Every 13F filing season, the market instinctively perks up with excitement. Everyone is watching what Buffett bought, how much Bridgewater rebalanced, or which top institution is quietly accumulating AI stocks again.
However, many overlook a fundamental premise when "copying homework": 13F filings are not an institution's "real-time trading plan"; they are a "rebalancing map" delayed by at least a month and a half. Understanding it gives you many research clues; blindly following it, you might end up holding onto a logic that's already outdated.
We are currently in the 2026 Q2 13F filing season. Over the past month, the U.S. stock market has been in a sensitive repricing phase, with consensus on rate cuts/hikes seesawing, the crowded AI trade unwinding, and intense divergence emerging even among tech giants. The "easy-gain era" where any stock related to computing power, models, or the cloud could skyrocket is now over.
Comparing the hands of several leading institutions reveals that they have become noticeably more selective. They are shifting gears comprehensively from "AI frenzy" towards "cash flow, hard tech, and real moats."
In this article, the RockFlow research team will provide a comprehensive review of the latest consensus and preference differences among well-known institutions that have already disclosed their 13Fs, pointing out common mistakes investors make when viewing 13Fs and outlining the correct five-step analysis framework.
Rebalancing Signals from Five Core Institutions
Temasek: Doubling Down on SpaceX, Google, ASML, Betting on Hard Tech
Temasek's moves this time are highly indicative. Its top ten holdings concentration reached 51.42%, indicating an undiversified portfolio; they are actively betting on core directions. The three most important actions:
Significantly increased holdings in SpaceX
SpaceX jumped to become Temasek's fifth-largest holding in Q2, with about 9.83 million shares worth approximately $1.68 billion. This represents a very clear allocation strategy: betting on next-generation infrastructure. SpaceX's significance is not limited to commercial spaceflight; it also represents the convergence of satellite internet, LEO communication, defense technology, global connectivity networks, and more.
Significantly added to Alphabet holdings
Google surpassed Visa to become Temasek's second-largest U.S. stock holding, with a market value of about $2.53 billion. This shows Temasek wants not only hard tech but also stable cash flow. Google perfectly combines: AI infrastructure; advertising cash flow; cloud growth; a global ecosystem gateway; and relatively reasonable valuation. Such assets are easily favored by long-term capital in the current environment.
Bought ASML
ASML is a key player in global advanced lithography equipment. Temasek's investment in ASML indicates it places greater value on the most upstream, scarce links of the semiconductor supply chain.
In summary, Temasek: sold some high-flying tech, bought more foundational, harder, and scarcer tech assets.
Bridgewater: One Hand on Tech Leaders, One Hand on Defensive Positioning
Bridgewater has not yet disclosed its latest Q2 holdings, but its Q1 equity portfolio was valued at about $22.4 billion. Its core five holdings include: SPY; IVV; Amazon; Nvidia; Alphabet. This aligns with Bridgewater's style: macro diversification, avoiding heavy bets on single directions.
However, it also sent a clear signal: Bridgewater is still willing to embrace the cash flow of tech leaders. Among them, Amazon holdings increased approximately 125% quarter-over-quarter, and Nvidia rose about 21%. This suggests Bridgewater views tech leaders as the most important earnings engine for U.S. stocks.
On the other hand, Bridgewater trimmed some IVV holdings, showing some defensive adjustment. Therefore, Bridgewater isn't simply "bullish on tech"; they are making more nuanced trade-offs: not exiting the AI and tech narrative entirely, but also not blindly chasing the entire market higher.
Greenwoods Asset Management: Sharp Contraction, Exiting Nvidia and Meta, Shifting to Equipment and Optical Communications
Greenwoods Asset Management (often referred to as "景林" in Chinese context) made aggressive moves in Q2. Its total U.S. equity portfolio value dropped from $3.88 billion to $2.19 billion, a decrease of about 43%. This indicates a significant reduction in risk exposure.
Most notably: Greenwoods exited Nvidia and Meta. This could easily be interpreted by the market as "bearish on AI." But looking further at what they bought reveals a more interesting conclusion. Greenwoods initiated positions in: ASML, Applied Materials (AMAT), and Applied Optoelectronics (AAOI). Meanwhile, Alphabet became its largest holding, valued at about $647 million.
Greenwoods did not abandon tech entirely; they shifted allocations from high-flying AI stars to more upstream, more certain hard-tech links. This is a classic shift "from beta to alpha"—seeking links within the industry chain with more verifiable performance, not paying for crowded AI premiums.
Berkshire Hathaway: Ending the Wait, Making a Rare Large-Scale Purchase of Google
Berkshire Hathaway's change is the most talked-about. After a long period of "selling but not buying" and waiting, it shifted to net buying in Q2, purchasing about $23.5 billion. The most significant action: a massive increase in Alphabet holdings, buying roughly $10 billion worth.
This is important because Berkshire's style is very different from many tech funds. It doesn't chase trends. Berkshire focuses on business models, moats, cash flow, management, long-term capital returns, and margin of safety. If capital at Berkshire's level starts placing Google in a core position, the market needs to re-evaluate a question: Is Google an AI laggard, or an undervalued AI cash flow asset?
Of course, one shouldn't buy just because Berkshire bought. But this signal deserves a place on the research list. Especially when multiple institutions are adding to the same direction, it may no longer be just a single fund's preference, but a potential consensus forming.
Baillie Gifford: SpaceX Becomes Largest Holding, Continuing to Bet on Long-Term Growth
Baillie Gifford has always favored long-term growth and innovative assets. Its Q2 portfolio value rebounded to about $110 billion. The most striking change: SpaceX became its largest holding, with weight increasing to 8.66%.
Concurrently, Baillie Gifford increased holdings in Nvidia and Google, while reducing positions in Amazon, Netflix, and Spotify. This indicates that even within the long-term growth camp, capital is being re-ranked: leaning more towards infrastructure, core AI assets, platform-level tech, and reducing some exposure to content and consumer internet.
The Most Important Conclusions from This Round of 13Fs
Analyzing all institutional actions together using our in-depth analysis, we find this round of 13Fs yields several core conclusions:
Google is Becoming a Consensus Institutional Asset
Temasek added. Bridgewater holds it heavily. Greenwoods' largest holding. Baillie Gifford increased holdings. Berkshire made a large-scale purchase.
This is no coincidence. Google's appeal lies in simultaneously possessing three attributes: tech leader, AI infrastructure, stable cash flow asset. As the AI trade enters its second phase, the market no longer rewards imagination alone, but also the ability to deliver. Google sits right at this intersection.
Nvidia Remains Core, But Its Crowding is Starting to Be Scrutinized by Institutions
Some continue to buy, some partially sell, some exit entirely. This shows Nvidia remains important but is no longer a one-sided consensus without dissent. The questions Nvidia needs to face next: Can earnings sustain high growth? Can gross margins hold? Can customer capex continue? Will competition intensify? Can valuation be further justified by earnings?
Upstream Semiconductors are Gaining More Attention from Long-Term Capital
ASML, AMAT, optical communications, equipment chain—these are entering more institutional radar screens. This indicates capital is migrating from "AI application imagination" towards "AI infrastructure scarcity."
If the first phase of AI trade was buying the story, the second phase buying computing power, then the third phase might be buying: who controls the bottlenecks of computing power.
SpaceX Represents Another Type of Institutional Preference—Next-Generation Infrastructure
Both Temasek and Baillie Gifford place SpaceX in very high positions. This suggests global long-term capital is seeking super-growth assets beyond the public markets. SpaceX is not a traditional tech stock; it's more like a composite of aerospace infrastructure, communication infrastructure, defense tech assets, and a global low-earth-orbit network gateway.
The 3 Most Common Mistakes Investors Make When Looking at 13Fs
13Fs are useful but easily misused. Especially when seeing clickbait news, many rush to "copy the homework." This is very dangerous.
Mistake One: Treating 13F as Real-Time Trade Records
13Fs have a reporting lag. What you see today is the fund's holdings at the end of the last quarter. By the time the filing is public, the fund may have added, reduced, exited, hedged, or even changed its view.
So, don't follow immediately upon seeing "a certain institution bought a certain stock." Ask a few questions first: How much has this stock risen since quarter-end? Is the current price far from the institution's likely entry range? Has the market already front-run the disclosure? Do the latest earnings still support the original buy thesis?
13F is a clue, not a buy button.
Mistake Two: Only Looking at "What Was Bought," Ignoring "How Much Was Bought"
Adding a new stock isn't necessarily significant. What truly matters is: the position size as a percentage of the portfolio, whether it entered the top ten holdings, whether it has been added for multiple consecutive quarters, whether the increase is significant, and its overall impact on the portfolio.
A 0.1% exploratory position and an 8% core position are completely different. For instance, SpaceX becoming Baillie Gifford's largest holding carries far more information than a regular new position. Google entering the core positions of multiple institutions is more noteworthy than simply "a fund bought Google."
Mistake Three: Overlooking Fund Style Differences
The same stock appearing in different funds' 13Fs carries entirely different meanings:
For Berkshire Hathaway, it might be a long-term core asset.
For Bridgewater, it might be part of macro asset allocation.
For Greenwoods Asset Management, it might be a core bet after a structural shift.
For Baillie Gifford, it might be a ten-year growth narrative.
For an event-driven fund, it might be just a short-term catalyst trade.
Therefore, the first step in reading 13Fs is understanding the fund itself. Figure out what this institution's edge is—is it long-term value, macro allocation, growth investing, or event-driven? What is its turnover rate? How long has it typically held similar assets in the past?
Without understanding the fund's style, it's easy to misread the signals.
As an AI research assistant designed for U.S. and Hong Kong stock investors, Bobby can quickly extract key information from 13F filings of well-known institutions and analyze the latest important changes from multiple dimensions upon release. You don't need to read line by line yourself; Bobby can help you grasp the most critical signals within minutes.
The 5-Step Framework for Correctly Analyzing 13Fs
If you don't want to just watch from the sidelines, you can use the following framework.
Look at the portfolio structure, not just the headlines
First look at the overall portfolio, then at individual stocks. Who are the top ten holdings? What is the concentration? Is the portfolio more concentrated or diversified? Are there new core assets? Temasek's top ten concentration reaching 51.42% means changes in its major holdings carry significant information.
Look at marginal changes
The most important aspect of 13Fs is "what changed." Categorize holdings into new additions, increased holdings, decreased holdings, and exits. Then ask: Which assets saw sustained buying? Which saw profit-taking? Which were abandoned entirely? Where did capital flow from and to? This is more useful than just looking at single stocks.
Look at sector and thematic migration
The real institutional view often lies in sector migration. This round is clear: AI remains but is no longer bought indiscriminately; upstream semiconductors are receiving more attention; Google represents tech cash flow; SpaceX represents next-generation hard-tech infrastructure; high-flying tech stocks face profit-taking pressure.
Look at consensus and divergence across multiple funds
A single fund buying is just a clue; multiple funds adding simultaneously warrants deeper research. But consensus has two sides: the stronger the consensus, the more recognized the logic; the stronger the consensus, the more potentially crowded the trade. Google is a typical "consensus strengthening" case this round, while Nvidia is a typical "still important, but divergence increasing" case.
Return to price, valuation, and earnings
13Fs only tell you what happened in the past; they don't judge whether it's worth buying now. Before actually investing, verify: Has the stock price already surged? Is the valuation still acceptable? Do the latest earnings support the institutional thesis? What are the risks? Under what circumstances would this judgment be proven wrong?
Conclusion: 13Fs Are the Footprints Institutions Leave Behind
The RockFlow research team believes one of the most dangerous moves in the market is treating someone else's holdings as your own logic. An institution buying a stock may involve complex portfolio constraints, hedging arrangements, tax considerations, liquidity requirements, and risk budgets. You see its footprints, not its complete thought process.
But footprints still have value. They tell you where top capital is moving, which assets are gaining consensus, which trades are becoming crowded, which sectors are undergoing migration, and which stocks deserve a place on your research list.
Don't rush to copy the homework. First, learn to read the clues. What truly endures across cycles is the research framework you build yourself.
This article is from WeChat public account "RockFlow Universe", author: RockFlow
