六巨头诸神之战：13F 美股持仓全景，顶级机构开始互为对手盘？

The results of the first-quarter 2026 portfolio adjustments by six major fund managers are out.

As is well known, mid-May each year is one of the most noteworthy time windows for the global US stock market. At this point, major institutions must submit their 13F filings to the SEC, disclosing their holdings at the end of the previous quarter. Although the 13F itself has a lag, typically filed within 45 days after the quarter ends, making it unsuitable for real-time copy trading, it is excellent for observing how institutional funds re-understood the market's main theme in the previous quarter.

And the most important change in the Q1 2026 13F filings is not which stock was bought by whom, nor which big shot cleared out what, but rather that the consensus among Wall Street's top capital is starting to fracture.

Because in the past few years, there was a very clear common narrative in the US stock market: buy the Magnificent Seven, buy AI, buy platform leaders, buy high-quality tech. While the timing of capital deployment varied, the general direction was the same. But this time is different – for Google, some are frantically adding while others are nearly exiting; for Amazon, some have completely sold off while others continue to hold heavy positions; for Microsoft, some are establishing core positions while others are directly liquidating them; traditional SaaS was heavily liquidated by Bridgewater, but AI hardware and computing infrastructure were concentrated on by another group of funds.

This indicates that their judgments on "which layer will ultimately capture AI profits," "whose moat will be reassessed by AI," and "which valuations have already over-promised the future" are beginning to split significantly.

So, this batch of 13F filings is not a simple list of holdings; it's more like a map of Wall Street's opposing hands.

1. The Core Change: Consensus Shifts from 'What to Buy' to 'Who is Holding Whose Bag'

The most notable trend in these 13F filings is that institutions have started to become counterparties to each other within AI targets.

In the past, institutional trading in US stocks was more like a big river, with everyone heading in a similar direction, just differing in position size and timing. Now, it's more like a fork in the road. Everyone knows AI is the main theme, but no one is willing to pay the same valuation for the same story anymore:

• Some buy Google because it's cheap, has strong cash flow, and YouTube and Search still have moats; others sell Google because AI search could potentially disrupt its core business model.
• Some buy Microsoft because the certainty of Azure and enterprise AI adoption is higher; others sell Microsoft because the market has already priced in excessive AI premium.
• Some buy Amazon because AWS remains a core platform for AI cloud capital expenditure; others sell Amazon because they no longer want to bear the risk of such high-valuation platforms in their portfolio.
• Some are fleeing Salesforce and ServiceNow because the intermediary value of traditional SaaS is being compressed by AI; others are buying NVIDIA, TSMC, Micron, SanDisk, because regardless of which AI application wins, the underlying hardware must be purchased first.

So, the core of these 13F filings is not "buying AI."

It is that AI as a monolithic concept is disintegrating. Institutions are starting to split it into platform layer, application layer, hardware layer, industrial capex layer, and financial toll road layer, repricing each.

Let's break down each one in detail.

1. Berkshire Hathaway: Redrawing the Lines in the Post-Buffett Era

Objectively speaking, Q1 2026 is the first complete observation window since Berkshire Hathaway entered the post-Buffett era.

The most interesting aspect of this 13F is that it did two seemingly contradictory things simultaneously: sharply simplifying the portfolio while significantly increasing its position in Google.

According to public reports, Berkshire significantly increased its Alphabet holdings in Q1, while establishing new positions in Delta Air Lines and Macy's, and liquidating holdings in Amazon, Visa, Mastercard, UnitedHealth, and others:

• Exiting Amazon, Visa, Mastercard indicates it doesn't want to retain all seemingly high-quality business models from the past.
• Adding to Google means it hasn't moved away from tech, but is searching for assets within tech that are closer to Berkshire's traditional aesthetic: strong cash flows, not too expensive, sufficiently controversial, but whose underlying business hasn't been completely disproven.

This is also why Google became the biggest point of divergence in these 13F filings. Berkshire isn't buying the "AI story"; it's buying a cash flow giant that the market has begun to doubt, betting on the side that "Google's moat still holds value."

2. Pershing Square: Ackman Takes the Opposite Side of Buffett

If Berkshire was one of the biggest buyers of Google this quarter, then Bill Ackman is the quintessential counterparty.

Pershing Square's most stunning move in Q1 was nearly liquidating its Alphabet position and establishing a new one in Microsoft. In his public explanation, Ackman emphasized that Microsoft's valuation is more attractive after its price correction, and the long-term growth potential of Azure, Microsoft 365, and enterprise AI remains strong. In other words, he shifted his tech exposure from Google to Microsoft.

This forms a stark contrast with Berkshire. Ultimately, Berkshire sees the resilience of Google Search, YouTube, Cloud, and advertising cash flow, while Ackman sees the risk that generative AI could disrupt the search gateway.

One thinks Google is undervalued; the other thinks Google's moat is being repriced. To put it bluntly, Ackman hasn't given up on AI; he just doesn't think Google is the highest probability bet in this AI trade.

3. Bridgewater Associates: Dalio Sells Software, Buys Hardware

Bridgewater Associates' 13F is always complex, as it involves macro allocation rather than single-company bets.

But this time, Bridgewater's direction is very clear: sell traditional software, buy AI hardware.

Public 13F tracking shows that Bridgewater exited Salesforce in Q1 and clearly rotated towards AI hardware and infrastructure names like NVIDIA, TSMC, and Amazon. Some market reports also mention that TSMC became one of Bridgewater's significant new positions this quarter, while Salesforce was a major exit. This trend is very important.

It indicates that Bridgewater isn't simply bullish on tech; it's performing a supply chain rotation *within* tech. Over the past decade, traditional SaaS was one of the most comfortable business models: subscription revenue, customer stickiness, high margins, strong cash flows. But with the advent of AI, the valuation logic for traditional SaaS is being reassessed.

If large language models can automatically generate code, automate processes, and replace some enterprise software functions, then the intermediary value of traditional SaaS will be compressed. So, Bridgewater isn't retreating from tech stocks; it's more like rotating from "software middlemen" to "AI hard currency."

Assets like NVIDIA, TSMC, Micron, Broadcom, Oracle, and Amazon represent computing power, wafer foundry, memory, networking, cloud, and infrastructure. Their commonality is that regardless of which AI application ultimately wins, the underlying capital expenditure will likely have to pass through these links first.

To summarize in one sentence: Bridgewater isn't buying the AI concept; it's buying the money AI *must* spend.

4. Appaloosa Management: Tepper Bets on the 'Hardware No One Can Bypass'

David Tepper's Appaloosa Management also showed a very strong direction in Q1.

Public reports indicate that Appaloosa significantly increased positions in Amazon and Uber in Q1, exited airline stocks, added SanDisk, and continued to raise exposure to semiconductor and AI hardware chain assets like Micron and TSMC.

Tepper's logic is similar to Bridgewater's and very direct: It doesn't matter who ultimately wins in AI; buy everything that all winners must procure.

• Micron represents HBM and memory.
• TSMC represents advanced process nodes and foundry capacity.
• SanDisk represents the storage chain.
• Amazon represents AWS cloud infrastructure.

These aren't pure AI application stories; they are the hardware, cloud, and infrastructure of the AI arms race. Of course, while this overlaps with Bridgewater's thinking, Tepper is more concentrated and aggressive.

Where Bridgewater engages in macro-style "add hardware, reduce software," Tepper seems to be directly betting that the AI computing cycle is not over, and the ones who will truly capture orders and cash flow are the hardware and infrastructure chain.

To summarize in one sentence: Tepper is buying the shovel sellers, specifically those closest to the computing power bottleneck.

5. Duquesne Family Office: Druckenmiller's Signal is 'Don't Chase the Hottest Spots'

Stanley Druckenmiller's Duquesne Family Office is a bit different from the previous firms.

It wasn't the most typical major buyer of AI hardware this quarter, but its significance lies in representing another institutional mindset: don't stay too long in the most crowded trade.

Druckenmiller has previously reduced or exited popular AI concepts like NVIDIA and Palantir, while consistently watching more upstream parts of the supply chain like TSMC. Public reports also show that macro-trading oriented funds like Duquesne are characterized by rapid adjustment, not long-term commitment to a single narrative.

This is highly consistent with the main theme of these 13F filings: When the market has driven AI application plays into a consensus, truly sensitive macro capital has already started moving further upstream, down the stack, and to cheaper links.

To summarize in one sentence: he doesn't stand where the crowd is thickest; he goes to places the market hasn't fully priced in yet.

6. Egerton Capital: Sells Microsoft, Buys Google, NVIDIA, and Industrial Hard Assets

Egerton Capital not only bought AI but also financial infrastructure and industrial hard assets. More importantly, it liquidated Microsoft, standing on the opposite side of Ackman.

Public 13F tracking shows Egerton Capital's Q1 13F portfolio was around $9 billion, with top five holdings including Visa, Alphabet, Moody's, Linde, and Carpenter Technology. It also established or increased positions in NVIDIA, Linde, Devon Energy, Canadian Natural Resources, among others.

This set of holdings is very interesting. It didn't simply buy the Magnificent Seven but split the portfolio into several lines:

• Line 1: Financial Infrastructure – Visa, Moody's, CME, Interactive Brokers, Mastercard.
• Line 2: AI Platform and Compute – Alphabet, NVIDIA.
• Line 3: Industrial Hard Assets and Capex – Linde, Vulcan Materials, Carpenter Technology, Amphenol.
• Line 4: Energy and Resources – Devon Energy, Canadian Natural Resources.

This suggests Egerton isn't buying a single AI story; it's betting on the intersection of the AI cycle, industrial capital expenditure, and financial toll roads.

Most crucially, it liquidated Microsoft, creating a clear counterparty trade against Ackman.

Ackman believes Microsoft has a higher probability of winning in enterprise AI entry points and Azure, so he established a new position. Egerton chose to exit Microsoft entirely, placing more tech exposure into Google and NVIDIA.

Same AI, same quality growth, completely opposite answers from different institutions.

2. Who is Playing Against Whom? A Comparative Analysis

1. Google: The Biggest Divergence in This Round of 13Fs

It can be said that Google is the asset most worthy of a standalone analysis in these 13F filings: Berkshire significantly added, Egerton also increased its Alphabet position, but Ackman nearly liquidated his.

This shows Google has transformed from a consensus tech leader into a contentious asset. The bullish side believes Google Search, YouTube, Cloud, and advertising cash flows remain strong; its valuation is relatively less expensive; and the disruptive impact of AI is being overstated by the market. The bearish side believes generative AI could change the search gateway; the advertising business model faces repricing; Google Cloud and Gemini need to prove their commercialization efficiency; and high capital expenditure could compress margins.

So, Google isn't a simple case of "institutions are all buying" or "institutions are all selling." It's more like a moat stress test.

Those buying Google are buying cash flow and low valuation recovery. Those selling Google are selling the risk of AI search disrupting the old gateway.

2. Microsoft: Some See an Enterprise AI Gateway, Others See a Priced-in Story

Microsoft is also a target with significant divergence this time.

Ackman built a position in Microsoft, seeing long-term certainty in Azure, Microsoft 365, and enterprise AI. But Egerton liquidated its Microsoft position, indicating that another class of institution is unwilling to continue paying a high AI platform premium for Microsoft.

This divergence is critical. Microsoft hasn't been abandoned by the market, but it is no longer a consensus asset without controversy.

Its problem is whether a good company's potential has been preemptively discounted by its good price. This is the most common challenge for high-valuation tech stocks. The business can continue to perform well, but the stock price may have already priced in several years of future growth.

3. Amazon: Berkshire Sells, Ackman and Tepper Buy

Amazon is not a consensus either.

Berkshire exited Amazon, but Ackman and Tepper still value it highly. Bridgewater also keeps it in a significant position, and Egerton retained its position even after trimming.

The divergence behind this: Is Amazon a high-valuation platform risk, or a core infrastructure in the AI cloud capex cycle?

Bullish investors see AWS, the e-commerce base, advertising business, and AI cloud demand. Bearish investors see portfolio simplification, valuation discipline, and the need to rebalance platform assets.

So, Amazon isn't like Google with its "moat controversy." It's more of a "portfolio position controversy." Institutions will question whether, at the current price and portfolio structure, they still need such a large exposure to Amazon.

4. Traditional SaaS: From Safe Asset to Audited Asset

Bridgewater's liquidation of Salesforce and ServiceNow is one of the most structurally significant moves in these 13F filings.

It's not just selling two stocks. It represents the market beginning to re-evaluate the business model of traditional SaaS. In the past, SaaS was considered a high-quality asset, generating revenue through subscriptions, stickiness, data, and process automation. But with the generalization of AI large models, the intermediary value of enterprise software is being challenged.

If many processes can be automated by AI, much code can be generated by AI, and many software functions can be directly invoked by large models, then the high valuations of traditional SaaS need to be justified anew.

This is why, in these 13F filings, traditional software and AI hardware have formed a very clear counterparty trade. Selling the software middleman, buying the computing, memory, foundry, cloud, and hardware infrastructure.

5. Financial Infrastructure: Berkshire Sells, Egerton Buys

Visa and Mastercard also show an interesting divergence.

Berkshire liquidated its positions in Visa and Mastercard, but Egerton Capital's top holding is Visa, alongside other financial infrastructure assets like Moody's, CME, and Interactive Brokers. This suggests that financial toll roads haven't lost their value.

Institutions just disagree on their role within a portfolio. Berkshire might be engaging in portfolio cleaning and pruning old positions. Egerton treats assets like Visa and Moody's as long-term cash flow anchors.

So, this isn't a simple case of "payment stocks are failing." It's that financial infrastructure is no longer a no-brainer consensus, but it remains a high-quality anchor position in the eyes of some institutions.

3. How Should We Interpret These 13F Filings?

Many might say: "These 13Fs still look like buying AI. How can you say consensus is gone?"

The key point is that consensus on the AI *direction* remains, but the AI *beta consensus* is gone.

As we know, in the past, buying AI, buying the Mag 7, buying tech leaders, buying semiconductor ETFs, would generally align with the main trend. But it's different now.

AI is being disaggregated. The platform layer, application layer, hardware layer, cloud layer, industrial capex layer, and financial toll road layer are each being priced independently. This means the future won't be "everything AI goes up." The market will ask increasingly granular questions.

In short, the real signal from these 13F filings is: The AI trade is