How I found NVIDIA before it became obvious

Hi {{first_name | Investor}},

In 2017, I was a PC gamer and a Google employee — two things that had nothing obvious to do with investing.

But they gave me a lens most people didn't have. And that lens led me to a position that changed how I think about investing.

Here's what happened — and the framework I built from it.

As a gamer, I knew NVIDIA made the best GPU on the market. That wasn't an opinion — it was hardware reality. Around that same time, I started noticing something else: crypto miners were buying NVIDIA cards by the pallet. GPUs were selling out. Prices were climbing. Two completely different industries pulling on the same hardware.

Interesting. But that still wasn't the thesis.

The thesis was CUDA.

CUDA is NVIDIA's parallel computing platform — the software layer that lets developers write programs specifically designed to run on NVIDIA hardware. By 2017, it had quietly become the foundation that machine learning researchers and AI developers built everything on top of.

At Google, I worked on the Android team. I understood from the inside what it meant to own a developer ecosystem. Android was one of the most powerful software platforms in the world. But here's the thing: Google had the software. We didn't have the hardware.

NVIDIA had both.

The GPU was the hardware moat. CUDA was the software moat. And the two were reinforcing each other in a way that was going to be extraordinarily hard to displace. Developers had years of work built on CUDA. Switching platforms meant rebuilding from scratch. That's not a switching cost. That's a wall.

I bought NVIDIA in 2017. It eventually became a seven-figure position in my portfolio — not because I predicted how AI would explode, but because I understood what kind of company I was looking at. One that had built something genuinely hard to take away.

"The best moats aren't built in boardrooms. They're built one developer, one tool, one workflow at a time — until displacement requires dismantling an entire ecosystem."

That kind of thinking has a name. I call it the Wall Test. It's not about finding exciting companies. It's about finding companies that are structurally hard to displace.

Today I'm walking you through all four questions. By the end, you'll be able to run any company through it yourself.

Passing three gets a second look. Passing two or fewer is a no. Run this every time you consider adding a position — or reviewing one you already hold.

Question: What would it cost a well-funded competitor to take this company's customers?

If the honest answer is "not that much" — that's a red flag. Businesses without moats compete on price, and competing on price is a race to the bottom.

Moats come in a few distinct forms:

The most powerful moats aren't ones a company claims — they're ones you can observe in customer behavior. Do customers stay even when cheaper alternatives exist? Do prices go up without churn?

The strongest companies have two moats that reinforce each other. NVIDIA's GPU hardware made CUDA the obvious platform to build on. CUDA's dominance made NVIDIA the only serious choice for AI workloads.

Question: Is the business growing consistently, and does the growth make sense?

You're not looking for the fastest-growing company. You're looking for growth that's believable, sustainable, and internally consistent.

Where to find the numbers: revenue and earnings history lives in a company's annual report — the 10-K in the US, the 20-F for foreign companies listed on US exchanges. Both are free on the SEC's EDGAR database (sec.gov/edgar).

Question: Do the people running this company have skin in the game, and have they earned trust?

Management is the multiplier. A great business with bad leadership deteriorates. A good business with exceptional leadership compounds.

You don't need to love the CEO. You need to trust their judgment with your capital. Those aren't the same thing.

Question: Can this company raise prices without losing customers?

Pricing power is the clearest observable signal of a real moat. If a company raises prices and customers stay anyway, it means those customers have no better option.

NVIDIA is one of the more extreme examples. GPU prices surged as AI demand exploded. Cloud providers, AI labs, and enterprises paid anyway. There was no real alternative.

Company A — Enterprise SaaS | WATCHLIST

High switching costs. Revenue +22% YoY, consistent over four years. Founder still CEO, 8% insider ownership. Raised prices 15% last year; net revenue retention above 120%.

Passes all four tests.

Company B — Retail Tech | CUT

Low moat — competitors offer similar features at lower prices. Revenue +31% YoY, but down from +58% the prior year. Professional management, founder exited 2021, low insider ownership. No price increases in three years.

Passes one test. Fails three.

Company B's 31% growth looked attractive at first glance. The Wall Test surfaced why that number was misleading. Company A's 22% was less exciting on paper and more durable in practice.

"Growth gets your attention. Quality earns your capital."

Passing the filter doesn't mean you buy. It means the company earns a deeper look. The Wall Test handles the qualitative case — is this business real, durable, and well-run? The next step is running the numbers to make sure the financial statements actually back up the story. That's what we'll cover in a future newsletter.

For now: pick one company you already own or have been watching. Run it through the four questions. Write down your honest answers. That exercise alone will tell you more than most research reports.

Anyone can buy a stock. Not everyone can explain why the business wins.

If you can't answer that — the Wall Test already has.

Stay disciplined,

Koh