How do I apply the NVIDIA framework to find the next winner?

Apply the four-question screen: Does the company have an infrastructure monopoly? Is demand secular and multi-year? Are margins expanding, not just revenue? Is the stock undervalued relative to growth (PEG ratio below 1.5)?

Is it better to invest in a good company or a good stock?

The pattern identifies great businesses. Valuation determines great investments. A company can match the NVIDIA pattern perfectly and still be a bad investment if the stock price already reflects everything.

What is a PEG ratio and why does it matter?

PEG ratio is the price-to-earnings ratio divided by expected growth rate. Below 1.5, the market hasn't fully priced in growth. Above 2.0, the stock needs to beat expectations to justify the price.

Stock Selection

The Next NVIDIA: A Framework for Finding It

Q: What made NVIDIA's run identifiable?

Four traits defined NVIDIA's run: infrastructure monopoly position, secular demand driver, margin expansion runway, and management execution. All four had to be true simultaneously.

NVIDIA returned 22,000% in a decade. That run was identifiable early if you knew what to look for. This essay reverse-engineers the pattern and screens today's companies against it.

The Template

What Made NVIDIA's Run Identifiable

In 2016, NVIDIA was a $30 billion company making GPUs for gamers. By 2026, it's worth over $3 trillion, the standout among the Magnificent Seven. The stock returned roughly 22,000% in a decade. Most of that return came after 2022, when the AI training demand explosion hit.

Looking backward, four things made NVIDIA's run possible. All four had to be true simultaneously. Missing even one would have capped the outcome.

NVIDIA Revenue Trajectory

Annual revenue in billions, showing the inflection point

FY 2017

$6.9B

FY 2019

$11.7B

FY 2021

$16.7B

FY 2023

$27.0B

FY 2024

$60.9B

FY 2025

$130B+

Revenue doubled from FY2023 to FY2024, then doubled again to FY2025. That acceleration is what drove the stock price. But the acceleration was enabled by a decade of positioning that happened before the demand arrived.

The Four Criteria

The NVIDIA Pattern

Four traits defined NVIDIA's run. Together they form a screen you can apply to any company claiming to be "the next big AI play." For the quantitative companion to this qualitative screen, see the 10 numbers that tell you everything about an AI stock.

Infrastructure Monopoly Position

NVIDIA doesn't sell AI applications. It sells the tools everyone else needs to build AI. When every company needs GPUs to train models, NVIDIA sits at the tollbooth. The key was that alternatives were poor. AMD was years behind on software. Google's TPUs were internal only. Intel fumbled. An infrastructure monopoly means customers can't easily switch, and demand is driven by the entire industry, not just one use case.

Secular Demand Driver

AI training demand isn't cyclical. Every major tech company is building AI capabilities simultaneously. Microsoft, Google, Meta, Amazon, Apple, and hundreds of startups all need the same hardware. This is secular demandLong-term demand driven by fundamental shifts that persist across economic cycles.: it doesn't depend on one customer or one application. When the demand driver is broad-based and multi-year, the revenue growth sustains longer than analysts model.

Margin Expansion Runway

NVIDIA's gross margins expanded from 59% in FY2019 to 75% in FY2025. That's unusual. Most companies see margins compress as they scale. NVIDIA's margins expanded because demand exceeded supply, CUDA lock-in reduced price sensitivity, and each new chip generation (H100, B100, B200) commanded premium pricing. Margin expansion on growing revenue is a double engine for earnings growth.

Management Execution

Jensen Huang bet the company on AI before anyone else did. NVIDIA invested in CUDA (its GPU programming platform) starting in 2006, a full 16 years before the ChatGPT moment. When AI demand arrived, NVIDIA had the hardware, the software ecosystem, the developer community, and the manufacturing relationships. The window between "the trend is obvious" and "it's too late to build" is narrow. NVIDIA was already through it.

Infrastructure monopoly plus secular demand plus margin expansion plus execution. All four together produced a 22,000% return. Three out of four would have been a good investment. All four made it a generational one.

The Screen

Applying the Framework Today

Which companies today match the NVIDIA pattern? Here's a screen of five candidates, scored against the four criteria. This is a framework, not a buy list.

NVIDIA Pattern Scorecard

Current companies scored against the four criteria (Strong / Moderate / Weak)

Company	Infrastructure Monopoly	Secular Demand	Margin Expansion	Execution
TSMC	Strong. Fabricates 90%+ of advanced chips.	Strong. Every AI chip needs TSMC.	Moderate. Margins improving but capex-heavy.	Strong. Decades of execution under Morris Chang's framework.
ASML	Strong. Only maker of EUV lithography machines.	Strong. Every advanced fab needs EUV.	Strong. 52% gross margins, expanding.	Strong. 20+ year monopoly maintained.
Broadcom	Moderate. Custom AI chip design (XPUs) for hyperscalers.	Strong. Growing custom silicon demand.	Strong. 75%+ software margins, improving.	Strong. Hock Tan's acquisition + integration track record.
Vertiv	Moderate. Data center power and cooling infrastructure.	Strong. Every new data center needs power.	Strong. Margins expanding from 12% to 20%+.	Moderate. Turnaround story, less track record.
Palantir	Weak. Software, not infrastructure. Competitors exist.	Moderate. AI platform demand growing but competitive.	Strong. Reached profitability, margins expanding.	Moderate. Government to commercial pivot ongoing.

TSMC and ASML score highest because they hold genuine infrastructure monopolies. Every AI chip passes through TSMC's fabs and ASML's lithography machines. Like NVIDIA, they sit at chokepoints that the entire industry depends on. The question is whether their stocks already reflect this.

Broadcom is interesting because it's building a custom chip business for hyperscale customers who want alternatives to NVIDIA. If Google, Meta, and Amazon shift spending toward custom silicon, Broadcom captures that demand. The risk is that NVIDIA's CUDA moat keeps customers locked in.

Vertiv plays the picks-and-shovels angle one layer deeper. AI data centers need massive amounts of power and cooling. Vertiv supplies that infrastructure. The secular demand is clear, but the competitive position is less monopolistic than TSMC or ASML.

Palantir has momentum but fails the infrastructure monopoly test. Software companies face competition in ways that hardware infrastructure monopolies don't. Palantir could be a great investment, but it's a different pattern than NVIDIA's.

The Valuation Check

Good Company vs. Good Stock

A company can match the NVIDIA pattern perfectly and still be a bad investment if the stock price already reflects everything. The pattern identifies great businesses. Valuation determines great investments.

Pattern Score vs. Current Valuation

Forward P/E ratio and PEG for each candidate

Company	Forward P/E	Expected Growth	PEG Ratio	Assessment
TSMC	22x	25%	0.88	Reasonable. Growth justifies multiple.
ASML	32x	20%	1.60	Premium. Monopoly premium may be justified.
Broadcom	30x	22%	1.36	Fair. Custom silicon optionality priced in partially.
Vertiv	38x	30%	1.27	Reasonable for growth rate. Execution risk.
Palantir	120x	25%	4.80	Expensive. Needs to exceed expectations by a lot.

TSMC stands out. Strong pattern match, reasonable valuation (PEG 0.88), and 25% expected growth. It has the infrastructure monopoly, the secular demand, and the stock price doesn't yet fully reflect it. The geopolitical risk (Taiwan) is the discount.

Palantir is the opposite. Even if the company executes perfectly, the stock at 120x forward earnings already prices in years of exceptional growth. At a PEG of 4.80, you need Palantir to grow far faster than analysts expect. The company might be good. The stock is expensive.

The next NVIDIA won't look like NVIDIA does today. It will look like NVIDIA did in 2016: a company with a clear infrastructure position, growing demand, expanding margins, and a stock price that underestimates all three.

Your Checklist

The Four-Question Screen

Apply this to any stock pitched as "the next NVIDIA." It takes five minutes.

Infrastructure monopoly?

Does the company sell tools the entire industry needs? Can customers switch easily? If competitors are close, it's not a monopoly. Look for 60%+ market share in a critical input.

Secular demand?

Is demand driven by a multi-year trend affecting many customers, or by one big contract? Secular demand means revenue grows even if individual customers slow down. Cyclical demand means you're one downturn from a revenue cliff.

Margin expansion?

Are gross margins going up, not just revenue? If margins compress as the company scales, revenue growth gets diluted. The magic is when margins expand with revenue, creating a double engine for earnings.

Undervalued?

Check the PEG ratio. Below 1.5, the market hasn't fully priced in the growth. Above 2.0, the stock needs to beat expectations to justify the price. A great business at a stretched valuation is still a mediocre investment.

4 of 4

Infrastructure monopoly. Secular demand. Margin expansion. Reasonable valuation. NVIDIA had all four before the run. The next one will too. Find the chokepoint, check the price, and size the position so you can hold through the volatility.

How I Built This

Financial data from company earnings reports and FactSet consensus estimates. Framework derived from NVIDIA's 2016-2026 trajectory analysis.

NVIDIA Revenue Data

SEC filings, fiscal years ending January

NVIDIA's fiscal year ends in January. FY2025 refers to the fiscal year ending January 2025. The $130B+ figure is an estimate for FY2026 based on the first three quarters of reported results plus analyst consensus for Q4. Exact final figures will differ.

Framework Limitations

Post-hoc pattern recognition, not predictive

This framework was reverse-engineered from NVIDIA's success. Survivorship bias applies: we built the model from the winner. Companies that had three of four traits and still failed aren't in the sample. The framework is a screen, not a guarantee. Use it to narrow the field, not to make final decisions.

Candidate Valuations

FactSet consensus, February 2026

Forward P/E and growth estimates change daily. Palantir's 120x forward P/E fluctuates significantly. The PEG ratios are snapshots. Check current values before acting. The directional assessment (expensive vs. reasonable) is more durable than the specific numbers.

Geopolitical Risk Not Modeled

TSMC's Taiwan risk is acknowledged but not quantified

TSMC's discount relative to its pattern score partly reflects geopolitical risk. If cross-strait tensions escalate, the investment thesis breaks regardless of fundamentals. This framework does not model tail risks. That's what position sizing is for.

Jesse Walker has been an individual investor for 30 years. Before that, he was a poker professional, which is where he learned that the best decision and the best outcome aren't always the same thing. He writes about financially navigating the uncertainties of AI.