Behind the $13.9 trillion in assets under management: Why is BlackRock adjusting its AI stock allocation?

On July 8, 2026, Rick Rieder, Chief Investment Officer of BlackRock, the world’s largest asset management firm, revealed in an interview with CNBC that the firm has moderately reduced positions and rebalanced its portfolios in some companies with a high direct linkage to artificial intelligence. As of March 31, 2026, BlackRock’s total assets under management reached a record high of $13.9 trillion. This scale exceeds the GDP of most countries, and every move it makes is enough to move the nerves of global capital markets.

In the interview, Rieder made it clear that this adjustment is a “rebalancing operation, not a reversal.” He added that the team trimmed positions in companies whose earnings rely most on AI development, while also reducing overall equity exposure. BlackRock emphasized that dynamic portfolio adjustments are an ongoing process designed to optimize risk-adjusted returns. This localized rebalancing within the AI theme does not represent a bearish view on the sector; rather, it reflects a tactical allocation optimization based on the current market environment.

This statement has drawn widespread attention not only because of BlackRock’s sheer scale—managing more client assets than any competitor—but also because it comes at a special point in time. Over the past year, the AI sector has experienced unprecedented gains. Since September 2025, the Philadelphia Semiconductor Index has risen cumulatively by about 123%, but since setting an all-time high in June 2026, it has corrected by nearly 14% in total. The market’s core contradiction is shifting from “Does AI have a future?” to “Has AI’s valuation already priced in that future?”

Rebalancing, Not Bearish: The Triple Logic Behind BlackRock’s Reduction

To understand BlackRock’s reduction, it is necessary to distinguish three different logics: portfolio rebalancing, profit-taking and risk control, and industry weight adjustment. These three are interconnected, but they point to different aims.

Portfolio rebalancing is standard practice in institutional asset management. When the weight of a certain asset in a portfolio passively exceeds the target due to price increases, institutions reduce holdings to bring the proportion back into a preset range. BlackRock’s reduction this time targets “those companies whose earnings depend most on AI development.” These are the names that saw the biggest gains over the past year, and their weights in the portfolio are likely to have deviated from the initial settings. Rieder described this change as “trimming the winners, not exiting the theme.”

Profit-taking and risk control are another layer of consideration. The valuations of some AI-related companies have climbed sharply, prompting the management team to adopt a more prudent approach to position adjustments. In a CNBC event in June, Rieder previously pointed out that the current price-to-earnings ratio of the “Magnificent Seven” is about 26x, with expected earnings growth rates exceeding 20%. Although valuations are not at bubble levels supported by earnings growth, market concentration risk is rising—investors have begun to discuss whether the market’s concentration on a few AI winners is excessive.

Reducing the weight of a single industry is the direct result of rebalancing. BlackRock reduced its exposure to AI-related stocks and overall equities. Rieder noted that the company may redeploy capital to beneficiaries of AI adoption with lower costs, including power producers, industrial companies, and infrastructure builders—areas that may be able to capture the next wave of data center spending.

Together, the three logics form a complete picture of BlackRock’s reduction: not a bearish view on AI, but an active optimization of the risk-return structure after a major rally in the AI sector.

Wall Street’s Divergence and Consensus: AI Investing Enters the “Selection Era”

BlackRock’s rebalancing is not an isolated event. In early July 2026, multiple Wall Street institutions spoke out intensively. Although there are differences in specific strategies, a consensus is forming: AI investing has moved from the previous “broad-based sector rally” into a new phase of “selecting leading players.”

Goldman Sachs said in its global equity strategy report released on July 7 that AI chip trading has entered a stage that is even more focused on selection, and it does not recommend continuing to “buy a basket” of semiconductor stocks. The firm continues to like the sub-sectors such as CPU, ASIC, memory, and semiconductor equipment, and it also specifically names AMD and Applied Materials. Goldman Sachs simultaneously forecasts that from 2026 to 2031, global AI capital expenditures centered on computing, data centers, and electricity will reach approximately $7.6 trillion, with annual spending rising from $76.5 billion in 2026 to $1.64 trillion in 2031. Hyperscale cloud providers’ AI investments may exceed $6 trillion by 2030.

JPMorgan has taken a more proactive stance. Strategist Mislav Matejka said on July 6 that the recent pullback in semiconductor stocks should be viewed as a buying opportunity. The bank’s priority for positioning in the technology sector is clear: “semiconductors outperform hyperscale cloud operators, and hyperscale cloud operators outperform high-risk AI concept stocks.” JPMorgan believes that AI chip demand is still in a long-term upward cycle, and that new capacity is not expected to be released meaningfully until around 2028, keeping the industry’s supply-and-demand landscape healthy. The firm expects global stock markets to hit new highs in the second half of the year.

Morgan Stanley is more cautious. In a report dated July 6, strategist Michael Wilson said that as investors pull back from the strongest-performing technology stocks this year, the U.S. market may face headwinds in setting new all-time highs again. The firm believes capital is rotating from chip stocks to hyperscale cloud providers—including companies such as Microsoft, Amazon, and Meta. Morgan Stanley recommends that investors place more importance on the realizability and quality of earnings. The firm maintains its year-end S&P 500 target at 8,000 points.

Bank of America and UBS both maintain an optimistic outlook on the long-term favorable cycle for AI semiconductors. Bank of America believes the industry is still in the middle stage of an 8-to-10-year growth cycle; UBS said that the long-term investment logic for AI has not changed, and short-term volatility in the semiconductor sector instead provides opportunities for long-term investors to build positions gradually.

Overall, Wall Street’s divergence is concentrated in “what to buy” rather than “whether to buy.” Goldman Sachs advocates selecting leading hardware sub-sectors, JPMorgan suggests positioning in semiconductors on dips, and Morgan Stanley leans toward rotating into cloud service providers—but no institution suggests fully exiting.

From Concept Trading to Earnings Trading: The Deeper Evolution of AI Investment Logic

BlackRock’s rebalancing and Wall Street’s divergence both point to a deeper shift: the driving logic of AI investing is moving from “theme narratives” to “fundamental verification.”

Over the past two years, the rise of the AI sector has mainly relied on two narratives: the exponential growth in large model parameter scale, and the explosive expansion of demand for computing power. These narratives still hold true. Goldman Sachs expects that in 2026, global hyperscale cloud providers’ capital expenditures will exceed $760 billion, equivalent to about $2 billion per day; UBS’s model predicts that in 2026, HBM demand will be equivalent to 8.5 million Nvidia AI GPUs; the World Semiconductor Trade Statistics organization forecasts that the global semiconductor market size could reach $1.51 trillion in 2026.

But the market is raising new questions: can massive AI capital expenditures translate into sufficiently rich profits? This question is driving more capital to move from mega-cap technology stocks into a broader range of stocks. Morgan Stanley pointed out that although mega-cap technology stocks posted strong performance in the third-quarter earnings, the rise in stock prices has already lagged noticeably, leading valuations to decline. The market wants convincing evidence that massive AI capital expenditures can translate into sustained returns—not just ever-increasing spending figures.

This is the essence of “moving from concept trading to earnings trading.” In the first stage, the market rewarded any AI-related names—“if it’s AI, it goes up.” In the second stage, the market begins to distinguish who truly benefits from AI commercialization and whose earnings can actually be realized. The latest statements from Goldman Sachs, JPMorgan, Morgan Stanley, and BlackRock all point to this shift.

Which AI Sub-Sectors Are Still Worth Watching?

Within the framework of the “Selection Era,” capital is more likely to flow into areas that truly benefit from AI commercialization. Based on views from multiple institutions, the following directions are worth paying attention to:

AI chips (GPU and ASIC). Goldman Sachs believes that areas such as CPU and ASIC benefit more directly from the expansion of AI infrastructure, with higher visibility into demand. The firm expects the penetration rate of ASICs in AI servers to rise significantly in 2026. JPMorgan designated Broadcom as a “strong buy” for the remaining portion of 2026.

High-bandwidth memory (HBM). AI training and inference continue to drive demand for high-end storage such as HBM. In 2025, the HBM market was basically dominated by SK hynix. Together with Samsung, it holds more than 80% of the global HBM market share. UBS predicts that total HBM industry demand in 2026 will grow 90% year over year.

AI data centers and cloud computing. Morgan Stanley believes capital is rotating from chip stocks to hyperscale cloud providers. HSBC believes that after an adjustment of about 20%, valuations for AI hyperscale cloud service providers have regained attractiveness. Goldman Sachs expects that by 2026, the combined share of capital expenditures in four major global areas—data centers, semiconductors, utilities, and defense—in total capital expenditures will rise sharply from 25% in 2022 to over 40%.

Semiconductor equipment. Long-term equipment procurement plans of leading global companies lock in long-term demand, and the semiconductor equipment sector has strong certainty for an upward cycle in 2026 to 2027. Goldman Sachs likes Applied Materials’ advantages in advanced process and memory capital expenditure.

Power and infrastructure. The expansion of AI computing power is continuously reshaping energy demand, and growth in electricity usage for data centers has become an important incremental driver of global electricity demand. Rieder himself also noted that power producers, industrial companies, and infrastructure builders may be able to capture the next wave of data center spending.

Conclusion

BlackRock’s trimming of some AI stocks is not a signal of the end of the AI rally; it is a landmark event marking AI investing’s transition from the first phase to the second phase. The world’s largest asset manager with $13.9 trillion in assets rebalances; Goldman Sachs recommends “selection rather than buying a basket,” JPMorgan calls for “buying on dips,” and Morgan Stanley advocates “rotating to cloud service providers.” Although these views appear to differ, they share the same premise: AI’s long-term logic has not changed, but the way to invest needs to be adjusted.

For investors, this means they can no longer simply “buy the whole sector” and expect a broad-based rally. Future returns will depend more on in-depth assessment of company fundamentals, the ability to deliver earnings, and the visibility of demand across sub-sectors. AI remains one of the most important technology trends of the next decade, but the trend’s beta returns are giving way to alpha returns from individual stocks.

As Rieder said in his January outlook, 2026 will “reward earnings and selective allocations.” This judgment is being validated by the market.

FAQ

Q: Does BlackRock’s reduction of AI stocks mean it is bearish on the AI sector?

No. BlackRock’s Chief Investment Officer Rick Rieder explicitly stated that this reduction is a portfolio rebalancing operation, not a reversal. The company emphasized that this is a tactical allocation optimization based on the current market environment, and it is not a denial of AI’s long-term outlook.

Q: How do Goldman Sachs, JPMorgan, and Morgan Stanley differ in their views on AI investing?

Goldman Sachs advises against “buying a basket” of semiconductors and favors sub-sectors such as CPU and ASIC. JPMorgan believes semiconductor pullbacks are a buying opportunity and likes the long-term demand for chips. Morgan Stanley suggests rotating capital from chip stocks into hyperscale cloud providers. The three institutions agree that the long-term outlook for AI remains positive, but they differ in the specific allocation directions.

Q: Has AI investing truly entered the “Selection Era”?

The latest statements from multiple Wall Street institutions confirm this trend. The market is shifting from “AI concept trading” to “AI earnings trading”—funds are no longer blindly chasing every AI-related name, and are paying closer attention to a company’s ability to deliver earnings and the quality of its fundamentals.

Q: How big is the scale of AI infrastructure investment?

Goldman Sachs estimates that from 2026 to 2031, global AI capital expenditures centered on computing, data centers, and electricity will reach approximately $7.6 trillion, with annual spending rising from $76.5 billion in 2026 to $1.64 trillion by 2031. Hyperscale cloud providers’ AI investments may exceed $6 trillion by 2030.

Q: What role do crypto assets play in institutional investment portfolios?

In its broader portfolio guidance, BlackRock recommends allocating 1% to 2% of bitcoin alongside top AI stocks. The institution positions bitcoin as a “complementary diversification tool” for portfolios, believing that a moderate allocation may influence a portfolio’s return potential while maintaining an appropriate level of risk tolerance.