The AI Trade Has Three Chapters — a Weekend Audit to Check All of Them in One Sheet

This was, by most measures, a one-theme week in markets. The AI trade fired on all three of its chapters within five days: Microsoft's $30B Azure run-rate re-anchored compute, Micron crossed $1,000 for the first time as the memory shortage kept paying off, and NVIDIA's Actuate 26 conference planted the flag on embodiment. If you hold any AI-exposed name — and most retail investors do, directly or through a fund — this is a good weekend to find out which of those chapters you are actually trading. Here is a fifteen-minute audit to run before markets open again.

The idea of splitting the AI trade into chapters is not academic. Each chapter has a different driver, a different risk, and a different set of earnings ahead. Compute is driven by data-center capex. Memory is driven by supply tightness — Micron has already sold out its 2026 HBM capacity. Embodiment is driven by robot and vehicle deployments, which lag and are lumpy. When all three rally at once, as they did this week, they look like one trade. When one of them stumbles — and with NVIDIA reporting August 26, one of them is about to get tested — they stop looking like the same trade. Knowing which you hold is the only way to react to that with a plan instead of a guess.

The three chapters, in one line each

Compute is the data-center buildout — the GPUs that train large models, visible in the $30B Azure AI run-rate Microsoft reported in early August. Memory is the high-bandwidth memory that feeds those accelerators, the shortage that lifted Micron past $1,000 this week. Embodiment is the next step — robots, autonomous vehicles, and industrial automation that consume chips on-device, the Physical AI thesis NVIDIA pushed at Actuate 26. Each of these is the same underlying AI demand showing up in a different ledger, and each re-rates a different slice of the market.

How to audit all three chapters in one sheet

Step one is seeing what you actually hold across every brokerage, instead of logging into each one separately and adding up the same ticker by hand. That is exactly what InvestSheet removes: it syncs positions from Robinhood, Fidelity, Schwab, and 35+ other brokerages into a single Google Sheet, so each name shows up once with live formulas no matter which account holds it:

Step two is bucketing. Once the raw values are in one sheet, assign each position to a chapter — compute, memory, embodiment — and total the buckets. The key is to count the overlap honestly. NVDA is in every chapter, so if you hold it plus a robotics fund and a semiconductor ETF, your true embodiment exposure is not just the NVDA line — it is the NVDA value plus the robotics weight inside the fund. Add a column for the fund weight of each chapter and you will often find your real exposure is two or three times the direct holding.

Step three is sizing. Compare each chapter total against your single-name cap and your overall theme limit. A rule investors like Bernstein and the theme-concentration writers converge on is capping any single theme — "AI" counts as one theme — at a level you are comfortable losing a fifth of without changing your life. The chapters matter because they are not one position: a memory bust and an embodiment stumble are different risk events, and a portfolio that blends both into one "AI" line is the one that gets caught flat-footed when they diverge.

Why this weekend, specifically

The next two weeks carry the hard catalysts that make a written-down audit worth doing now. NVIDIA reports on August 26, the first earnings call to test whether the embodiment re-rating from Actuate 26 has any revenue behind it yet. The retail chapter — a fourth, more defensive part of the market — just wrapped its own earnings week, with small-ticket spending holding and big-ticket renovation names cutting guidance. Neither of those prints has to move your positions, but both will move markets, and weekends are when you get to write down what you would do before the opening bell forces the decision.

The quiet advantage of the weekend is that there is no tape. No flashing prices, no red-green panic, no reason to convince yourself a concentration is fine because it went up today. It is the one window where your portfolio can be looked at the way you would review anyone else's — calmly, by the numbers, across every account you own. Fifteen minutes now, with all three chapters in one sheet, is the difference between deliberating on Tuesday and reacting on Tuesday.

The through-line of this month's posts — the Azure run-rate, Micron's $1,000 crossing, Actuate 26, and the retail tape — is that the market keeps segmenting the AI trade into new phases. Compute, memory, embodiment. The investor who can see exactly what they hold in each slice, across every brokerage, is the one who gets to rotate deliberately. The weekend is when that seeing happens.

Frequently asked questions