A single $13.7 billion GPU contract shows how fast the AI compute build-out is reshaping markets, but the bigger risk is that this theme may already dominate your portfolio without you realizing it. This post explains how to measure your true exposure to AI infrastructure across stocks, funds, and even crypto, and how to use AI-powered analysis to flag dangerous concentration in a single trade.
What is AI compute exposure and why does it hide in your portfolio?
AI compute exposure is the share of your portfolio whose value depends on the build-out of data centers, GPUs, networking, and power that trains and runs AI models. It hides because it is spread across dozens of names and inside funds you never chose for that reason.
When RUM Group signed a six-year $13.7 billion GPU services contract on August 25, it was one link in a supply chain that touches chipmakers, cloud providers, utilities, and memory suppliers. Your portfolio likely touches several of them at once.
The problem is that these positions do not announce themselves as a single theme. You may own:
- Direct chip plays like NVDA, AMD, and AVGO
- Hyperscalers like MSFT, GOOGL, and AMZN that fund the data centers
- Memory and storage names like MU that feed GPU servers
- Index funds where the same megacaps sit at the top of the weighting
Add these together and a portfolio that feels diversified across 15 tickers can be a concentrated bet on one macro story.
How much of the S&P 500 is already an AI compute bet?
Roughly a third of the S&P 500 by market weight now sits in a handful of AI-linked megacaps, which means most broad index investors are heavily exposed by default. If you own VOO or SPY, you already own a large AI compute position whether you planned it or not.
This is the same lookthrough problem we cover in our breakdown of how much Apple you really own across your funds. The mechanics are identical for AI compute names.
The double-counting trap
Investors often stack direct shares on top of fund holdings without noticing the overlap. A typical example:
- You hold NVDA directly as a conviction position
- Your S&P 500 ETF holds NVDA near the top of its weighting
- Your Nasdaq-100 fund holds it at an even higher weight
Your true NVDA exposure is the sum of all three, not just the shares in your direct account. Miss this and you underestimate concentration by a wide margin.
How do you map AI compute exposure across every account you own?
You map it by consolidating all holdings into one view, then tagging each position by its role in the AI compute supply chain. This is nearly impossible if your assets are scattered across three brokers and a spreadsheet.
PortfolioTrackr handles this by pulling every account into one place. You can connect brokers through the SnapTrade bridge covering 35 brokers, plus three direct integrations with Alpaca, Bybit, and Interactive Brokers. Connecting a broker is optional though, and manual entry, CSV import, voice, text, and broker screenshots all work on every plan.
Once everything is in one view, group your holdings into supply-chain layers:
- Compute silicon: GPU and accelerator makers
- Infrastructure and cloud: hyperscalers and data-center REITs
- Enablers: memory, networking, cooling, and power
- Downstream demand: software firms monetizing AI features
If you also hold crypto, do not skip it. GPU demand and AI narratives spill into tokens, and our guide on how to track crypto and stocks together in one portfolio shows why a unified view matters here too.
Which holdings count as AI infrastructure exposure?
AI infrastructure exposure includes any company that earns meaningful revenue from building, powering, or supplying AI data centers, not just the chip designers everyone names first. The list is broader than most investors assume.
| Layer | Example tickers | What drives them |
|---|---|---|
| Compute silicon | NVDA, AMD, AVGO | GPU and accelerator orders |
| Cloud and data center | MSFT, AMZN, GOOGL | Capex on AI capacity |
| Memory and storage | MU | High-bandwidth memory demand |
| Power and utilities | VST, CEG | Data-center electricity load |
Notice the last row. Power is the quiet part of the AI compute story, because data centers need enormous, reliable electricity. A utility can be an AI trade even though it never appears on a tech screener.
How do you spot dangerous concentration in a single theme?
You spot dangerous concentration by measuring what percentage of your total portfolio moves on one shared catalyst, then comparing it against a threshold you set in advance. A common rule of thumb is that no single theme should drive more than 20 to 25 percent of portfolio value.
PortfolioTrackr uses AI-powered analysis to surface this for you. It reads across your consolidated holdings and flags when a cluster of positions, including fund lookthrough, adds up to an outsized bet on one theme like AI compute.
Signs your AI exposure is too concentrated
- Your top five holdings all move together on chip earnings days
- A single sell-off in NVDA swings your whole portfolio by several percent
- Your index funds and direct positions overlap heavily on the same megacaps
- You cannot name a holding that would rise if AI capex slowed
If more than one of these is true, you are running a thematic bet, not a diversified portfolio. That is fine if it is intentional. It is dangerous when it is accidental.
Why this matters for correlated drawdowns
Concentrated themes fall together. When a growth scare or a rate shock hits, AI compute names tend to sell off as a group, so the diversification you thought you had disappears exactly when you need it. We saw a related dynamic in our analysis of when rising oil and yields hit stocks and bonds together.
How can alerts help you manage AI compute risk without watching screens?
Alerts let you define price levels for your key AI holdings in advance, so you get notified the moment a position hits a level you care about instead of monitoring quotes all day. PortfolioTrackr monitors prices continuously through market hours and fires the alert as soon as your level is reached.
Set them up around your own plan, for example:
- A stop-loss level on your largest single AI position
- A Target 1 level where you plan to trim concentration
- A rebalance trigger if the theme grows past your chosen weight
PortfolioTrackr reports status against these levels, such as still below target, Target 1 reached, or stop-loss level reached. It does not give buy or sell advice. The decision stays yours, and pairing alerts with a plan is why we compared trailing stops versus bracket orders in a separate guide.
Why do broker apps miss thematic concentration?
Broker apps miss thematic concentration because they only see the assets held at that one broker and rarely offer fund lookthrough or theme tagging. If your AI exposure is split across three brokers and two fund families, no single broker app can show the full picture.
A dedicated tracker solves the fragmentation problem in a few ways:
- It consolidates every account, across up to 95 stock exchanges and 67 currencies
- It applies lookthrough so fund holdings count toward your true exposure
- It runs AI-powered analysis to flag single-theme concentration
If you are weighing tools for this, our data-driven comparison of six portfolio trackers walks through which ones handle multi-account lookthrough well and which do not.
The bottom line
The AI compute build-out is real, and deals like RUM Group's $13.7 billion GPU contract show the momentum behind it. The risk for retail investors is not missing the theme, it is owning far more of it than they realize across chips, clouds, memory, power, and index funds.
Consolidate every account, tag holdings by their role in the AI supply chain, and let AI-powered analysis flag concentration before a correlated drawdown does it for you. Set alerts around your own levels, keep the decisions yours, and you turn a hidden blind spot into a position you actually control.
Find out what you are actually exposed to
Sector and currency concentration across every account you hold, benchmarked against the S&P 500, NASDAQ and gold.
Check My Exposure
Frequently asked questions
How do I check how much AI stock exposure I really have?
Consolidate every brokerage and fund into one view, then add up direct shares plus the AI names hidden inside your index funds. Double-counting is common because megacaps like NVDA sit in both your direct account and your S&P 500 ETF. PortfolioTrackr applies fund lookthrough so your true theme exposure is visible in one place.
What counts as an AI infrastructure stock?
AI infrastructure stocks include chipmakers, hyperscale cloud providers, memory and networking suppliers, data-center REITs, and the utilities powering those data centers. The theme is broader than chip designers alone, because power and cooling are essential to running GPUs at scale. A single AI catalyst can move all these groups together.
How much of my portfolio should be in one theme?
A common guideline is keeping any single theme below 20 to 25 percent of total portfolio value, though your own risk tolerance sets the number. Concentrated themes fall together in a drawdown, which erases diversification when you need it most. PortfolioTrackr can flag when a cluster of holdings crosses your chosen threshold.
Why does my broker app not show my thematic concentration?
Broker apps only see assets held at that single broker and rarely offer fund lookthrough or theme tagging. If your AI exposure is spread across three brokers and two fund families, no one app shows the full picture. A dedicated tracker consolidates everything and measures your real theme weight.
Can PortfolioTrackr alert me when an AI holding hits a price level?
Yes. Set your own price levels and PortfolioTrackr monitors them continuously through market hours, firing the alert as soon as the level is reached. It reports status such as Target 1 reached or stop-loss level reached against your levels, without giving buy or sell advice. The decision stays entirely yours.