Mutual funds underweight, institutions return to net long positions with real capital! Mag 7 led by Nvidia (NVDA.US) returns to the spotlight of capital flows

According to reports from Zhitong Finance APP, based on the latest chart compiled by Wall Street financial giant Goldman Sachs and shared by renowned U.S. financial journalist and investment analyst Mike Zaccardi, mutual fund institutions focused on U.S. large-cap stocks remain significantly underweight in the current $5 trillion market cap “AI chip superpower” Nvidia (NVDA.US). The chart shows that among the AI-related stocks tracked, Nvidia experiences the most notable underweighting, with large-cap mutual funds holding an average underweight of about 100 basis points.

Other AI computational power industry chain leaders are also underweighted by some mutual fund institutions. For example, Nvidia’s strongest competitor in the PC and high-performance computing chip sectors—another GPU technology leader, AMD (AMD.US)—is also among the most underweighted stocks, with an underweight of about 60 basis points. Among the “Magnificent 7” U.S. technology giants, both Google (GOOGL.US) and Microsoft (MSFT.US) are unusually underweighted, by roughly 70 and 50 basis points, respectively.

Meta Platforms (META.US), Facebook’s parent company, is underweighted by about 30 basis points. U.S.-based memory chip giant Micron Technology (MU.US) is an exception, as U.S. large-cap mutual funds are overweight by approximately 40 basis points.

The chart also shows that several hot AI computational power stocks are unexpectedly on the mutual fund “overweight” side, including Seagate Technology (STX.US), the U.S. HDD storage giant, overweighted by about 15 basis points, and Sandisk (SNDK.US), the leading U.S. NAND flash memory provider with a 575% year-to-date gain—the highest in the S&P 500 index—underweighted by about 25 basis points on the chart's scale.

Meanwhile, Goldman Sachs’ compiled statistics also show that in August, the historically hottest U.S. market sector—TMT (Technology, Media, and Telecom)—saw capital flows almost entirely driven by actively increased long positions, not by short covering. The “Magnificent Seven” net allocation benchmark indicator rebounded from about 14% to 18%, strongly indicating that risk capital is returning to the AI computational power investment theme. The underweighting of Nvidia, AMD, Google, and Microsoft relative to the benchmark by large-cap mutual funds means that, should these stocks continue to outperform and the AI computational power industry chain is further boosted by catalytic events, there remains considerable potential capital demand from funds needing to cover positions to reduce tracking error and relative performance pressures.

After the summer lows, the “Magnificent Seven” and TMT stocks dominate major capital inflows in August

Goldman Sachs’ recent prime brokerage compilations show that, in August, Wall Street institutional investors have been significantly returning to U.S. large-cap technology stocks, with technology, media, and telecom (TMT) stocks leading this wave of capital re-entry.

So far this month, the TMT sector has posted the largest net purchases among all U.S. equity market sectors. Remarkably, this capital inflow is almost entirely driven by institutions proactively taking long positions, rather than simply unwinding shorts. This suggests a rapid resurgence of investor optimism about the growth prospects of these technology-focused sectors.

This aggressive buying is also flowing into the seven mega-cap “Magnificent Seven” U.S. tech giants—Google (GOOGL.US)(GOOG.US), Amazon (AMZN.US), Apple (AAPL.US), Meta (META.US), Microsoft (MSFT.US), Nvidia (NVDA.US), and Tesla (TSLA.US). After a pullback in early summer, these tech giants have seen significant net inflows since the beginning of August. The data shows that hedge funds and other institutional investors have increased their net allocation to the “Magnificent Seven” to about 18%, a significant rebound from the year’s low of about 14% at the end of June.

The so-called “Magnificent Seven”—Nvidia, Apple, Microsoft, Google, Tesla, Amazon, and Meta Platforms (the parent of Facebook)—constitute over 40% of the S&P 500 and Nasdaq 100 indices, serving as the core driving force behind the S&P 500’s record highs. They are regarded by top Wall Street investment institutions as the group most capable of delivering significant returns to investors amid the largest technological transformation since the Internet era.

The S&P 500's “AI super bull market” has surged by more than $30 trillion over the past three years, largely driven by these global tech giants, seen as the biggest beneficiaries of the AI boom. This rally is also strongly supported by chip companies (such as Micron, TSMC, and Broadcom) that benefit from massive global AI computational infrastructure investments, the “big three” of storage products (Sandisk, Western Digital, and Seagate), and power system suppliers (like Constellation Energy).

Globally, under the “AI faith” narrative, massive capital continues to flow specifically towards AI computational infrastructure construction and the revenue-generating AI application track—currently the two hottest investment sectors. For the U.S. stock market and the MSCI global equity benchmark index, both repeatedly setting record highs and now entering new long-term bull markets, the increasingly fervent “AI faith” has been the central and strongest bullish driver. As long as this “AI faith” trend remains strong and continues to sweep global equity markets, both U.S. and global bull runs are set to continue their frenzied trajectories.

From Micron’s supply-demand gap to Nvidia’s underweighting—the AI capital frenzy is far from over

The AI computational power trading theme has not reached its market-wide allocation limit, and there remains a high probability that further capital will continue to flow into this theme. However, this next phase will shift from broad-based gains driven by premium on computational power, to structural overweighting driven by order visibility, free cash flow, and irreplaceable bottlenecks.

According to Wall Street strategists, the most important change in the global semiconductor sector is that the “AI crowded trade unwind” in July is increasingly appearing to be a rapid unwinding of extreme leverage and speculative positions, rather than a fundamental reversal for the AI computational power industry chain. Latest market data shows the Philadelphia Semiconductor Index (SOX) plunged nearly 29% from its June 22 historic high to the July 29 low, but then rebounded almost 20% in just three weeks—technically re-entering bull market territory in the short term.

South Korea, known as the “AI computational power barometer,” amplified this “fundamentals intact, AI leverage burst” semiconductor rally: the benchmark KOSPI index rose 11.5% last week, ending a seven-week losing streak, with Samsung Electronics and SK Hynix surging 19% and 16% respectively. From the 5,593.56 close on July 30 to 6,977.94 by August 14, the KOSPI rallied about 24.75%. Notably, South Korean single-stock leveraged ETF assets plummeted from about $50 billion to $17 billion, and JPMorgan predicts hedge fund deleveraging is roughly 90% complete. With foreign capital net buying of about 3 trillion KRW on August 14, the global semiconductor rally is shifting from “forced selling” to institutions like hedge funds resuming risk exposures. If underallocated funds chase further, a “rally—covering—short squeeze—further rally” bull market feedback loop is possible.

Goldman Sachs data shows large-cap mutual funds are underweight Nvidia by about 100 basis points, and also underweight AMD, Google, and Microsoft. This does not mean the funds are bearish or hold net shorts—it simply indicates their holding weight is below the benchmark. Should these leaders continue to outperform, funds will face pressure from tracking error, relative performance, and career risk, which could trigger “benchmark-chasing” buy action. Meanwhile, Micron is already overweighted by about 40 basis points, suggesting that capital is not merely chasing the entire AI sector indiscriminately, but is flowing preferentially to more certain areas like storage bottlenecks. Combined with institutions’ net allocation to the “Magnificent Seven” rebounding from about 14% at end-June to 18%, mainly driven by proactive long buying, it’s clear the process of restocking the AI computational power theme has begun, but is still in progress.

Fundamentals support this capital reversal as earnings backstop: Micron’s customer demand is approximately 150% of its promised supply, implying a supply shortfall of about 50%; five-year strategic agreements further transform volatile spot demand into long-term order visibility. As HBM, DRAM, and NAND together determine GPU utilization, inference throughput, and per-Token costs, memory is upgrading from a standardized commodity to a system-level infrastructure that must be co-designed with accelerators. This benefits product mix, bargaining power, and profit margin stability. Micron’s $10 billion AI storage R&D investment also shows management is allocating capital based on long-term structural needs, not short-term cycle peaks.

Nvidia’s robust fundamentals and its influence on the entire AI computational power industry chain further reinforce a powerful “underweight + valuation discount” catalyst: Bank of America analysts believe that even accounting for about $300 billion in ecosystem investments, guarantees, and backstops, a sum-of-the-parts valuation for its free cash flow still indicates the stock is undervalued by 34%–50%, hence maintaining a “Buy” rating and a $350 price target. Thus, if GPU rental prices, cluster utilization, and AI capex remain strong, and Nvidia returns more free cash flow to shareholders via buybacks, institutional restocking could lead to a re-rating of valuations; conversely, if AI project financing cycles, off-balance-sheet guarantees, or capex returns worsen, buying interest may quickly cool.