World model companies are keeping a lot of secrets

World-modeling startups — the labs building AI that simulates environments — are raising easily but revealing almost nothing about what they're actually building, a strategy that is less about stealth than about delaying the moment when well-funded rivals realize the path to market is open.

Published: 2026-09-22 Category: Quick Take

The pattern

The world-modeling space — AI that learns to simulate or imagine environments — is flush with funding, but the companies inside it are conspicuously vague about what they ship. At a recent confab, AMI Labs co-founder Michael Rabbat dodged questions about the product, saying "We'll talk about it when we're ready to talk about it," and clarified by email that the startup is "still in a research and building phase, so we're not talking publicly about any product plans or timeline." AMI is less than a year old, so the quiet is partly fair — but the caginess spans the whole space.

Even the suppliers are in the dark

World Labs' Marble is probably the most developed product in the field, yet its demos read more as capability showcases — media creation, explorable game environments, CGI effects, robotics use cases — than a clear commercial anchor. The secrecy extends to suppliers: Alex de Vigan, CEO of Physicl, a data supplier for the world-model business, says he knows his data is being used but doesn't know for what. "I wish they would tell us more. We could build more useful data if we knew what they were working on," he said.

Why the silence

There are plenty of viable businesses to build on world-model tech, and as long as fundraising is easy there's little pressure to commit to one. There's an active incentive not to. If AMI announced tomorrow it had built a humanoid robot or a next-gen Hollywood rendering system, other labs would suddenly move into the space — the neolabs, other world-model firms, even OpenAI and Anthropic.

This is the flip side of easy money. The same capital that lets a company build under the radar also funds potential rivals the moment the path to market becomes obvious. Better, then, to stay quiet and delay that clock for as long as possible.

The sharp read

Secrecy here is not a sign of weakness; it's a rational response to a market where the moat is timing, not technology. The value of a world model is realized when you're the first to show a compelling use case — but declaring that use case is precisely what invites the competition that eats the margin. So the rational move is to keep the application undisclosed until it's defendable. The catch is that this dynamic also keeps customers, data partners and regulators guessing, and it makes every grand claim in the space unverifiable until someone finally blinks.

Source: TechCrunch