At a world models panel at the All In conference, I stepped into one of AI’s most mysterious corners. The leading names are Yann LeCun’s AMI Labs and Fei-Fei Li’s World Labs. Both have buzz and funding, yet they are not chasing revenue. World models promise automated spatial intelligence, spanning robotics, interactive video, and more complex self-driving. But where does it turn into business? The harder you press, the thicker the fog and the more careful the silence.
What are world models, and why do they matter now?
World models aim to automate spatial intelligence and make it practical. The idea points toward robotics, interactive video, and advanced self-driving systems. Players like AMI Labs and World Labs fuel expectations, yet they show little focus on monetization. That raises the central question: the potential is clear, but where is the real market?
At the core is a model’s ability to understand environments and predict actions in space. The simplest mental picture is a navigable map, similar to approaches behind self-driving. The same principles can transfer across tasks. That range of possibilities attracts investors and teams alike.
This breadth keeps promises high. World models can treat video as an environment, not just media. They also aim at actions in the physical world, where prediction and control matter. Yet the same breadth makes choosing an initial commercial niche harder.
One thing is obvious: demand for these abilities will exist. Still, a fast push into a product needs tight focus. That focus is barely visible right now. The field hovers between capability showcases and careful quiet about plans.
Where will commercialization show up, and why is the answer so vague?
There is no clear answer yet, and that shows even in public forums. When you ask for specifics, companies clam up. The closest thing to an “official line” came from AMI Labs co-founder and VP of World Models Michael Rabbat — and even he was highly cautious.
On stage, the response was guarded: they would talk when ready. Later, he clarified over email that AMI is in a research and building phase. So the company is not discussing public product plans or timelines. For a team less than a year old, that stance makes sense.
“We’ll talk about it when we’re ready to talk about it.” And later by email: “We’re still in a research and building phase, so we’re not talking publicly about any product plans or timeline.”
Still, the caution is not limited to one company. The “say less, buy more time” mood pervades the entire world-modeling space. The market is only forming, and early publicity raises both stakes and risks.
Companies avoid premature announcements to avoid spotlighting the path. Early declarations draw competitor attention and change the race’s pace. As long as capital is available, waiting becomes a rational tactic.
What products are visible, and what do they really teach us?
World Labs’ Marble is likely the most developed product on display, yet it mainly shows, not sells. Demos range from media creation to explorable game environments and CGI effects. There are robotics use cases too, but the platform seems geared to demonstrate capabilities rather than validate a business model.
This makes sense at the validation stage. Teams want to prove “can do” before narrowing to “must do.” Within that logic, Marble works as a showcase for engineering feats. It shapes the language the field uses to discuss spatial modeling.
However, demos do not answer questions about price, service model, or scale. They rarely reveal the needs of specific segments. Even a strong showcase is not the same as a validated revenue model. That is why the product path still looks uncertain.
While the field collects examples, business logic waits its turn. Practical niches exist, but they need verification through usage data and real contracts. Until then, Marble and similar projects remain primarily class demonstrators.
Why does secrecy reach even the data suppliers?
The category’s closed nature is felt even by its suppliers. Physicl provides data to the growing world model business, yet gets little visibility into end goals. They know their datasets help, but they remain in the dark about exactly how.
Physicl CEO Alex de Vigan speaks directly about the information gap. He believes clearer plans would help them make more useful data. Yet transparency is still scarce. This is a classic early-stage symptom, with companies guarding their idea vectors.
“I wish they would tell us more. We could build more useful data if we knew what they were working on,” de Vigan told me.
The paradox is that suppliers also optimize costs and quality. Without clarity, it is hard to tailor collection or structure for target tasks. Secrecy at the top turns into inefficiency below. Still, players accept the tradeoff to avoid revealing too much.
It underscores how sensitive product-strategy information is. Even neutral partners receive only the “need to know.” The market is being built with limited access to plans, and that is a deliberate choice by its leaders.
How flexible is the world-model idea, and how does that complicate focus?
The idea’s versatility is striking, and that is exactly what tempts teams. The simplest use is a navigable map, like models used for self-driving. The same approach could help a humanoid carry boxes or turn minutes of video into an explorable space. Examples port easily across domains.
AMI has already dipped its toe into several pools: manufacturing, biomedicine, robotics, and even AI software for doctors through its Nabia partnership. It seems clear not all can fit the first wave. The question remains: which one or two vectors are standing out?
No one doubts there are many viable businesses here. While fundraising is easy, there is little pressure to narrow focus. In fact, there is good reason not to rush. Premature specificity opens the door to competition when teams still want quiet and room to maneuver.
In that logic, demos are a map of possibilities, not a market contract. Companies watch reactions and collect signals without naming the endgame. Versatility lets them slow any lock-in to a vertical until a clear winner emerges.
The dark forest as strategy: why do leaders stay quiet?
Public specificity can instantly reshape the game. Imagine AMI announcing a humanoid OpenClaw or a next-generation Hollywood rendering system tomorrow. Other labs would quickly focus on that route. Competition would emerge not just from fellow world-model companies but also from neolabs and even OpenAI and Anthropic.
Fundraising success has a flip side: competitors can fundraise too. The same money that lets you build under the radar also funds rivals once the path to market is clear. If competition is inevitable, it is best to delay it as long as possible. The simplest way is to stay quiet about build details.
This is the “dark forest” scenario familiar to Cixin Liu fans. If you do not know who else is in the woods, do not draw attention. Silence becomes a tool of strategic delay rather than a sign of weakness.
That stance also explains why even Marble mostly demonstrates rather than reveals market logic. Showcases build momentum and thought-leadership while keeping the final bet hidden. It is a compromise between displaying progress and avoiding a roadmap for competitors.
In the end, world models are a testing ground for focus and pacing. The current consensus is simple: the potential is broad, the market is maturing, and it is better to allow more time before naming the prize. In that frame, silence is not a barrier but a strategy for survival and positioning.
Based on the source text.
