JP Morgan: The de-leveraging phase for hedge funds has come to a close, and AI computing investments have finally achieved a profitable cycle.

The Three Layers of Underlying Logic Behind the U.S. Stock Market Technology Rebound Overnight, the Nasdaq rose by 3.3%, while the Philadelphia Semiconductor Index hit an 8% gain, and momentum factor baskets increased by 5.4%. Morgan Stanley has broken down the core support of the market-wide rebound into three tangible industry signals. The extreme position reduction in the market that persisted for two months is returning to normal. The extreme selling data for technology and Korea-related holdings in Q2 has already been repaired to a historically neutral range. The continued pressure from institutional funds is no longer amplifying, which is the most fundamental funding condition for this rebound. OpenAI's operational data has provided positive catalysts for the entire industry's AI commercialization. After the enterprise tool ChatGPT Work was implemented, the companys revenue scale in July exceeded the overall level of Q2, with the monetization speed of AI applications surpassing previous market expectations. The large model sector had long been trapped in a cycle of high costs and slow revenue realization; a single month's revenue exceeding that of an entire quarter can directly alter downstream cloud providers' judgments on expanding computational power. Reports from large-scale cloud service providers have provided empirical proof that capital investment can be profitable. Microsoft Corporation (MSFT.US) recorded its highest single-day increase since 2008, largely driven by Azure, which delivered its fastest growth in four years. The management's subsequent guidance indicates that growth will continue to accelerate. Meanwhile, Amazon.com, Inc. (AMZN.US) surged 8% in after-hours trading, with AWS's year-on-year growth rate hitting 37%, higher than the consensus expectation of 35%. The company also raised its capital expenditure target for 2027 from $200 billion to $220 billion, with management stating that AWS demand is saturated and that existing computational power cannot cover customer orders, necessitating additional hardware investments. In the U.S. stock market sector breakdown, the storage chip sector saw a 19% increase, and hardware associated with computational power rose by 12%, while the semiconductor sector overall outperformed the software sector. This order of increases corresponds precisely with the sequence of transmission within the industry chain, where demand for computational power initially drives growth in storage and chip hardware, and the monetization of software applications experiences a lag. ARM (ARM.US) also rose by 7%, with data center business growth continuing to rise. There were also signals of differentiation within the same trading day, as Meta (META.US) saw a single-day drop of 8%. The core issue is the company's disclosure of significant long-term capital expenditure plans without a clear accompanying path for revenue realization; the market's tolerance for pure input without returns has sharply declined. Apple Inc. (AAPL.US) slightly dropped before the market closed and declined further by 6% after hours, as the component supply gap continues to suppress end sales expectations. Starbucks Corporation (SBUX.US), stepping away from the tech sector, achieved a 2% gain based on its profitability and earnings data. Macroeconomic data also provides a mild environment for risk assets. The U.S. core PCE month-on-month growth rate was 0.1%, with the consensus expectation being 0.2%. The weakening inflation reading led to a decline in U.S. Treasury yields, with the two-year yield down 2 basis points to 4.25% and the ten-year yield down 1 basis point to 4.67%. The VIX index fell by 17%, dropping back below 20, partially correcting the panic pricing resulting from the Federal Reserve's policy actions. Crude oil prices also declined, creating a friendly macro environment for growth assets. Hedge Fund Position Data: Deleveraging Has Completed Its Bottom Morgan Stanley's proprietary trading team provided three clear signals in their position monitoring data, showing that the z-score for the combined holdings of hedge funds and factor performance in the entire market has dropped to -2, which is historically at an extremely low level. The reduction rate for North American hedge funds over five days correlates to three times the standard deviation, while over four weeks, it corresponds to two times the standard deviation, indicating that the continuous and large-scale selling behavior is reaching its end. The momentum sector has become a focal point for hedge funds' concentrated sell-offs, with outflows over a twenty-day period corresponding to -2.2 times the standard deviation, highlighting a notable phenomenon of crowded trades collectively closing positions. Over the rolling twenty-day window, long positions decreased by 6.8%, while short positions only shrank by 4.6%, resulting in a differential of 2.2 percentage points. The disparity within the technology sector is even greater, with long positions declining by 30% and shorts only down by 11%, resulting in a 20 percentage point difference. Comparatively, in June, this differential was positive at 40 percentage points. A direct comparison clearly shows a reversal in funding behavior; previously, the market was heavily focused on long positions in tech growth, but entering July saw a concentrated cutting of longs while retaining shorts. After the extreme liquidations were complete, market sell pressure naturally subsided, which constitutes the report's reasoning for concluding that "momentum sector holdings have bottomed out." The Current Industry Fundamentals of the Emerging Market After a 15% Decline The MSCI Emerging Markets Index has declined by 15% from its peak in June, with adjustments concentrated in technology and semiconductor-related assets. Morgan Stanley's emerging market strategists have provided a historical reference: a 10% pullback during an emerging market bull cycle is a common phenomenon, with an average single adjustment magnitude of 21%. Since the onset of this current upward cycle in October 2022, there have already been five instances of similar deep adjustments. During this adjustment process, both retail investors and hedge funds accelerated their position reductions, with the adjustment rhythm displaying characteristics of panic selling. The report assesses that prices are now close to the phase bottom. In terms of valuation, the index currently has a price-to-earnings ratio of 10 times, deviating one standard deviation from the historical average. Since 2015, this valuation point has triggered price rebounds multiple times. Regarding earnings growth forecasts, the expected growth rates for earnings per share in 2026 and 2027 are 65% and 24%, respectively, showing significant discrepancies compared to developed markets. The report breaks down three categories of industries with changing fundamentals: the first category includes oversold momentum assets, such as Samsung, SK Square, Accelink Technologies, and Cambricon; the second category consists of leading manufacturing and financial companies with stable growth attributes, including Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR (TSM.US), Delta Electronics, ASE Technology Holding Co., Ltd. (ASX.US), and India's ICICI Bank (IBN.US); the third category comprises yet-to-recover platform companies such as Tencent and Trip.com Group Ltd. Sponsored ADR (TCTP.US). On the eve of the Bank of Japan meeting, the Japanese yen rose 2.5% due to intervention strategies. The report coincided with the Bank of Japan's monetary policy meeting, shortly after a rate hike was completed last month. The market consensus is that this meeting will maintain interest rates. Attention is focused on the future path of rate hikes, and whether Governor Ueda will signal a September increase; currently, market pricing implies only a 30% probability of a September hike, while Morgan Stanleys baseline view is that the window for rate hikes will be in October, aligning with full pricing. The yen rose by 2.5% against the dollar overnight, with the Nikkei newspaper reporting that the Japanese government and central bank intervened to support the yen, while U.S. regulators conducted market inquiries to verify dollar-yen quotations. Morgan Stanley expresses skepticism regarding the sustained effectiveness of this intervention, mainly constrained by the difference in real interest rates between the U.S. and Japan. As long as the Bank of Japan's rates cannot match domestic inflation levels, the yen is unlikely to experience sustained upward momentum, and asset prices in Japanese banks will exhibit volatility during the phase of exchange rate adjustment. U.S. AI Infrastructure Sectors New Pricing Rule: Capital Expenditures Must Align with Return Visibility Morgan Stanleys analysts on U.S. stocks have proposed a new pricing logic for the AI infrastructure sector. The current rebound in U.S. stocks is largely driven by position repair, as prolonged deleveraging by hedge funds at three times the standard deviation had driven asset prices down. Persistently rising long-term U.S. Treasury yields will compress the valuations of growth assets through discounting models. This does not mean that in a high-interest-rate environment, growth assets lack upward potential entirely, but the market has introduced previously absent constraints into pricing. High financing costs compel companies to establish stronger controls over capital expenditures. The market will actively differentiate between two types of companies: cloud providers with clear return paths and sustained growth statistics will receive positive pricing, while companies with merely large-scale investments and no visible revenue transformation paths will face valuation pressures. The different trends of Meta (META.US), Microsoft Corporation (MSFT.US), and Amazon.com, Inc. (AMZN.US) serve as direct examples. GEO Group Inc. and Three Long-Term Industrial Threads of Regional Macro Strait Control Strategy in the Strait of Hormuz: Shifting from Blocking Shipping Lanes to Collecting Service Fees In traditional market understanding, Irans control over the Strait of Hormuz has only one option: a complete blockade of shipping lanes. The report proposes a new operational path, stating that Iran will not cut off shipping but will instead charge service fees for regulated access to the lanes. Turkeys existing shipping lane management model may serve as a legal reference, where the fees will not be defined as tolls but as navigation guarantees, vessel scheduling, escort emergencies, and environmental governance service fees. Iran plans to collaborate with Oman to establish a complete legal framework, emphasizing that the fee standards apply equally to all shipping entities, with funds used for the safety and ecological maintenance of the waterways, in an effort to gain tacit approval from the United Nations and the International Maritime Organization. This model will change the cost structure of global oil transportation. The U.S. 301 Tariff Policy Will Not Exit in the Short Term; It Represents a Long-Term Baseline Tax Rate The multiple rounds of U.S. tariff policies have distinct phases, with the initial implementation relying on the International Emergency Economic Powers Act and Clause 122, and the current enforcement stage based on Section 301 investigations, which sets tariffs in the range of 10%-12.5% against the worlds six largest trading partners, ostensibly due to those economies' failure to implement bans on forced labor goods. GEO Group Inc. analysts bluntly state that forced labor is merely a legally feasible excuse; the U.S.'s underlying demand is to set a fixed baseline tariff range against major trading partners. This tariff will become a long-term policy baseline; even if trading partners adjust in response to U.S. concerns, there will be no short-term tariff reductions. Julys ZZJ Meeting and the Policy Context for the October Fifth Plenary Session The July ZZJ meeting concluded economic policy for the second half of 2026, maintaining a stable policy orientation and focusing on the execution of existing fiscal policies without introducing any new large-scale expansion tools. The policymaking level has made it clear that the economy is currently under downward pressure, allowing for the introduction of counter-cyclical adjustment tools but emphasizing targeted and precise support measures. Priorities for industries remain locked on AI and manufacturing upgrades, while consumer-side support policies are centered on optimizing supply-side initiatives, promoting trade in services, and balancing foreign trade development to counter external pressures arising from global trade imbalances. The core topics discussed at the Fifth Plenary Session in October will revolve around improving discipline among cadres, corresponding to the local leadership reshuffle cycle; this political backdrop will amplify market risk aversion and may slow the pace of investment recovery. Panorama of Monthly and Year-to-Date Yield Divergence Across Asia-Pacific Assets Monthly Yield Divergence of APAC Industry Indices The oil refining, gold mining, oil production, and shipping sectors recorded positive monthly yields, with oil refining increasing 21% in a single month and domestic gold-related indices rising 30%, ranking them among the top performers across the region. The semiconductor, optical communication, AI hardware, and storage chip sectors have seen significant monthly pullbacks, with the storage chip sector retracting 45% in one month, optical communication down 42%, and semiconductors overall down 36%. This data contrasts sharply with the earlier rebound of the U.S. semiconductor sector and reflects regional differences in funding structures; the U.S. market is mainly focused on institutional position repairs, whereas APAC semiconductor holdings were previously overcrowded and are continuing to digest excess shares. Complete industry yield data is sourced from Bloomberg terminals, covering five time dimensions: one month, three months, year-to-date, one year, and two years, clearly illustrating the market shifts between cyclical and technology hardware sectors. Comparison of Asset Returns By Country/Region Total Monthly Asset Returns by Country (in USD) On a monthly basis, assets in Singapore, Indonesia, mainland China, and Hong Kong recorded positive returns, while Korean assets experienced the highest market pullback, with a single-month decline exceeding 30%. Taiwan, the broader Asia-Pacific region, the Nasdaq, and the emerging markets index weakened in conjunction, closely related to the significant pullback of Korean semiconductor assets. Total Year-To-Date Asset Returns by Country (in USD) From a year-to-date perspective, the market sentiment has completely reversed course, with assets in Korea, Taiwan, and Thailand performing well overall. In contrast, Indonesian assets have experienced the most significant pullbacks, with Chinese, Indian, and South African assets maintaining small negative yields for the year. Developed markets and various European countries have stayed minimally positive for the entire year. A comparison of these two charts reveals the rotational characteristics of the Asia-Pacific market in 2026; the first half saw continuous uptrends in semiconductors and Korean manufacturing, followed by a concentrated deep adjustment in July, with cyclical commodities like oil and gold becoming switches for capital flow during the adjustment phase. Morgan Stanley's Two Major Asia-Pacific Industrial Conferences in the Second Half The 11th Annual India Industry Summit will be held in Mumbai from September 21 to 22, with participants including policymakers and business leaders from various countries, covering 100 enterprises and over 1,000 attendees. The main venue will feature keynote speeches, industry roundtables, and one-on-one business exchanges, followed by three field visits to industries. The Technology Innovation and Growth Summit will take place in Hong Kong from November 16 to 18, lasting three days, with the first two days dedicated to keynote shares and industry forums, providing an open platform for investor-business matchmaking. It is expected that 300 enterprises and 2,000 industry practitioners will attend, with core topics centered on industry changes in the Asia-Pacific region. This report is sourced from "Research Memo Updates," written by Lao Jin, with editing by Zheng Yuyang.