I am completely unpacking every aspect of the Physical AI thematic, and building out my portfolio in real time.

I believe Physical AI can become one of the major investment themes of the next several years.

But to invest in it properly, we have to build the stack layer by layer.

I want to understand where the dollars can accrue, which parts of the stack are already monetizing, which parts are still early, and which companies have enough exposure for the theme to actually move the stock.

The first report in this series unpacked Layer 1: Training and Model development.

That layer was about helping the model learn before the machine ever moves.

Now we move to Layer 2: Simulation and Synthetic Data

Think of it like Layer 1 teaches the model and Layer 2 gives it a world to practice in.

There is naturally overlap here, as there will be throughout this series. But this is a different angle.

Before we deploy our Physical AI systems, can we test this first?

The real world is expensive and messy. Humans are messy (no offense). You’ve got unpredictable weather and light changes. You’ve got safety rules. Etc. etc. Real-world learning alone is too slow and too expensive for many Physical AI use cases. So Physical AI needs digital practice.

Here we are talking about things such as simulated environments, synthetic data, digital twins, virtual commissioning etc. Tools that let engineers, operators, and AI agents test decisions before they touch the physical system.

We got a good bit of signal from NVIDIA around this.

This Cosmos 3 announcement frames the new model around physical AI reasoning, world simulation, action generation, synthetic data generation, and policy model development for physical AI systems.

NVIDIA also describes Omniverse as libraries and microservices for industrial digital twins and physical AI simulation applications. Basically, the AI industry is moving from text and software into world understanding, and physical systems need a world where they can be trained, tested, and validated before deployment.

After going through the research, earnings calls, product pages, and public-market screen, I landed on three companies that each give a different version of this layer.

One is a quality infrastructure digital twin and engineering simulation business.

One is a small-cap semiconductor simulation specialist.

One is a high-torque operational digital twin name with solid proof points.

The public-market universe here is thinner than I expected. Many of the best simulation assets have already been acquired. Synopsys completed its acquisition of Ansys in July 2025. Siemens completed its acquisition of Altair in March 2025. Emerson completed its acquisition of the remaining AspenTech shares in March 2025.

So I have to accept some imperfection and be precise about what kind of exposure each company actually gives us. I also might have to come to the conclusion that this might not be an investible. But we will dig into that more in this report.

Below the paywall I walk through all three companies in full: the product fit, the financial profile, what I like, what I need to watch, and how I am ranking them for deeper research.

Read more

Now I want to make that process really easy to follow.

This page will become the live home for my Physical AI model portfolio.

The goal is to show how I am personally building exposure to the theme over time:

• which companies I am tracking

• which companies I am adding

• how I am sizing them

• where they fit in the stack

• how my 12-month and 24-month frameworks change as the facts change.

This will be a multi-quarter and multi-year research tracker.

I am building this slowly. I want the portfolio to reflect actual research, actual conviction, and actual risk management rather than just filling a table with names.

The first version starts with two names:

One existing full position.

One target position I want to build.

The rest remains in cash while I continue working through the Physical AI stack.

For paid subscribers, the full tracker is below.

Disclaimer:

This document is educational commentary and reflects my personal research process, opinions, portfolio construction, and interpretation of public information. It is general market commentary, not individualized financial advice, tax advice, legal advice, or a recommendation to buy, sell, hold, or short any security.

I am not a financial advisor. This is not financial advice. Please conduct your own research.

Read more

This is the first company I want to build into that portfolio.

I’m starting with the learning layer.

If you haven’t read my introduction yet here, read this:

That may sound less exciting at first, but I think it’s the right place to begin. The robot is the visual. The intelligence behind the robot is the value. Before any physical system can operate in the real world, it has to learn what the world looks like, how objects move, what good outcomes look like, what mistakes look like, and what action should come next.

Physical AI is about the machine and the full loop behind the machine. See the world. Understand the task. Decide what to do. Act. Measure what happened. Improve.

That loop needs data, human demonstrations, synthetic examples, expert feedback, evaluation, safety testing, monitoring, and constant improvement. I already laid this out in the Layer 1 report: Physical AI starts before the machine moves, because a physical system has to understand space, motion, timing, force, safety, and consequences.

That’s where I want my first position.

The company I’m writing about today is already being paid by major AI customers for data engineering, model evaluation, trust and safety, expert feedback, and agentic AI workflows. The pure Physical AI revenue piece is still early, but the company sits in exactly the layer I want exposure to first.

It’s a company getting paid today from the AI learning cycle, with a credible path into Physical AI as AI moves from digital tasks into real-world action.

The goal here is to build the Physical AI portfolio with discipline.

I am building this from scratch. I am going to create a spreadsheet with all the positions I take over the coming weeks and months, weightings, price targets, and more. This is going to be a real reflection of my own personal portfolio that I will be building out. I am going to explain this in full in a standalone article this weekend, but just know this is a really exciting update to the platform.

Read more

I was fortunate enough to be positioned in the optics super cycle before most and it has been life changing.

What if we can find the next theme of this magnitude and position ourselves early?

I believe we can.

And that is what this report is about.

After really digging into this one, I believe this is it.

So in this report I am going to lay out the theme, why it’s investible now, where it’s investible, all the layers + supply chain, and what my plan is going forward.

Optics / Photonics is still my main theme that I will be covering in depth on this page.

But there is opportunity out there just waiting for us.

And this theme I am going to lay out here is going to be the 2nd main coverage universe on my Substack over time.

We are still early and we have time to position ourselves.

Let’s figure out the best way to do that.

Read more