BofA: Funds rotate back into software and banks in September, while semiconductor holdings continue to be reduced

Which sectors reversed their earlier selling trend?

In September 2026, global long-only funds tracked by BofA bought a net 13.7 billion dollars of software, 8.5 billion dollars of banks, and 6.1 billion dollars of media and entertainment, reversing their selling of these sectors earlier in the year. They continued to sell semiconductors and technology hardware, with net sales of 9.3 billion dollars and 3.9 billion dollars, respectively. The return of funds was selective, and trends also diverged within technology.
The buying of software came mainly from the United States: net purchases of US software stocks in September were approximately 14 billion dollars, slightly above net buying of the global software sector, while other regions combined remained net sellers.
A similar divergence appeared across themes: funds bought AI platforms, quantum computing, and air defense, began selling AI computing infrastructure and photonics, and continued to sell SaaS and luxury lifestyle themes. Buying in the broadly defined software sector can occur alongside selling of the SaaS theme.

In September, long-only funds bought a net 18.3 billion dollars in the United States, 8.3 billion dollars in Japan, and 5.3 billion dollars in Europe, while selling a net 23.6 billion dollars in emerging markets. Japan recorded cumulative net buying of 48.2 billion dollars for the year to date, the strongest figure among the regions compared in the report.
But net buying in the United States was driven by passive funds: in September, passive funds bought a net 49.6 billion dollars, while active funds sold a net 31.3 billion dollars. Europe and Japan also saw passive buying alongside active selling, so aggregate flows masked differences between fund types.
Trading direction must also be considered separately from existing allocations. Active long-only funds are most overweight Europe and emerging markets relative to their respective benchmarks; even though emerging markets were sold during the month, the existing allocation remained overweight.

BofA believes that stocks with many holders and high allocations can continue to outperform as long as their fundamentals remain attractive. It therefore combines holdings with “triple momentum”: earnings estimate revisions, share-price trends, and news sentiment together help distinguish positive from negative catalysts.
The “crowded positive” group combines high fund ownership, an overweight position relative to benchmarks, and positive momentum. In September, the group returned −2.3%, better than the “crowded negative” group’s −5.5%, a lead of 3.2 percentage points. Over the past 12 months, the groups returned 18.3% and 7.9%, respectively, a gap of 10.4 percentage points. The relative advantage came alongside a decline in absolute returns in September.
The current crowded positive list includes SK Hynix, Samsung Electronics, TSMC, Seagate, and Siemens Energy. The momentum used in the screening can change quickly with unexpected events, and the return figures do not deduct transaction costs.

TSMC is on the crowded positive list, but over the past 12 months through the end of September, long-only funds cumulatively sold a net 37.5 billion dollars of its shares. There is no contradiction: overweight measures the current weight relative to a benchmark, while net selling measures the change in the number of shares held over a period.
Active exposure is also affected by benchmark weights and share-price performance. Even without selling, a rise in an underweight stock can widen the underweight, so BofA separately estimates trading flows using changes in the number of shares held, avoiding attribution of all weight changes to buying and selling.

The report’s data generally run through 30 September 2026, but the most common lag in the underlying holdings disclosures is 2–3 months. After the disclosure date, month-end updates mainly revalue holdings using prices and do not include subsequent undisclosed portfolio changes.
Buying and selling amounts are estimated by multiplying the change in disclosed shares held by the month’s volume-weighted average price, and include active and passive funds; the overweight and underweight analysis focuses on active funds. Retail investors, small funds, hedge funds, and sovereign funds are outside the sample, so these figures describe trading by the funds covered.