The "split season" in 13F: Chinese concept stocks are collectively increased, Nvidia becomes the single stock with the most divergent opinions
As the third quarter of 2026 13F disclosures comes to an end, a seemingly contradictory picture of holdings emerges.
According to Zhitong Finance APP, as the curtain falls on the Q3 2026 13F disclosures, a seemingly paradoxical portfolio picture emerges. Institutions such as Berkshire Hathaway, Bridgewater, Renaissance Technologies, Hillhouse, Greenwoods, Gao Yi, and H&H under Duan Yongping have successively revealed their cards, with Chinese ADRs becoming one of the few consensus directions. Pinduoduo (PDD.US), Alibaba (BABA.US), Trip.com (TCOM.US), and BOSS Zhipin (BZ.US) were all added to varying degrees. For Nvidia (NVDA.US), institutions have engaged in an almost "directional war"—some liquidated 160 million shares, while others increased their positions by over sixfold. The coexistence of consensus and divergence is precisely what makes this quarter’s 13F filings particularly worth studying.
Chinese ADRs: From “Avoidance” to “Group Rebuilding of Positions”
The most coordinated moves in the third quarter involved Chinese ADRs. Hillhouse’s HHLR increased its holding in Pinduoduo from 500,000 to 4.447 million shares, an increase of 789%, making Pinduoduo its largest holding again with an end-of-quarter market value of $347 million, accounting for 22.46% of the portfolio. In the same quarter, HHLR increased its position in Alibaba by about 2.3059 million shares, an increase of over 1,250%, making Alibaba the second-largest holding. Gao Yi Asset acted on both Alibaba and BOSS Zhipin: its Alibaba position grew from about 3,399 shares at the end of Q2 to 493,400 shares, essentially rebuilding the position from a very low base; BOSS Zhipin was increased by about 2.2504 million shares, becoming the largest holding at 22.37% after adjustment, while Trip.com and NetEase also saw position increases.
Duan Yongping’s moves were even more consistent. H&H increased its Pinduoduo holdings by 2.9688 million shares in Q3, raising Pinduoduo’s portfolio weight to 12.69% and making it the third-largest holding, with an end-of-period market value of about $2.182 billion. This marks Duan Yongping's fourth consecutive quarter of increasing his stake in Pinduoduo, with holdings growing from 7.23 million shares at the end of 2024 to 27.99 million shares, nearly quadrupling the original amount. Li Lu’s Himalaya Capital was also present, adding another 1 million shares of Pinduoduo in Q3, bringing total holdings to 11.761 million shares, now the largest individual position by stock. Additionally, Li Lu increased Tencent Music from 6.59 million to 14.824 million shares, an increase of about 125%.
Foreign institutions are also gravitating towards Chinese ADRs. Fidelity increased its positions in JD.com, Vipshop, and KE Holdings; BlackRock increased its holdings in JD.com, Trip.com, and JinkoSolar; JPMorgan increased its stakes in Pinduoduo and BOSS Zhipin. After several quarters of valuation compression, Chinese ADRs have become one of the few "consensus lines" this quarter. However, beneath this consensus, choices vary: Gao Yi increased Alibaba, Hillhouse added both Pinduoduo and Alibaba, but both institutions also made selective position adjustments among other Chinese ADRs within their portfolios. The rebalancing within Chinese ADR positions is far more nuanced than a simple “collective bullishness.”
Nvidia: Between Liquidating 160 Million Shares and Increasing Holdings Sixfold
According to holdingsintel.com’s compilation of 497 institutions’ 13F data, Nvidia is this quarter’s most divisive single stock: one end sees about 160 million shares liquidated—almost a 99.7% reduction; the other end shows a 648% position increase, with both sides delivering significant moves.
The reduction side’s strength is clear: Sixth Street Partners liquidated about 160 million shares. Renaissance Technologies slashed its holdings from 7.0913 million shares at the end of Q2 to 23,200 shares, a reduction of about 99.67%, with a previous market value of around $1.42 billion—virtually a complete exit. Simultaneously, it made Micron Technology its largest holding and significantly reduced positions in Meta, Apple, and Amazon. H&H under Duan Yongping reduced Nvidia by about 3.9612 million shares, taking its portfolio weight from 6.58% to 3.08%, but Duan later stated: "The moat and demand remain strong—not bearish, but won’t add for now," also noting that some of the reduction might relate to options operations. Bridgewater reduced about 1.178 million shares, a cut of roughly 30%; Soros Fund cut nearly 20%; Gao Yi Asset completely liquidated its previous 80,000 Nvidia shares. Wellington Management reduced about 2.0153 million shares, but still held about 120 million shares at quarter-end, making Nvidia its largest holding.
The increasing side was more about “rebuilding or multiplying positions.” ARK increased by 85.43%, Point72 expanded its holdings by roughly fourfold, Third Point rebuilt a 1.5 million share position after exiting in Q2, DekaBank increased by 648%, and Tepper was also in the accumulating camp.
Perhaps the most notable detail is the time window: these 13Fs reflect holdings as of September 30, while the Jensen Huang resignation issue occurred on November 12. Institutional rebalancing of Nvidia in Q3 was thus more about re-assessing valuation, portfolio concentration, and AI computing expectations—not a “pre-emptive move” for subsequent management changes. Interpreting Nvidia’s reduction simply as “bearish on AI” misses the real signal.
Rotation, Not Exit: Position Reallocation Within the AI Value Chain
If we view Nvidia reductions within the larger context of portfolio moves, institutional directionality becomes clearer: capital is not systematically exiting technology, but reallocating within the AI value chain.
Bridgewater is a typical example. While reducing Nvidia, Amazon, Alphabet, and Microsoft, it increased Qualcomm from 162,000 shares to 878,000 shares, up about 440%, raised Micron Technology by 120,500 shares, increased Applied Materials by 272,000 shares, and built a new position in Cisco. What’s being reduced are “compute leaders that have risen significantly,” while additions target storage, equipment, RF, and other upstream hardware. Soros’s approach is similar, but even more extreme: cutting Nvidia by about 20% while boosting Intel from 100,000 to 2.83 million shares—an increase of about 2,730%—and adding SK Hynix. The logic points to “selling expensive compute, buying cheaper logic chips and storage.”
Greenwoods chose another type of “internal differentiation.” In Q3, it increased Taiwan Semiconductor by about 892,200 shares, taking holdings to around 11.74 times the Q2-end level, making TSMC the third-largest position. It also added 1.3241 million Intel shares, an increase of about 48.1%, and established a Meta position of about 304,200 shares. With Nvidia, they only bought back 118,000 shares, worth about $27.04 million, making it more of a token exposure than a core bet. Gao Yi’s path is nearly the opposite of Greenwoods: liquidating Nvidia, Micron, and SanDisk, reducing TSMC, but more than doubling its Alphabet C shares. On the same AI value chain map, some bet on foundry, some on storage, some on software platforms—the divergences here rival those on Nvidia itself.
Evergreen, another Hillhouse filing entity, increased Google A shares about 288-fold, making it its top holding, and also increased Nvidia by 477%. Within the same institutional family, HHLR heavily concentrated on Chinese internet, while Evergreen bet big on Google and Nvidia—two parallel lines. This further demonstrates that Q3’s portfolio rebalancing is not a simple “tech vs non-tech” binary, but reflects fine-grained division of labor within different products and strategy frameworks under the same institution.
Berkshire’s “No Action” Is Also a Signal
Amid all the portfolio shifts, Berkshire Hathaway’s choices stand out for their quietness. At Q3's end, its US equities position was about $304 billion, with the only notable move being another reduction of Bank of America—selling another 27.533 million shares, dropping the holding ratio from 9.20% to 8.16%. The 78.79 million shares of Google A and 27.19 million shares of Google C were untouched, with a combined market value of about $36.375 billion, still Berkshire’s third-largest position. Apple also remained unchanged, with the share price rising about 15% in the quarter, boosting its market value from $65.95 billion to $75.901 billion—and its portfolio weight back to 24.95%.
The more striking numbers aren’t in the 13F: Berkshire repurchased about $9.6 billion of its own shares during the quarter, its net external equity trades were only slightly negative, and cash plus short-term Treasuries remained as high as $377.4 billion. While others debate the direction of AI compute and reallocate technology stocks, Berkshire expresses its view with buybacks and cash: when there aren’t enough attractive external opportunities, its own stock is the best allocation.
The True Signal Isn’t About the Direction Itself
This quarter’s 13F has three aspects that are easily misinterpreted: equating “Nvidia reduction” with “bearish on tech”; reading “increased Chinese ADR positions” as a “trend reversal” for Chinese ADRs; and interpreting Berkshire’s “no action” as “no information.”
But by looking at the three sets of data together, the conclusion is actually clearer than “bull vs bear divergence”: institutions are not leaving AI but are rotating within the AI value chain. The divergence on Nvidia is really about valuation and position concentration, not the direction of AI itself. Chinese ADRs represent a crossroad of valuation recovery and position replenishment, not a wholesale reversal. Berkshire’s inaction isn’t a lack of conviction; it expresses another choice through buybacks and cash.
Disclaimer: The content of this article solely reflects the author's opinion and does not represent the platform in any capacity. This article is not intended to serve as a reference for making investment decisions.
You may also like
Goldman Sachs, Morgan Stanley, and the AI market are deeply linked—How much longer can Wall Street investment banks enjoy good times?
Paul J. Davies, a columnist for Bloomberg, pointed out that the AI financing boom has driven a significant increase in performance for Wall Street investment banks. The five largest U.S. banks are expected to see stock trading revenue grow by more than 40% year-on-year this year, reaching around $85 billion. However, the IPO financing wave may be approaching a short-term peak, and demand for leverage from hedge funds is also cooling down, making it difficult for investment banks to replicate this year’s high growth by 2027.
Citigroup lowers 12-month Bitcoin target to $99,532, citing weak ETF demand
Bitcoin and Ethereum prices dip as World Bank explores stablecoin payments
BlackRock Holds 123,056 MSTR Shares Ahead of MSCI Removal
