U.S. institutional portfolios remained committed to artificial intelligence-linked equities through the second quarter of 2026, but the latest regulatory filings show that the trade is becoming far more selective. Semiconductor stocks drew the clearest aggregate demand, while major managers took sharply different views on Alphabet, Nvidia, Broadcom and other large technology names.
A tally based on 13F filings from 6,371 institutional managers found that about 48% increased semiconductor exposure during the quarter, compared with 34.5% that reduced it. Positioning in the “Magnificent Seven” was almost evenly divided: nearly 44% of filers cut exposure, while about 42% initiated or added holdings. Software was more balanced still, with 28.2% net sellers and 26.3% net buyers.
The disclosures cover portfolios held on June 30 and were due to the U.S. Securities and Exchange Commission by Aug. 14. They offer a delayed snapshot of long U.S.-listed equity positions, rather than a complete account of a fund’s trading, hedges or cash holdings. Even so, the filings show that large managers have not broadly abandoned AI-related stocks. Instead, they have increasingly treated the sector as a collection of separate bets on chips, foundries, infrastructure, networking, software and corporate execution.
Alphabet becomes a dividing line among major managers
Alphabet stood out as one of the quarter’s most contested large-cap technology holdings.
Berkshire Hathaway increased its Alphabet exposure across the company’s Class A and Class C shares from roughly 57.84 million shares at the end of the first quarter to about 106 million shares at June 30, according to its SEC filing. The increase exceeded 80% and placed Alphabet among Berkshire’s more substantial changes during the period.
The conglomerate also added Delta Air Lines and homebuilder Lennar, showing that its second-quarter activity extended beyond technology. Those additions suggest Berkshire’s portfolio managers were willing to pair a larger position in an AI-linked platform company with exposure to travel and housing-sensitive businesses.
Tiger Global Management took the opposite direction on Alphabet. Its filing showed Alphabet shares falling from about 10.63 million to roughly 5.81 million, a reduction of 45.4%. Tiger Global also cut Broadcom by nearly half and trimmed Taiwan Semiconductor Manufacturing and several other large-cap technology holdings.
Yet Tiger Global’s moves did not amount to a wholesale retreat from AI infrastructure. The firm opened positions in AMD, Applied Digital and Cerebras, while adding compute and data-center-related holdings including Cipher Digital and Core Scientific. It also increased Intel from about 1.64 million shares to roughly 4.25 million shares.
That mix points to a portfolio rotation toward companies more directly tied to AI computing capacity, chip competition and power-intensive data-center buildouts. Alphabet and Broadcom remain central AI beneficiaries, but their scale and prior market performance appear to have led some managers to seek more differentiated exposure elsewhere in the supply chain.
Chip portfolios shift from established leaders to narrower bets
Several prominent hedge funds used the quarter to reshape semiconductor exposure rather than simply raise or lower it.
Third Point exited Nvidia, Broadcom, KLA, Lam Research and the VanEck Semiconductor ETF, known by its ticker SMH, according to its June 30 filing. The fund also sold its Meta position. At the same time, Third Point increased holdings in Alphabet and TSMC, opened stakes in Keysight Technologies and Flex, and expanded into non-technology names including Warner Bros. Discovery, Capital One and Norfolk Southern.
Warner Bros. Discovery became Third Point’s largest U.S.-listed equity position by the end of the quarter. The shift gives the portfolio greater exposure to media, consumer finance and transport while retaining targeted links to AI hardware and semiconductor manufacturing through companies such as TSMC and Keysight.
Duquesne Family Office made a similarly selective set of changes. The firm removed Broadcom and Micron Technology from its reported 13F holdings after owning both at the end of the first quarter. It initiated positions in Alphabet, AMD and Palo Alto Networks, the cybersecurity company, while increasing TSMC from about 495,000 shares to 590,000 shares.
Duquesne also raised its position in STMicroelectronics from roughly 2.61 million shares to 3.10 million shares. The additions to TSMC and STMicroelectronics place more emphasis on foundry capacity and industrial or automotive chip demand than on the highest-profile U.S. AI-chip names.
Semiconductor demand remains broad, but not uniform
The aggregate semiconductor figures are stronger than the individual fund stories might initially suggest. Nearly half of the 6,371 filers in the tally were net buyers of chip stocks, a much higher share than net sellers. But managers did not concentrate their activity in a single company or sub-sector.
Some bought or expanded exposure to AMD, Intel, TSMC and STMicroelectronics. Others exited Nvidia, Broadcom, Micron and semiconductor-equipment stocks. The spread of decisions reflects the expanding range of businesses affected by AI spending: chip designers compete for accelerator demand, foundries benefit from manufacturing volumes, equipment makers depend on factory investment, and infrastructure companies face rising demand for data-center capacity and electricity.
This makes the sector harder to read through a single headline position. A sale of Broadcom or Nvidia can reflect profit-taking, valuation discipline or a preference for smaller rivals and adjacent infrastructure providers, rather than a negative view on AI spending itself.
Portfolios add exposure beyond technology
The filings also show several managers directing capital toward industries outside the AI theme. Berkshire added airline and homebuilding exposure. Third Point took positions in media, financials and rail transportation. Tiger Global’s additions included infrastructure-oriented companies alongside semiconductor names.
Those decisions reduce the degree to which portfolios depend on a narrow group of technology leaders. They also indicate that some managers are seeking businesses tied to consumer demand, housing, transportation or company-specific restructuring opportunities while preserving selected AI exposure.
Second-quarter filings therefore portray an institutional market that remains engaged with AI-linked equities but less willing to treat technology as a single trade. Semiconductors retained the strongest broad support, while Alphabet’s sharply divergent treatment and the rotation among chip names show that portfolio managers are increasingly choosing where in the AI buildout they expect returns to be concentrated.
To see how AI is reshaping broader digital assets, explore our web3, AI and crypto market dynamics overview.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
