Tesla, Meta, and Figure – a "photon battle" is underway

The field of artificial intelligence robots is undergoing an unprecedented "photon competition," with major tech giants frantically collecting visual data from the real world to train AI robots.

According to Hard AI, Morgan Stanley stated in its latest research report that with the development of AI robots and embodied artificial intelligence, companies like Tesla, Meta, and Figure AI are massively collecting visual data to train Visual Language Action (VLA) models.

Specifically, Tesla is shifting to a "pure vision" training method, Meta is collecting data on daily activities through smart glasses, while Brookfield is collaborating with Figure AI to deploy data collection across a vast real estate portfolio.

This trend means for investors that visual data has become a new "gold mine" for AI training, and companies with data collection capabilities will hold an advantageous position in the AI robot race.

Morgan Stanley used the metaphor of "fat tuna" to explain the value of visual data: A 612-pound bluefin tuna sold for $3.1 million at an auction in Tokyo in 2019, but without fishing tools, the value of that fish is zero. Similarly, without processing capability (yottaflops-level computing power, 1 yottaflop = 1 trillion teraflops), the world's visual data also holds no value. However, once the ability to collect and process is established, this data becomes extremely valuable.

Tesla's Strategic Shift: From Remote Control to Pure Vision Training

Morgan Stanley stated that Tesla is undergoing a significant strategic shift in the training of its Optimus robots.

According to Business Insider, internal sources at Tesla revealed that the company has shifted the training of the Optimus robots to a "pure vision" approach, abandoning traditional remote control, motion capture suits, and VR technology, and instead recording videos of workers performing tasks as training data.

In May 2025, the former head of Optimus at Tesla released a series of video clips on the X platform, showcasing Optimus performing tasks that it allegedly learned from human videos. These videos were initially shot from a first-person perspective (with the camera mounted on a human demonstrator), but the ultimate goal is to expand to third-person perspectives captured by "random cameras" and content from the internet.

This strategic shift highlights the core value of visual data in AI robot training. As stated in the Morgan Stanley report: "When you drive a Tesla, you are not just moving through physical space; you are also playing a video game... feeding data into a simulated world to train Tesla's latest FSD model."

Meta's Smart Glasses: Transforming Daily Life into Training Data

Morgan Stanley's internet team believes that while Meta's wearable devices are "long-term bullish options" and unlikely to impact financial data in the coming years, their strategic significance cannot be underestimated. Meta is advancing its long-standing vision of integrating leading large models and agent capabilities into the next generation of wearable devices.**

Morgan Stanley report points out:

When you wear Meta glasses, you are teaching the model how to play the piano, knit, pour coffee, or take out the trash.

Imagine if 20 million such devices are put into operation within 2 years—this is almost twice the number of Tesla vehicles on the road—each Meta glasses user could potentially train a humanoid avatar in the metaverse that iterates across billions of scenarios.

Brookfield and Figure AI: The Data Collection Network of the Real Estate Empire

Morgan Stanley's alternative investment team views Brookfield as a leader in executing large-scale AI infrastructure solutions. The partnership between Brookfield and Figure AI is seen as an important step in creating expertise in the rapidly evolving field of humanoid robotics.

Brookfield's vast global footprint makes it a unique partner to help Figure AI build the largest pre-trained dataset. Brookfield is one of the largest real estate owners, with over 100,000 residential units, more than 500 million square feet of commercial office space, and 160 million square feet of logistics office space.

This collaboration will allow Figure AI to accumulate critical AI training data to teach humanoid robots how to move, perceive, and act in various human-centered spaces. Data collection efforts have already begun in Brookfield environments, and the project is expected to scale up in the coming months.