AI-generated summary: After Argentina’s loss in the World Cup final, Messi placed another bet—his investment platform, Play Time, invested in World Labs, founded by Fei-Fei Li. The company raised $230 million in its founding year with a valuation exceeding $1 billion; its first product, Marble, is now live, converting images and text into editable 3D scenes, with applications spanning gaming, film and television, architectural design, and robotics training. It has attracted investment from multiple institutions and currently holds a valuation of approximately $5 billion. Giants like Google and NVIDIA are also entering this space, making the spatial intelligence sector highly competitive—Messi’s investment remains unproven.

Article author and source: Blue Letter Project

Messi's next match has just begun.

Messi hasn't lost yet.

This World Cup final, 39-year-old Messi failed to lift the World Cup trophy again; but immediately after the tournament ended, media reported that Play Time, one of its investment platforms, invested in World Labs, a company founded by Fei-Fei Li that uses AI to generate 3D worlds at scale.

For top athletes, using earnings from the field to make more money is nothing new.

LeBron James, widely regarded as the second-greatest in NBA history with aspirations for the top spot, built his business empire early through equity investments, sports assets, and a film production company; another football superstar, Cristiano Ronaldo, has expanded his business ventures into hotels, fitness, and healthcare; and tennis legend Serena Williams has even founded a dedicated venture capital firm.

In comparison, Messi started much later in the venture capital field.

It wasn't until 2022 that he co-founded Play Time with partners to begin investing in sports, media, and technology companies. Although he has made several investments over time, none have yet emerged as a standout success that clearly demonstrates his investment acumen.

Moreover, the company Messi bet on this time, World Labs, is not a mature company in the AI circle that can reliably turn a profit.

The company was valued at over $1 billion in its founding year, before even having a product; now, with its first product just launched, reports indicate its valuation has reached approximately $5 billion.

On one side is a football legend whose investment performance has yet to be proven; on the other is an AI company whose product has just launched and whose valuation has already soared to billions of dollars—it’s hard not to worry for Messi.

Is Messi’s move on the field this time a good deal?

Messi voted for another GOAT

GOAT stands for "Greatest of All Time," meaning the best of all time.

If Messi is the GOAT of football, then Li Fei-Fei, whom he just invested in, is roughly of the same stature in the field of AI vision.

Li Fei-Fei's most significant contribution was leading her team to create ImageNet.

In the late 2000s, researchers had begun teaching computers to identify cats, dogs, cars, and pedestrians, but they had too few training images, and each team used different datasets, making it difficult to determine which algorithm was stronger.

Li Fei-Fei decided to first create a unified set of visual questions for AI worldwide.

She led her team in collecting images from the internet and enlisted tens of millions of ordinary people to label them. Ultimately, ImageNet included over 14 million images spanning more than 20,000 categories.

It was precisely this thing, which sounds like "organizing images," that changed the course of AI development.

In 2012, AlexNet entered the ImageNet image recognition challenge and achieved a classification error rate more than 10 percentage points lower than the second-place entry. This result was the first to clearly demonstrate to the industry that, with sufficient data combined with neural networks and GPUs, machine image recognition capabilities could undergo a qualitative leap.

Since then, an increasing number of researchers have turned to deep learning, accelerating advancements in technologies such as facial recognition, autonomous driving, and AI-generated images.

ImageNet is therefore regarded as one of the key starting points of the modern AI wave, and Li Fei-Fei has become an indispensable figure in the field of computer vision. Stanford University once entrusted her with leading its AI laboratory, and Google Cloud also invited her to serve as Chief Scientist for AI.

She previously taught AI to “see.” Now, Li Fei-Fei wants AI to further understand the world it sees.

In 2023, at a robotics forum held at Stanford University, she stated that while ChatGPT demonstrated a breakthrough in linguistic intelligence, AI needs more than just understanding text to enter the real world.

Humans can quickly determine where furniture like tables and chairs are placed, whether a door should be pushed or pulled, and where they can move through a room. Even if robots recognize these objects, when placed in an unfamiliar environment, they may need to re-perceive, re-plan, or even re-train.

Li Fei-Fei refers to the ability to understand objects, space, and the relationships between people and their environment as spatial intelligence. In 2024, she co-founded World Labs with three experts in computer vision and graphics to equip AI with this capability.

In its founding year, the company secured $230 million in funding. Leading investors included top Silicon Valley venture capital firms a16z, NEA, and Radical Ventures, with investment arms of NVIDIA, AMD, and Intel also participating.

By 2026, NVIDIA and AMD continued to invest, and Autodesk, one of the world’s largest design software companies, made a one-time investment of $200 million.

Based on the founder’s background and investment lineup, Messi has at least aligned himself with the right people; but being with the right people doesn’t necessarily mean investing in a good business.

AI version of "Minecraft"?

Determining whether an investment is worthwhile is only the first step—choosing the right person. However, World Labs must still prove that Li Fei-Fei’s spatial intelligence can be turned into a product people are willing to pay for.

At the end of 2024, World Labs publicly demonstrated its technology for the first time. Users simply upload a regular image, and the system transforms it into a freely explorable 3D scene.

After processing, a photo of a living room allows users to walk around the sofa and see furniture that was originally blocked; a city street scene can also be explored from different angles to discover buildings and roads.

In 2025, World Labs officially launched its first product, Marble. In addition to images, it can generate 3D worlds from text, video, and simple 3D layouts, allowing users to continue editing, expanding, and combining different scenes.

In simple terms, Marble is like an AI-powered “simulation engine”: give it a sentence or a photo, and it will instantly generate a fully immersive 3D world you can explore and keep modifying.

Currently, Marble is most easily adopted in industries such as gaming, film and television, architectural design, and virtual reality.

Maps and buildings in games, virtual scenes in movies, spatial designs presented by architects, and digital models of cars and industrial products all require substantial 3D content. In the past, these assets were often painstakingly built, textured, and lit by professionals, resulting in lengthy production cycles and high costs.

If Marble could turn a picture directly into a modifiable 3D world, what it would sell first is one of the most expensive commodities in these industries: production time.

Autodesk is willing to invest $200 million upfront in World Labs, which is also tied to this real-world demand. Its customers span the architecture, manufacturing, film, and gaming industries; once World Labs’ models are integrated into these professional workflows, spatial intelligence could become a directly monetizable tool.

Greater potential comes from robotics.

In July 2026, World Labs acquired the robotics simulation platform SceniX to enhance its physical simulation and robotics training capabilities. It aims to create virtual worlds that not only look realistic but also obey the laws of gravity, collision, and motion.

This allows robots to practice grasping, moving, and obstacle avoidance in AI-generated factories and homes, while autonomous driving systems can beforehand experience extreme scenarios such as heavy rain, construction zones, and pedestrians suddenly stepping onto the road.

Once this path is validated, World Labs will be able to expand its offerings from a 3D creation tool to the environments, simulators, and data needed for robot training.

However, this potential is still largely confined to the future.

Marble can generate a virtual house, but it still cannot directly make a robot enter a real house and confidently pick up a cup from the table. There is still a long way to go from drawing a 3D world, to accurately simulating physical laws, and finally to controlling a robot to complete tasks.

Meanwhile, Google DeepMind has launched the Genie series of models capable of generating interactive environments, and NVIDIA is building the digital worlds needed for robot training through Omniverse and Cosmos.

Google owns models and data, while NVIDIA controls computing power, chips, and the robotics ecosystem. Their entry signals that spatial intelligence is worth betting on, and it means World Labs must compete in the market against giants with far greater resources.

If spatial intelligence becomes the next major AI infrastructure following large language models, and World Labs emerges as a leading platform in this space, its current valuation may still have room to grow; but if Marble ultimately amounts to nothing more than a more advanced 3D generation tool, the company’s prematurely inflated expectations may be difficult to realize.

The World Cup has already decided its winner, but Messi’s match on AI has just begun.