Nick Bostrom, Unity, and the market for simulated worlds

Once, Zhuang Zhou dreamed he was a butterfly, a butterfly flitting and fluttering about, happy with himself and doing as he pleased. He didn’t know that he was Zhuang Zhou. Suddenly he woke up and there he was, solid and unmistakable Zhuang Zhou. But he didn’t know if he was Zhuang Zhou who had dreamt he was a butterfly, or a butterfly dreaming that he was Zhuang Zhou.

That is the dream argument in a nutshell.

Which begs the question: Are we living in a simulation? According to philosopher Nick Bostrom, we might be. If there were a substantial chance that our civilization will ever get to the posthuman stage and run many ancestor‐simulations, Nick explains, then how come you are not living in such a simulation? Either we never make it to build simulated worlds, or we choose not to, or simulated minds within simulated worlds will so vastly outpace base-layer inhabitants that our odds of living in a dream inexorably approximate to 1.

The thing is, we are already building simulated worlds. To some extent, we might say that the history of humanity is the history of simulation. If we push Harari’s argument to the extreme, every facet of our culture, including the substrate that it runs on, is a figment of simulated worlds. “Homo sapiens rules the world, Harari says, because it is the only animal that can believe in things that exist purely in its own imagination, such as gods, states, money, and human rights.” Simulation is necessary for cooperation. “You could never convince a monkey to give you a banana by promising him limitless bananas after death, in monkey heaven.”

Now, as we approach the Gentle Singularity, simulation has become so sophisticated, so economically meaningful that it will soon outgrow the value of base reality. How big is the market for simulated worlds? If we count just movies and video games — as if we were still living in the 90s — the number appears comfortably small, but the closer we get to AGI, the larger and more nonsensical it becomes.

Or, is it? As humanity moved from speech to text to multimedia, so is AI moving from large language models, with text at its cornerstone, to world models, where fiction is no longer a story you simply tell to someone else: it turns into something you can live in.

Just a few days ago, World Labs, a spatial intelligence company by Fei-Fei Li, the so-called godmother of AI, launched Marble, a text-to-world application that turns words into 360° explorable environments. Genie 3, a maybe more powerful world model by Deepmind, “can generate dynamic worlds that you can navigate in real time at 24 frames per second, retaining consistency for a few minutes at a resolution of 720p.”

Marble, Genie 3 are the real deal: far from smoke and mirrors, they take us downstairs—from base reality down to its many simulated layers.

How many layers can there be? The answer to this question is a puzzle. Right now, the scope of the simulation can still be measured. But as we enter the second half of the chessboard, we are to lose the pathway out of the maze. Gradually, then suddenly, simulated worlds will dominate the Light Cone: VR; AR; LLM-powered video games; digital twins and industrial metaverses, where androids dream electric sheep; autonomous vehicles, spatial intelligence, physical AI; humanoids in charge of almost all of labor; holodecks and brain computer interfaces; wetware; edge AI; agents and their own training environments; world models; entire supply chains rehearsed in silico before a single atom moves.

As intelligence becomes too cheap to meter, not simulating will be ridiculous. From tokens to pixels to actions and consequences, everything will be simulated. No physical asset will be deployed, no single penny will change hands, if first we didn’t simulate it. This is how the pipeline flips: the simulation morphs into the substrate as reality becomes the commit stage.

Unity is a game engine. Together with Unreal Engine by Epic Games, Unity is the dominant player in the video game industry. Made with Unity games and apps now reach 3.6 billion devices worldwide, with billions of installs every month, and developers can ship to more than 20 platforms from a single codebase, including iOS, Android, PC, consoles, and all major VR headsets. Around 70 percent of the top 1,000 mobile titles, more than a quarter of the top PC games, and over 70 percent of the best selling VR games are built with Unity, and a community of millions of creators opens the editor every month to push new worlds into existence.

But, as they say, the devil is in the details. Unity is not a game engine. Unity is a simulation engine. The same stack that powers games is already used to build digital twins of factories and cities for Hyundai, airports, and defense bases; to prototype and test robots through Unity Robotics Hub and ROS, where arms and mobile platforms learn grasping, navigation, and manipulation before touching the real world; to generate synthetic datasets for computer vision with the Perception package, producing perfectly labeled images for object detection, segmentation, and pose estimation at scale; to train RL agents and embodied policies with ML agents in 3D environments where they learn to walk, drive, and explore; to create interactive training simulators for surgeons, mechanics, and plant operators; to design AR and VR workflows for field technicians, warehouse pickers, and remote assistance; to power automotive HMIs, cockpit visualizations, and UX prototypes for self-driving stacks; to visualize IoT streams, sensor fusion, and fleet operations in industrial metaverses; and to act as the front end for AI-driven analytics and control systems where models see, decide, and act inside a live 3D representation of base reality.

At this point, the question is not ‘how can Unity get a bigger slice of the gaming pie’ or ‘how can Unity ground-and-pound Applovin in the ads game.’ Like Search allowed Google to fund Other Bets that today are starting to pay out —Waymo, YouTube, Google Cloud, TPUs, Android, Transformers etc. — gaming and advertising give Unity the monopoly money it needs to transform into an AI-first simulation engine.

If simulated-world GDP grows larger than physical GDP, if every factory, city, warehouse, battlefield, environment, machine has a shadow twin where policies, agents and robots are stressed before they ever touch reality, then profit shifts to the engine that powers the simulation.

Most of the inhabitants will not be humans. They will be agents like Deepmind’s Sima 2, machines, pieces of software, fleets of humanoids and vehicles, all training and coordinating in closed loops. World models like Marble and Genie 3 generate vast implicit environments and pretraining curricula. They are powerful components, but the real value accrues above, at the application layer.

Marble, Genie 3—they set the stage where the magic happens, but they do not spell the incantation. They lack a general-purpose, cross-platform runtime that can serve as the system of record for how thousands of agents, robots and humans interact; they lack an editor and asset pipeline that millions already know; they lack deployment to billions of endpoints; they lack the unglamorous product infrastructure for input and UI, networking, replay, access control, analytics, patching, and everything else you need to turn a simulation into something the entire economy can live in.

Unity is the engine of creation. It is the general-purpose runtime where simulated worlds become products and systems of record. It already has the editor, the scene graph, the input and UI stack, the networking, the asset pipeline, and a deployment story that reaches 3.6 billion devices and more than 20 platforms. It already ingests CAD, BIM and ROS, already drives smart factories, airports, bases, medical and industrial training sims. Unity Studio pushes this further by moving the authoring layer into the browser and putting it in the hands of process engineers, city planners, logistics managers and AI researchers, who can pull in CAD models, Omniverse USD scenes or, in time, layouts born from world models, then wire agents, robots and data feeds into live applications without ever touching engine code.

Stripe wants to increase the GDP of the internet. Unity’s logical endgame is to increase the GDP of simulated worlds. The pot of gold at the end of the rainbow is in the trillions. That is the scale of the Unity wager.