Where androids dream electric sheep
And why Unity is their dream machine.
I like non consensus plays. After all, why pick stocks if not to beat the S&P 500? Sure, you could go the Joseph Carlson way and pick the best compounding machines the market has to offer, and rip the rewards. Or, you could go after momentum plays and chase the OKLOs and the Palantirs of the world, hoping the time bomb doesn’t explode right while you are holding it. Unity is neither of those. You are a lion running after a gazelle. You don’t follow the gazelle in its footsteps. You try to triangulate it and be there before it is. Unity is the gazelle. Right now, you can’t see it being valued at one trillion dollars, but that’s just because you’re following the market in its footsteps. Instead, use your imagination. After all, what is beating the stock market if not an exercise in imagination? You need to picture the future well before it takes shape. And in the future, 5, maybe 7 years from now, Unity is worth one trillion dollars.
This is why.
#1 Unity is hated - very much so
Nobody likes Unity. And why should anyone? The company is losing money hand over fist, and even tried to scam game developers with the runtime fee thing back in 2023. Which makes for a perfect setup for a moonshot type of return. Think about it. Moonshots are made by sifting through shit in the gutter. No one is gonna give you the perfect stock on a silver platter. It’s gonna be ugly, and you ain’t gonna like it. Needless to say, all that glitters is not gold, and sometimes, often times, almost always in fact, shit is just shit. So what makes Unity not shit?
First of all, management has been upended. Since Matt Bromberg became CEO in 2024, interim CEO Jim Whitehurst has moved to executive chair of the board, and Unity has fully killed the runtime fee, reverting to a simpler, seat-based subscription model. Workforce has been reduced by 25%, and the company has exited film/VFX bets like the Wētā professional services piece. A new CFO has been appointed (Jarrod Yahes (ex-Shutterstock)). The CTO has resigned, which is not good news per se, but we’ll see what the future holds. And, most importantly, Vector (more on this later).
Basically, it’s a whole new company. Sure, you could say that “that which we call a rose, by any other name would smell as sweet”. And to some extent, you would be right. Not everything has changed. In fact, what was good about Unity hasn’t changed. That is…
#2 Unity owns gaming (mostly of, at least)
Well, this might be a stretch. Unity has competitors: Unreal Engine in its Create segment, and AppLovin in its Grow segment. But it still owns 70% of mobile gaming and 51% of gaming on Steam, which makes investing in Unity right now the same type of bet you could do by investing in Netflix in 2022, when hating Netflix was popular. My point is, you don’t wanna be the new Bill Ackman and sell Unity at 45 dollars. Unity is the Netflix of gaming, kinda. It’s got a quasi-monopoly in the sector and why should you ever bet against a monopoly?
#3 Unity is a turnaround play
As I said, Unity is hated, which means it’s most likely mispriced. Remember Peter Lynch: people always misprice hated stocks. Since nobody likes them, nobody takes the time to price them correctly. You can do it. And not because you’re a phd in maths or anything like that. You don’t need to work at the Medallion Fund to recognize an opportunity. Temperament, like Buffet says, is the most important quality of a stock picker. It’s guts. It’s like what they say about streets and blood, the time to buy is when there’s blood in the streets. Right now, there is. The street is called Unity. And you know what’s good about this street? They’re cleaning it.
Not only has Bromberg gone back on the runtime fee shit, it also has admitted the company’s mistakes, which, if you ask me, makes for a good leader, and promised no further shenanigans in monetization schemes. Since Bromberg is at the helm, the company is “no longer at war with its customers” and has been shipping consistently. Unity 6, with mixed-to-improving sentiment from developers. And Vector, Unity’s new AI ad platform, which let me tell you, is a major deal: it’s boosted the ad network revenue 15% sequentially, with guidance for continued double-digit growth in Q3 - all while bringing 15-20% growth in the number of app installations and in-app purchase value. This is a turnaround in the making.
#4 Unity is a secret AI play
So I just told you why the company is a good investment: everyone hates its guts, it’s a quasi-monopoly, and it’s a turnaround play. What I didn’t tell you is, Unity is a secret AI play. Vector is just the start (now, bear with me, because this requires some imagination).
Everyone knows Unity as a gaming company, and as you know from Nvidia, gaming is AI’s secret cousin. Indeed, the largest pie of Unity’s future earnings won’t come from advertising, and neither from gaming: it will come from physical AI.
Physical AI - also called embodied AI - is a big pie. We’re talking high double-digit CAGR and a TAM as large as the entire physical labor market. There are several players in the space, the main ones being Tesla with Optimus, Figure with Figure, Apptronik with Apollo, 1X with Neo Gamma, and the Chinese company Unitree with a slew of humanoid and non-humanoid robots.
What is the opportunity? Perception. Synthetic data. Let me explain. Humanoids, like LLMs, need to be trained on massive datasets. Which begs the question: where do you get a big enough dataset to train these robots on common everyday tasks? I mean, there’s not plenty of POV videos of people doing the laundry on YouTube. But that hasn’t stopped Figure AI to train its humanoid to do just that. How did they pull it off? Human video collection. Which means: they used actual data coming from the real world.
Recently Figure CEO Brett Adcock announced Project Go-Big, the company's "push to build the world's largest humanoid pretraining dataset. We are scaling out human video collection and our F.02 humanoid can now learn directly from human video. This project is accelerated by our partnership with Brookfield, who owns over 100,000 residential units."
How scalable is this? Not much. First, usable videos of humans doing shit in first person is a scarce resource. Second, even if it were as big as the internet, it wouldn’t be enough. There’s been some chatter earlier this year about LLMs needing synthetic data after eating all the internet. The same thing will happen with robots. Unfortunately, LLMs and VLAMs (Vision-Language-Action-Models) are not yet able to generalize the concept of a dog after seeing a dog once, like humans do. I mean, my 4-yo nephew doesn't need to watch millions of dogs to understand what a dog is, but AIs do. They’re artificial intelligences, not artificial GENERAL intelligences, so to speak. My point is, it is OBVIOUS they’ll need synthetic data. Think about how big the human experience is. We can do millions of physical tasks, from petting a dog to folding clothes, and on and on. You can’t reasonably expect this humongous dataset to come from people labelling real-world videos. There aren’t enough videos to begin with.
And where will this treasure trove of synthetic data come from?
Digital twins - this is what you need. You need someone to build a virtual environment where digitized humanoids can roam freely and collect data about the world. This is the story of AlphaGo Zero. While tackling the challenge of building the world's strongest AI Go player, the genius people at Google's DeepMind taught their AI all the rules of Go and trained it with a dataset of expert human games. Then, they had a breakthrough. Why stop at that? Why not train the AI against itself? That's what they did. In October 2017, DeepMind unleashed AlphaGo Zero upon the world, a version created without using data from human games - without even teaching it the rules of Go! By playing only against itself, AlphaGo Zero was able to beat AlphaGo Lee in three days by winning 100 to 0, reached the level of AlphaGo Master in 21 days, and exceeded all previous versions in 40 days.
There is a lesson to be learned from this parable: to go superhuman you need synthetic data. Enters Unity. Unity is used to build interactive, physics-based environments where you can (a) simulate robots and sensors, (b) connect real robot stacks via ROS, and (c) generate data to train/validate physical AI before deployment. In short, Unity ships and maintains the core building blocks used in humanoid-training twins:
Unity Robotics Hub (open-source tutorials and packages for ROS, URDF, pick-and-place, pose estimation, etc.)
ROS-TCP Connector
Perception (Unity’s official synthetic-data toolkit)
Sensors/SystemGraph and docs for LiDAR/depth/IMU and authoring
Basically, Unity controls the entire software stack for humanoid training. And that's not all. Unity's digital twins deployments don't stop at humanoids. They're used at
Hyundai Motor, for building a real-time twin of its plants
Vancouver International Airport for staffing, passenger-flow, maintenance and decarbonization planning
Orlando Regional, which includes a 40-sq-mile high-fidelity core, used for coordinating agencies and exploring infrastructure scenarios
U.S. Air Force (Tyndall AFB) - we’re talking about a “largest-of-its-kind” twin used for design, resilience, day-to-day planning, and sustainment after Hurricane Michael
Sitowise, which built a Unity-based Hong Kong 5G radio-propagation simulator for planners, and earlier a district-scale city twin (Hervanta) to test urban scenarios
Bell Textron, where they built a program/turbine mockups & facility twin for design review.
Prespective (Unit040), where they built a Unity-based industrial simulation suite used by OEMs to prototype workcells, run throughput studies, and de-risk commissioning
What these names tell us is the massive opportunity at the heart of Unity's digital twin endeavor. Not just humanoids: live IoT/IT data in a 3D ops view; unified “single pane” for mobility, 5G, zoning, climate resilience, and more. The metaverse will get very real very fast, although it won’t resemble anything of what Zuck envisioned. It won’t be a B2C metaverse. At first, the metaverse will be B2B, and Unity will be its dream machine.
Disclaimer:
a) At the time of publishing this article, I own shares in Unity Software Inc. (U).
b) This article should not be misconstrued as financial advice. Do your own research.
c) I put a lot of effort into this article. If you like it, please help me grow my publication by subscribing: it’s free. I publish weekly, bringing you a glimpse of the extropian future, and the ideas and markets wiring it all together. And if you also decide that the article is worth sharing with someone in your life, that’d mean the world to me.