AI Investing 101

Start here: how to invest in the AI buildout

Most AI investing advice is a list of companies that mention AI a lot. This page teaches a different lens, the one this whole site is built on: follow the money, not the buzzword.

The world's largest technology companies have committed to the biggest capital spending cycle in the history of computing. That money does not stay with the buyers. It flows down a supply chain to whoever gets paid to build: chips, memory, networking, optics, servers, buildings, cooling, and the power to run it all. Every layer of that chain is a set of public companies whose revenue rises when buildout budgets rise, whether or not any particular AI product wins.

Supercycle tracks all of those companies across every layer of the buildout, scores how much of each business actually depends on AI, and follows the capex, projects, and power deals that fund it. This page walks the framework top to bottom, then shows you where each tool fits.

How this site works

Everything on Supercycle hangs off three ideas: a universe, a score, and the money that moves it. The tools below are different windows onto the same data.

📊 The stock universe

Hundreds of companies, every one either in the AI supply chain or with AI in the product. AI-washed names are deliberately excluded, and the bar is documented.

🎯 AI Exposure Score

A 0-100 answer to "how much of this business is actually AI?" It separates pure plays from giants where AI is one driver among many. The leaderboard ranks the whole universe.

🏗️ Buildout Tracker

The money itself: $657B in trailing-year capex from 7 major AI buyers, 92 tracked data center projects, and 47 power deals, each tied to the tickers that get paid.

🦅 Analyst Agents

The Capex Hawk is an AI analyst running this page's framework in public: a paper portfolio that reads the buildout data weekly, trades on it, and publishes every order, fill, and rationale.

🌡️ AI Fear & Greed Index

A daily 0-100 sentiment reading for AI stocks, overall and per buildout layer, so you can compare how the market feels against what is actually being built.

🔍 Screener, Heatmap & Compare

Work the universe yourself: screen by layer and exposure, see the day's money flow on the heatmap, and put any stocks side by side with compare.

The buildout, layer by layer

Think of the AI buildout as a stack. A dollar of capex enters at the top as an announced budget and lands, quarters later, as revenue somewhere below. Each layer is a category page with every tracked stock in it.

1.Physical inputs

Before a single chip is made, the buildout needs raw materials and, above all, electricity. Data centers are power plants with computers inside.

2.Making the chips

The compute layer: the machines that make chips, the companies that design them, and the memory that feeds them.

3.Housing and connecting the compute

Chips are useless in a box. They need buildings, cooling, assembly, and the networking that ties a hundred thousand GPUs into one machine.

4.Renting the compute

Most companies never buy a GPU; they rent one. The clouds, and the newer AI-only clouds, sit between the hardware and everyone else.

5.Models and software

The layer most people mean when they say "AI stocks": the model builders and the software that packages AI into products.

6.AI in the physical world

Where the models leave the screen: robots, vehicles, medicine, and defense.

Real exposure vs. AI-washing

In every bull market, companies discover they were an "AI company" all along. The test that cuts through it is one question: does money from the AI buildout actually land in this company's revenue? Either the company sits in the supply chain and gets paid to build, or AI is genuinely in the product customers buy. Mentioning AI on earnings calls is neither. This universe was pruned on exactly that bar, and names that failed it were removed.

The AI Exposure Score turns the same question into a 0-100 number, banded into tiers:

Tier	Score	What it means
Pure Play	90+	The business exists because of AI demand.
Core	70+	AI is the dominant growth engine and most of the thesis.
Strategic	50+	AI is a top-two driver alongside a large existing business.
Growing	30+	Meaningful, growing exposure that is not yet the main driver.
Peripheral	0+	Real but minority exposure, or optionality.

Browse the whole universe ranked by score on the AI Exposure Leaderboard.

Ways to actually invest

There are three broad approaches, in increasing order of effort and concentration:

1. Broad funds.An S&P 500 index fund already carries heavy AI weight through its largest holdings, and semiconductor ETFs like SMH concentrate the compute layer. Least effort, least control: you cannot lean into power or memory when those layers are where the bottleneck is.
2. Picking layers.Decide which layers of the stack are supply-constrained or under-priced and own baskets of names within them. This is the framework this site is built for: category pages, the screener, and the buildout tracker exist to answer "which layer is the money flowing to next?"
3. Picking names.Individual stocks, highest risk and highest specificity. If you go here, the exposure score matters most: two companies in the same layer can have wildly different sensitivity to AI spending.

For a read on how professionals are positioned, the Investor Portfolios tracker filters famous investors' quarterly SEC filings through this universe and shows the AI slice of each portfolio.

Nothing on this site is investment advice; it is a data and research tool. See the disclaimer.

Reading where we are in the cycle

The honest way to track a capex cycle is to watch commitments, not headlines. Capex guidance, data center projects breaking ground, and signed power deals are expensive to announce and embarrassing to reverse, which makes them better signals than anything said in an interview. The Buildout Tracker currently follows $657B in trailing-year AI capex, 92 projects totalling 86 GW, and 47 power deals.

Sentiment is the other half. The AI Fear & Greed Index reads how the market feels about the same stocks, daily, overall and per layer. The gap between the two is the useful part: euphoric pricing on slowing commitments is a warning, fearful pricing on accelerating ones is an opportunity. For the day-to-day picture, the heatmap shows where money moved today.

And if you want to see this entire framework applied rather than described: The Capex Hawk is an AI analyst agent that reads the buildout data every week, holds a paper portfolio of the suppliers collecting the checks, and publishes every trade, rationale, and mistake against benchmarks. It is the framework with receipts.

Glossary: terms you'll see everywhere

Capex: Capital expenditure: money a company spends on physical assets like data centers, chips, and power. The AI buildout is, at its core, a capex cycle.
Hyperscaler: The handful of companies operating cloud infrastructure at massive scale, such as Amazon, Microsoft, Alphabet, and Meta. Their capex budgets fund most of the buildout.
Pure play: A company whose business exists because of AI demand. If AI spending stopped, the business would too.
AI-washing: Marketing a company as an AI stock when AI is neither in its supply chain nor meaningfully in its product. Common in bull markets.
GPU: Graphics processing unit: the chip architecture that turned out to be ideal for training and running AI models.
HBM: High-bandwidth memory: specialized memory stacked next to AI chips. One of the tightest supply bottlenecks in the buildout.
Training vs. inference: Training is teaching a model, done once at enormous cost. Inference is running it for every user query, forever. Both consume compute; inference is the recurring bill.
Neocloud: A newer cloud provider built specifically to rent out AI compute, as opposed to the general-purpose hyperscaler clouds.
EMS: Electronics manufacturing services: the companies paid to assemble servers and racks. They build the hardware others design.
PPA: Power purchase agreement: a long-term contract to buy electricity, often signed years ahead. A signed PPA is one of the hardest-to-fake signals that a data center is real.
GW: Gigawatt: a billion watts. Data center projects are sized in power, not square feet, because electricity is the binding constraint.

Frequently asked questions

What are AI infrastructure stocks?

Companies that get paid to physically build AI capacity: chipmakers, memory suppliers, networking and optics firms, server assemblers, data center operators, cooling specialists, power producers, and utilities. They sit in the supply chain of hyperscaler capital spending, so their revenue rises when AI buildout budgets rise, whether or not any particular AI product succeeds.

How do I invest in AI without picking a single stock?

Semiconductor and technology ETFs such as SMH give broad exposure to the compute layer, and a plain S&P 500 fund already carries heavy AI weight through its largest holdings. The tradeoff is that broad funds dilute exposure to specific layers like power, cooling, or memory, which is where picking individual layers or names differs from buying the index.

What is an AI pure play?

A company whose business exists because of AI demand: if AI spending stopped, the business would stop with it. Pure plays offer the most direct exposure and the most concentrated risk. Most large companies with real AI exposure are not pure plays; AI is one driver alongside an existing business.

What is hyperscaler capex and why does it matter for AI investing?

Hyperscaler capex is the capital spending of the largest cloud and AI companies on data centers, chips, and power. It matters because it is the money that flows down the supply chain: announced capex today becomes supplier revenue over the following quarters. Tracking who receives that spending is a more grounded way to invest in AI than guessing which chatbot wins.

What is AI-washing and how do I avoid it?

AI-washing is presenting a company as an AI investment when AI is neither in its supply chain nor meaningfully in its product, usually by mentioning AI often in earnings calls. Avoid it by asking one question: does money from the AI buildout actually land in this company's revenue? If the answer requires a story rather than a customer, it is washing.

How do I know where we are in the AI cycle?

Watch commitments rather than headlines: announced capex guidance, data center projects that break ground, and signed power deals. Those are hard to fake and take years to reverse. Sentiment indicators show how the market feels about that same buildout, and the gap between the two, euphoric pricing on slowing commitments or fearful pricing on accelerating ones, is the useful signal.

Keep up with the buildout

The newsletter tracks where the money is flowing, in the same follow-the-capex lens as this page.

Get the Newsletter

Market data as of the close on September 18, 2026. Prices are end of day, not live.