visual data to create a "digital twin" of the physical world, which means constructing a virtual model of the physical world using high-precision visual data, leading to a global "photon race" for optical data.
The report also compares photon data to "Fat tuna," suggesting that if data cannot be captured or utilized, it is as worthless as tuna in the ocean; however, once the ability to capture and process this data is established, its value will significantly increase. This means that once AI technology matures, visual data will become an extremely valuable resource.
Morgan Stanley believes that this demand for data will encompass all aspects of life, including warehouses, factories, hospitals, schools, stores, vehicles, airports, mines, homes, and forests, with even biological visual data becoming an important resource.
The report specifically mentions Tesla's key role in this process, believing that the company is at the "center of the Cambrian explosion of technology," and its investment story has far exceeded the realm of electric vehicles. Musk's vast business empire under "DREAMS" (Data, Robotics, Energy, AI, Manufacturing, and Space) will drive further value release for Tesla.
The report states that as Musk makes further progress in building reasoning groups (cars, robots) and AI infrastructure, general AI and LLM may become the "key" to unlocking Tesla's value as an AI giant.