13F New Signal: AI Has Not Ebbed, Wall Street Is Just Becoming 'Picky'
- The article says 13F filings are lagging disclosures and should not be used for real-time copy trading because they report quarter-end holdings after a filing delay and do not fully reflect shorts or other positions.
- Reuters’ review of Q2 2026 filings from 6,371 institutions found that roughly 44% reduced exposure to the Magnificent Seven while about 42% added or initiated positions, showing a split rather than a broad exit from big tech.
- Institutional positioning in semiconductors remained more bullish, with about 48% of institutions net buying and 34.5% net selling, while major software names showed slightly more net sellers than net buyers.
- The piece argues that AI conviction has not disappeared; instead, investors are increasingly distinguishing between crowded winners and assets with better risk/reward inside the AI trade.
- Berkshire significantly increased its Alphabet stake, while Tiger Global reduced several large-cap tech holdings but added names tied to AI computing and data centers, such as AMD, Applied Digital, Cerebras, Cipher Digital, Core Scientific, and more Intel.
- The article frames the market effect as mixed for mega-cap tech and software, but still supportive for AI infrastructure, semiconductors, and related data-center names.
The article cautions that 13F filings are a delayed snapshot of institutional portfolios, not a real-time trading signal. Because they are filed after quarter end and do not fully capture shorts or other positions, the piece says they are better used to understand how major money has been repositioning over the past three months.
Based on Reuters’ review of Q2 2026 filings, the broad message is that Wall Street is not abandoning AI. Nearly 44% of institutions reduced exposure to the Magnificent Seven, but about 42% added or initiated positions, which points to rotation rather than a wholesale exit. Semiconductor holdings remained more clearly favored, while software showed a more mixed picture.
The report’s central argument is that the AI consensus is still intact, but investors are becoming more selective about where they want exposure. The market is now asking which companies have growth already priced in, which AI spending can become profit, and which crowded positions are most vulnerable if sentiment turns.
Several large investors are used as examples of different approaches. Berkshire increased its Alphabet stake materially, while Tiger Global reduced several large-cap tech positions but added to AI computing and data-center related names. The article frames the market impact as mixed for mega-cap tech and software, but still constructive for semiconductors, AI infrastructure, and data-center linked stocks.
13F新信号：AI并未退潮，华尔街只是变得更“挑剔”
文章首先提醒，13F披露是滞后数据，不适合拿来做实时跟单。由于其在季度结束后才披露，而且短仓等头寸无法完整反映，它更适合用来观察大资金过去三个月的调仓方向。
根据Reuters汇总的2026年二季度13F，华尔街并没有撤出AI。接近44%的机构减持“美股七巨头”，约42%的机构则新开仓或加仓，显示的是轮动而非整体离场。半导体仓位仍然更受偏好，而软件板块的分歧更明显。
文章的核心结论是，AI共识还在，但资金变得更挑剔。市场开始追问：哪些公司的增长已经被提前计价，哪些AI资本开支能够转化为利润，以及一旦市场回调，哪些拥挤持仓最容易先被兑现。
文中以几家机构为例说明这种分化。伯克希尔显著增持Alphabet，Tiger Global则减持多只大型科技股，同时增加了与AI算力和数据中心相关的标的。文章对市场影响的判断是：对大型科技和软件偏中性、分化，对半导体、AI基础设施和数据中心链条偏利好。