All the previous articles have been about industry logic. But whether that industry logic is actually right, and whether it’s moving fast or slow, is something you sometimes can’t judge on your own.
At times like this, one of the best ways to check is to see what top institutions are buying. They may not always be right, but the money they spent, what they bought, and when they bought it—these data are public, and they are mandatory SEC disclosures. That is the value of 13F and 13G.
Let’s start with 13F.
The SEC requires institutional investors with more than $100 million in assets under management to disclose their holdings within 45 days after the end of each quarter in a filing called Form 13F.
You can see what stocks they hold, how many shares, the market value, and whether they added or trimmed compared with the previous quarter. All hedge funds, mutual funds, and family offices have to report. This filing is the only information-equalizing tool between retail investors and institutions.
Here are a few of the most valuable things to look for in 13F.
First, track the complete AI holdings of a well-known fund. For example, technology-heavy funds like Tiger Global, Coatue, Citadel, and Lone Pine have position distributions across every layer of the AI value chain, which can help you reverse-check the judgment that “scarcity is at which layer.” If several top funds simultaneously shift positions from Nvidia into some upstream company, that is a signal worth paying attention to.
Second, look at new positions and major increases. The most valuable part of a 13F is the “change,” not the “existing holdings.” Existing holdings represent historical judgment; changes represent current judgment.
Third, pay attention to disclosure delays. 13F is disclosed 45 days after the end of the quarter, which means what you are seeing may be positions from as much as 135 days ago. It is a verification tool, not a copy-trading tool.
Now let’s talk about 13G.
This is another disclosure. When an institution holds more than 5% of a stock, it must file within 10 days.
13G is more real-time and more precise than 13F—you can know the exact date, price, and number of shares when the position was built. And the 5% threshold means this position is a “core holding” for the fund, not just something bought casually.
The most classic 13G disclosure in 2026 was Leopold Aschenbrenner’s Situational Awareness fund reporting a 5.6% stake in Nebius, worth about $2.6 billion. As soon as that disclosure came out, Nebius shares rose 12% that day.
Why did the market react so strongly? Because Leopold is the author of “Situational Awareness,” and his judgment on AI compute cycles is regarded in the industry as one of the most accurate. Putting $2.6 billion into Nebius is, in itself, a very strong industry signal.
This is not “copy-trading.” It is two independent lines of judgment arriving at the same conclusion—industry logic points to neo-cloud, and a top AI observer’s holdings also point to neo-cloud.
When reading 13F and 13G, there are several traps to watch out for.
First, institutions make mistakes too. Their holdings are not gospel; they are just data points.
Second, institutions may move in and out much faster than the disclosure frequency. The position you see may already have been closed.
Third, the cost basis of an institution’s position may be far below the price you can buy at now. Following a position does not mean following the returns.
Fourth, don’t look at just one fund—look at overlap across multiple funds. If the AI positions of ten top-tier funds all show the same new name, the signal strength is far greater than that of a single fund.
The right way to use 13F and 13G is to treat them as a verification tool, not a decision tool. First, get clear on your own industry logic, then look at whether top institutions’ holdings confirm your judgment. If the direction matches, it strengthens your confidence; if the direction differs, go back and check where your logic is missing something.
The real value of this approach is not copying someone else’s portfolio, but turning public data into part of your own understanding.
The course covered three judgment frameworks and one portfolio; that is only the beginning. Real AI investing is a long-term discipline—the industry changes every quarter, top capital rebalances every quarter, and your own understanding has to be updated every quarter as well.
Understanding 13F and 13G is the most basic brick for getting this kind of dynamic iteration running.