美股投资young

美股科技投资者 ai狂热者分享每日快讯. all in ai

299Following

673followers

Feed

$LRCX has declined for 10 consecutive trading days, with a cumulative drawdown of about 22%, $KLAC fell about 17% during the same period, and $AMAT also dropped about 22% within 9 trading days. It is difficult to simply explain such a level of sell-off among leading advanced semiconductor equipment companies as normal profit-taking. On October 21, $LRCX is about to release its new quarterly earnings report, which I believe could be an important basis for judging whether the semiconductor equipment sector can stabilize. $SOXX $QQQ Let's first look at the fundamentals. In the last quarter, $LRCX's revenue reached $6.722 billion, a quarter-on-quarter increase of 15.1%, with a Non-GAAP gross margin of 52% and an operating margin of 38.4%. The company previously guided this quarter's revenue to $8.1 billion, with a fluctuation range of $400 million, indicating that management originally expected the business to continue growing significantly. From these data, at least up to the last quarter, demand for semiconductor manufacturing equipment has not deteriorated comprehensively. $LRCX's competitive advantages focus on etching, deposition, and wafer manufacturing-related technologies. The more complex the advanced process, the more precise the transistor structure, and the higher the requirements for high-precision etching and deposition equipment in the manufacturing process. Especially with the increase in 3D NAND layers, upgrades in DRAM manufacturing technology, and continuous development of advanced packaging, there are opportunities to drive equipment demand growth. This is why I am long-term bullish on companies like $LRCX, $KLAC, and $ASML, which have high technological barriers. But now Wall Street is worried about 2027. If the financing environment for AI data centers continues to deteriorate and downstream customers slow capital expenditures, chip manufacturers may delay some expansion projects. Equipment companies are upstream in the capital expenditure transmission chain; once future order growth expectations are downgraded, even if the current quarter's performance is good, stock prices may continue to be under pressure. Therefore, this earnings report needs to focus on three key questions: whether this quarter's revenue can reach near the previous guidance midpoint of $8.1 billion, whether the gross margin can be maintained around 52%, and whether management's judgment on wafer manufacturing equipment demand in 2027 has changed. Particular attention should be paid to whether investment plans for DRAM, NAND, and advanced processes have been delayed, as well as the impact of demand in the Chinese market and export restrictions. Simply exceeding EPS expectations is no longer enough to dispel market concerns. Personally, I believe that if $LRCX maintains strong growth in this earnings report and management continues to release positive demand signals for 2027, then the valuation compression caused by the previous consecutive sharp declines may see a phased recovery. If companies like $KLAC and $AMAT also stabilize simultaneously, the credibility of the entire equipment sector reestablishing an upward structure will also increase. Conversely, if the earnings report exceeds expectations but the 2027 growth guidance weakens, or the market continues to fall sharply after the positive announcement, then this round of adjustment may not be over yet. I still believe that advanced semiconductor manufacturing equipment is one of the most worthwhile long-term holdings in the AI industry chain, especially optimistic about the technological barriers of $ASML, $KLAC, and $LRCX. But no matter how good the industry outlook is, valuation and profit cycles must be considered. The earnings report on October 21 may provide important clues, but a single report cannot confirm the entire sector's reversal alone. Only when earnings expectations stabilize, stock prices stop hitting new lows, and capital flows back in, will there be more sufficient reasons to judge that the semiconductor equipment sector is emerging from this extreme adjustment.

I'm actually a bit worried about a sudden surge in the US stock market on Monday. $SOXX $QQQ $NVDA $MU $SNDK It's not because I don't want the market to rise, but after experiencing such extreme declines consecutively, once a strong rebound occurs, many people might completely abandon risk management. $SOXX has fallen for 9 consecutive trading days, setting a rare record of continuous decline in nearly 20 years. The Nasdaq dropped 4.5% last week, with semiconductors, storage, optical communications, and software stocks all facing sell-offs. The price structure of the entire tech sector has clearly deteriorated, but the market has also accumulated a considerable degree of short-term oversold pressure. If on Monday South Korea, Japan, and Taiwan continue to be under pressure, but the US stock market suddenly opens low and then rallies, even with $SOXX rebounding 3% to 5% in a single day, I wouldn't find it strange at all. After continuous declines, short sellers taking profits, institutions readjusting risk exposure, and short-term funds chasing rebounds could all drive the market to rise quickly. Another factor that cannot be ignored is options hedging. When market makers are net negative Gamma, price increases may trigger trend-following buy hedges, further amplifying gains. But actual Gamma exposure cannot be fully confirmed just by public holdings, so not every sharp rise should be attributed to market makers being forced to buy. The problem is that a sharp rise and a true bottom are two different things. Having experienced the adjustments in March and July, I am very clear on how fast prices can rebound at the end of extreme market conditions. But precisely because of this, I won't simply assume that October will repeat the patterns of the previous two times. The market environment is different; interest rates, oil prices, credit risk, and AI capital expenditure expectations have all changed. We especially need to be wary of a scenario where the US stock market opens sharply lower on Monday, then quickly rallies, with stocks like $NVDA, $MU, and $SNDK collectively rebounding, and immediately on X there are many voices about a bull market return, short squeeze, and golden pit confirmation. Then the rebound fails to continue the next day, previous support levels become new resistance, and prices fall below the lows again. This is not uncommon in a downtrend. Because funds previously trapped may use the rebound to reduce positions, while genuinely willing new buyers may not have returned yet. So on Monday, I will focus on three questions. First, is the rebound driven only by a few heavyweight stocks pulling the index, or are sectors like semiconductors, software, and cloud computing all starting to strengthen synchronously? If $QQQ rises but most tech stocks remain weak, the foundation of the rebound is questionable. Second, can $SOXX regain and hold the key areas it previously broke below? Once prior support is lost, it often becomes resistance that must be retaken during a rebound. Only sustained recovery, rather than intraday rallies followed by pullbacks, is worth attention. Third, can the rebound gain confirmation in subsequent trading days? A 5% rise on Monday is certainly exciting, but if Tuesday and Wednesday see consecutive declines, the market still hasn't escaped its prior weak structure. I have always been a long-term AI bull and still tend to believe October is a deep pullback within a bull market. But being long-term bullish doesn't mean every rebound is worth chasing, nor does it mean one should bet on a market bottom with short-term options and leverage. I've been repeatedly emphasizing reducing positions recently because the risk-reward structure has changed. For those with heavy positions, if the rebound lacks sustained buying, I am more inclined to use the rise to reduce risk rather than buy back all previously reduced positions just because of a big green candle. For those with ample cash, there's no need to worry about missing the first day of a market rebound. If a new upward trend truly forms, it usually won't give investors only one trading day to act. What I least want to see now is people panic-selling during continuous declines, then chasing the rebound full throttle out of fear of missing out. The experiences in March and July tell us that pullbacks in a bull market can be very fierce, and rebounds can be very rapid. But historical experience is not a guarantee that October will necessarily rise. I still believe the foundation for AI's long-term growth hasn't disappeared, and I believe quality semiconductor companies have opportunities to rise again in the future. But right now, I'd rather miss out on a technical rebound than take on unbearable position risk trying to guess the bottom again. If there really is a surge on Monday, I'll be happy, but I won't change my risk management principles just because of one day's rise. What would truly make me increase my position again is not a big green candle, but the market starting to consistently prove that prior selling pressure is easing, new buying can absorb selling, and the downtrend is changing.

What is currently most frightening for the global semiconductor industry may not be negative news from any single company, but that South Korea, Japan, Taiwan, and the United States are all enduring the same round of capital sell-offs. South Korea falls, then the U.S. falls; after the U.S. falls, South Korea has to digest new selling pressure again. The entire AI industry chain seems to be forming a continuously transmitted negative feedback loop. $SOXX $QQQ $NVDA $MU $SNDK $TSM Look at how extreme last week's market was. On October 15, South Korea's KOSPI plummeted 7.57%, with SK Hynix and Samsung facing fierce sell-offs, and Japan's semiconductor sector was also under pressure. On Friday, the South Korean market finally rebounded 2.05%, but U.S. tech stocks continued to decline, with $SOXX plunging about 4.1% in a single day, marking a rare nine-day losing streak not seen in nearly 20 years. Asia barely caught a breath before the U.S. passed the pressure back. Why do I believe this round of declines deserves high vigilance? Because although the global AI industry chain is distributed across different countries, the underlying profit expectations and capital risks are highly correlated. South Korea's SK Hynix and Samsung supply memory chips, Taiwan's TSMC handles advanced wafer manufacturing, Japan's Tokyo Electron and others provide semiconductor equipment, and the U.S.'s $NVDA, $AMD, and $AVGO occupy the core segments of chip design and computing power demand. On the surface, these companies are in different markets, but in reality, many institutions hold the same AI growth logic. When the market starts doubting whether AI capital expenditures will continue in the coming years, institutions may not just sell one stock but might simultaneously reduce risk exposure across the entire industry chain. Here lies the problem. A decline in U.S. chip stocks may affect Asian investors' expectations for memory, equipment, and foundry sectors; concentrated sell-offs in Asian markets may further impact U.S. pre-market risk appetite. If combined with ETF redemptions, leveraged liquidations, foreign capital withdrawals, and exchange rate fluctuations, what would normally be a routine valuation adjustment could be amplified into a cross-market continuous decline. But I must emphasize, there is currently insufficient evidence to prove that all market sell-offs are passive liquidations, nor should every decline be attributed to malicious institutional short-selling. What we can observe is high price correlation, declining risk appetite, and some institutions reducing previously crowded AI positions. More interestingly, corporate fundamentals have not deteriorated to the same extent within just two weeks. TSMC and ASML recently still delivered strong operational data, and the long-term demand for advanced processes, HBM, and AI infrastructure has not suddenly vanished. But the market is now trading not just this year's earnings but the growth rate over the next few years and how much investors are willing to pay for that growth in valuation. This is also why I have always believed that trading based on news is easy to fall into misconceptions. During the March crash, there were many negative factors; similarly, in July's crash, there were also many negatives; now in October, various explanations are emerging again. News can explain why the market panics but rarely tells you when that panic will end. Ultimately, what determines whether a rebound can sustain is capital support, liquidity, valuation, and earnings expectations. I have always been a long-term AI bull and tend to believe that this still belongs to a deep correction within a bull market. But I will not assume that short-term further declines are impossible just because I am optimistic about AI long-term. Especially now that the global semiconductor industry has formed a highly synchronized weak structure, any market rebound may be affected by continued pressure in other markets. What I most want to see on Monday is not South Korea suddenly surging 5%, nor $SOXX opening with a 4% jump. I hope to see South Korean memory stocks stop hitting new lows, Taiwan and Japan's semiconductor sectors gradually stabilize, and U.S. tech stocks show sustained buying rather than every rebound being swallowed by new selling pressure. If only the U.S. rebounds while Asia continues to plunge; or if Asia just stabilizes and the U.S. breaks down again, then the global AI asset de-risking process may still not be over. So my strategy remains unchanged. If your position is too heavy, actively reduce risk; be especially cautious with short-term options and leverage; those with ample cash should patiently wait for truly competitive companies to fall to reasonable valuations. I believe the long-term AI bull market still has a foundation to continue, but what the market needs now is not more people shouting "buy," but real capital willing to keep buying. When will the global semiconductor industry end this round of continuous sell-offs? I think the most important signal to watch is when Asia and the U.S. no longer transmit new selling pressure to each other but begin to form synchronized and sustained capital support. Until then, don't mistake a sharp rebound for the entire market being safe.

If the Asia-Pacific markets collectively plunge again on Monday, can the U.S. stock market still hold up? I believe the next 24 hours could be one of the most important trading windows since October. $SOXX $QQQ $NVDA $MU $TSM $SNDK Last week, the global semiconductor sector showed extremely volatile performance. On October 15, South Korea's KOSPI plunged about 7.6%, Japan's Nikkei index dropped nearly 5%, and major memory giants like Samsung and SK Hynix faced intense sell-offs. Although the Asia-Pacific markets rebounded somewhat on Friday, U.S. tech stocks continued to decline, with the Nasdaq falling 4.5% for the week, and $SOXX recording a rare nine-day losing streak not seen in nearly 20 years. This means that when Asian markets open on Monday, they will need to digest the selling pressure from the U.S. on Friday again. But I don't think it’s as simple as South Korea falling means the U.S. will definitely fall, or the U.S. falling means South Korea will definitely fall afterward. Currently, the global AI industry chain holdings are highly correlated. South Korean memory, Japanese equipment, Taiwanese foundry, and U.S. chip companies are all affected by the same round of AI risk appetite changes. If institutions collectively reduce risk exposure, it can easily create cross-market negative feedback. But if Asia-Pacific continues to fall on Monday, we must distinguish whether new risks have emerged or if it’s just a catch-up drop from the U.S. sell-off on Friday. I mainly consider three scenarios. First, Asia-Pacific continues to plunge, U.S. futures weaken simultaneously, and $SOXX breaks down further after the open. This would mean the global AI sell-off pressure may not have fully released yet. If the VIX rises in tandem and tech market breadth continues to deteriorate, I would not rush to bottom-fish. Second, Asia-Pacific plunges, but U.S. stocks open lower then rally strongly, even rebounding 3% to 5%. This is entirely possible. After consecutive extreme drops, short sellers taking profits, short-term funds entering, and options hedging adjustments can all drive a rapid market rebound. However, I believe this kind of rebound requires even more sober judgment. Technical rebounds during a downtrend can sometimes be more intense than rallies in a normal bull market. If the rebound lacks sufficient volume and market breadth support, and previous support levels remain unrecovered, a big rally does not necessarily mean the correction is over. Third, Asia-Pacific stabilizes first, Samsung, SK Hynix, and TSMC-related assets stop hitting new lows, and U.S. semiconductor stocks begin to see broad buying. This is the scenario I most want to see. Especially if $SOXX can regain key levels and the rebound is no longer quickly swallowed by selling pressure, the market may gradually form a more reliable bottom structure. Why have I been emphasizing reducing positions recently? Because the issue now is not just how much semiconductors have fallen, but that the entire tech sector’s price structure is deteriorating. $SOXX keeps breaking down, software stocks are also catching down moves, and rebounds from the previous day are often fully erased the next day. It’s not that there is no bottom-fishing capital, but these funds have not yet shown strong and sustained buying power. Moreover, this round of correction has already shown clear signs of existing long holders reducing risk. Data previously disclosed by Goldman Sachs Prime Brokerage indicates that hedge funds’ selling pressure on tech stocks has persisted recently. But this does not mean shorts are not active, nor can we judge all institutional real positions solely by candlestick and options data. I have always believed the short-term risk to watch most closely is concentrated selling pressure after rebounds. Especially in an environment of negative Gamma risk, leveraged liquidations, and insufficient liquidity, price fluctuations can be amplified both up and down. I have been a long-term AI bull and still tend to believe October is a deep pullback within a bull market, not a sudden failure of the entire AI industry logic. But being bullish does not mean you must be fully invested, nor does it mean every plunge should be met with immediate buying. The market will not guarantee you won’t lose 30% or more in the short term just because you are right about the long-term direction. So on Monday, I will focus on three questions: whether selling pressure in Asia-Pacific further spreads, whether U.S. stocks can show sustained buying independent of Asia, and whether $SOXX’s rebound can hold through the close and be confirmed in the following trading days. If Asia falls, then the U.S. falls, then the U.S. rebounds but Asia falls again, the global AI asset de-risking process may not be over yet. The market has already fallen extremely far, so a sharp short-term rebound is not surprising. But I won’t assume all problems are solved just because of one big green candle. I still believe the foundation for a long-term AI bull market continuation exists and that this plunge may ultimately create long-term investment opportunities in quality companies. But the golden bottom may not be today; what’s truly worth waiting for is simultaneous improvement in valuation, capital, and price structure. Until the trend improves, reduce positions when needed, keep cash when needed. Short-term risk control and long-term AI optimism are not contradictory. The bull market can continue, but the premise is that you have enough capital to participate in the next real uptrend.

$100 trillion! Wall Street may be re-evaluating the largest infrastructure investment plan in AI history. But I believe the market has previously overlooked a very practical issue: how many of the data centers announced for construction in the coming years will actually be completed? $NVDA $AVGO $CRWV $VRT $MSFT $QQQ On October 18, the Financial Times published a major investigation pointing out that the expansion of AI infrastructure in the U.S. is encountering increasing real-world obstacles. First, let's clarify a concept: the $100 trillion is not already signed orders, nor is it money already paid by tech giants, but an estimate by research institutions of the AI infrastructure investment demand in the U.S. over the next several years. A study from Columbia Business School estimates that by 2032, AI construction in the U.S. may require investments exceeding $100 trillion, equivalent to about 3.6% of the U.S. GDP annually. How exaggerated is this scale? According to this study's estimate, the economic scale of AI infrastructure construction relative to GDP may even surpass the historical expansions of railroads, electrification, and internet infrastructure in the U.S. But massive investment plans and actual construction completion are completely different matters. There is also a very noteworthy news today. According to British media reports, $MSFT has withdrawn from a data center collaboration with AI cloud service provider Nscale in Loughton, Essex County, UK. The report points out that the project has been delayed multiple times, and the local power grid even expects that adequate power supply will not be available until the 2030s. Microsoft has not publicly confirmed the specific reasons for withdrawal, while Nscale stated it will continue to advance the project and seek alternative power supply solutions. But this incident itself already illustrates a problem. You can announce today the construction of a multi-billion-dollar data center and even sign long-term computing power contracts in advance, but you cannot rely on a business plan to make the local power grid have sufficient power capacity the next day. This is the risk of AI infrastructure construction that is most easily overlooked. In recent years, the market has mainly valued around GPUs, HBM, advanced packaging, optical modules, and server orders. But a large AI data center from planning to actual production must at least solve issues such as land, approvals, power access, equipment supply, construction costs, and project financing. Any delay in any of these links may affect the speed of fulfilling the entire investment plan. More importantly, these links are not independent of each other. If power supply is delayed by a year, the data center may not be able to go into production as originally planned. If the data center cannot go into production, the customer's computing power rental income may be delayed in recognition. Revenue delays, but financing costs may continue to accumulate. If the project still relies on debt financing, the originally reasonable investment return rate may significantly decline. Ultimately, a long-term contract that looks very attractive may not be able to convert into free cash flow within the expected timeframe. This is also why the AI data center financing market has recently started to show pressure. Reuters reported on October 15 that AI infrastructure-related companies such as AirTrunk, Firmus, and SB Energy have delayed or withdrawn their IPO or financing plans. Some investors have begun to more strictly scrutinize valuations, leverage, contract quality, and project execution risks. But this does not mean AI investment has completely stalled. During the same period, some data center mergers and acquisitions and debt financing activities remain active. I believe the market is shifting from simply believing in growth scale to studying whether this growth can truly be realized. So what does this mean for semiconductors? The impact may be more complex than many imagine. For upstream chip suppliers like $NVDA and $AVGO, continuous data center expansion can bring huge demand for computing and networking equipment. For power and cooling infrastructure suppliers like $VRT, data center construction also means long-term business opportunities. But if some projects are delayed and financing conditions tighten, customers may reschedule equipment procurement. Long-term industry demand has not disappeared, but that does not mean every quarter's orders can grow according to the most optimistic forecasts. Especially when stock prices have already priced in demand years ahead, even if actual orders only slow in growth, it may trigger significant valuation compression. This is exactly why I think the October tech stock adjustment deserves serious study. I have always been a long-term AI bull and have not changed my judgment on advanced processes, storage, and AI infrastructure long-term development. But I will not take every announced multi-billion-dollar investment plan as guaranteed future revenue. Going forward, I pay more attention to actual project start rates, power access progress, secured financing, irrevocable customer contracts, and how much cash flow projects can generate after production. For the market, the real risk may not be that AI is unused, but that the entire industry expanded too quickly before, and some projects' construction speed and commercial returns cannot keep up with capital market expectations. Both situations can have equally significant impacts on stock valuations. Therefore, I still believe that we cannot ignore short-term valuation and financing risks just because the AI industry looks promising long-term. When it's time to reduce positions, reduce them; cash and risk tolerance must be preserved. If in the coming years, those companies that truly complete construction, maintain high computing power utilization, and continuously generate free cash flow begin to stand out, I still believe the AI industry will present very valuable long-term investment opportunities. But the market may no longer be willing to pay a high premium for every construction plan that has not yet materialized. What the AI industry needs to prove in the future is not just how many data centers can be built, but whether these data centers can be built on time, put into use, and truly make money.

Anthropic is willing to offer Blackstone a financing yield exceeding 10% just to purchase AI chips. Keep in mind, this is one of the most closely watched AI companies globally, and now even it has to pay such a high capital cost for computing power. $GOOG $AVGO $NVDA $BX $CRWV According to the latest report from the Financial Times, Anthropic is advancing a chip financing plan worth about $60 billion, codenamed Project North Star. This financing is mainly for purchasing Google's TPUs, involving approximately 1.5GW of AI computing power. Of this, $42 billion is senior debt financing supported by Broadcom-related credit, and another $18 billion is subordinated debt. A Blackstone fund plans to invest $9 billion in this, demanding a comprehensive yield exceeding 10%, with the related debt primarily repaid through Anthropic's chip lease payments over the next five years. Note, this does not mean the entire $60 billion is financed at over 10% cost, nor does it mean the deal is fully completed. But the problem is already very clear. AI companies' revenue growth is indeed astonishing, but the capital required to build and acquire computing power is becoming increasingly expensive. Anthropic's Q3 revenue reportedly reached $16.8 billion, a 45% quarter-over-quarter increase. Such growth is outstanding in any industry. However, the capital markets are beginning to ask another question: with revenue growing so fast, why is there still a need for such large-scale financing to acquire computing power? If hundreds of billions of dollars need to be invested annually in purchasing chips, leasing servers, and building data centers, how much free cash flow can these investments ultimately generate? In the past two years, the market mainly focused on GPU supply, HBM capacity, advanced packaging, and data center power. As long as AI demand keeps growing, capital seems willing to pay increasingly higher valuations for the entire industry chain. But now the situation is changing. Broadcom CEO Hock Tan has publicly stated that the key constraint on AI growth in the future may be capital, not hardware supply like memory chips. Anthropic's financing now gives us a glimpse of how capital constraints might manifest. Even if AI models become more powerful and enterprise customers increase, it does not mean every new computing power investment will yield sufficiently high returns. Moreover, there is a market issue that I believe must be reassessed. For upstream players like $GOOG and $AVGO, AI chip procurement and computing power deployment bring business opportunities. But for companies bearing long-term lease payment obligations, future computing power utilization rates, unit inference prices, customer revenue, and equipment depreciation will all affect whether projects truly make money. The timing of upstream revenue recognition and downstream investment recovery may differ by several years. This is why I believe we cannot simply equate the growth prospects of all companies in the AI industry chain. Companies like $NVDA and $AVGO, which have core technology and mature businesses, face different financial risks than those heavily reliant on external financing to build infrastructure. But do not directly interpret a financing yield above 10% as an imminent credit crisis for Anthropic. This deal involves subordinated debt, which inherently carries higher risk. Financing structure, collateral arrangements, and debt seniority all influence the returns investors demand. What truly requires caution is that if the entire industry's financing costs continue to rise while AI service prices keep falling due to competition, some companies' investment returns may face double pressure. I remain long-term optimistic about AI and do not believe there is sufficient evidence now to prove the entire AI industry's demand is collapsing. But being optimistic about technology development does not mean all AI infrastructure investments are worth paying any price. Especially amid the ongoing adjustment of tech stocks and credit markets beginning to reassess AI debt risks, I prefer to focus on companies' actual cash flow generation capabilities. Going forward, I will focus on three key indicators: AI business revenue growth, capital expenditure growth, and the cash returns generated by new investments. If a company's revenue grows 30% but capital investment grows 80%, and financing costs keep rising, relying solely on revenue growth cannot determine if it is worth investing in. This AI bull market may be far from over. But Wall Street has already shifted from one question to another. In the past, the question was: how many GPUs will be needed in the future? Now, more and more investors are asking: who will pay for these GPUs, and ultimately, who will truly make money? AI changing the world can be true, and some AI investments yielding insufficient returns can also be true simultaneously. In the coming years, the valuation gap between companies that can consistently make money and those that can only sustain expansion through continuous financing may widen significantly.

Something very unusual is happening on Wall Street: tech stocks are being sold off wildly, yet global equity funds are still seeing substantial inflows. As the November 3 U.S. midterm elections approach, the market may be reconsidering the next phase of investment direction. $QQQ $SOXX $SPY $XLV Data released by Reuters on October 16 shows that for the week ending October 14, global equity funds had net inflows of about $20.01 billion, marking the fourth consecutive week of inflows. However, during the same period, tech sector equity funds experienced net outflows of about $4.75 billion. This set of data deserves serious study. The overall stock market has not seen a broad withdrawal of funds, but the tech sector is clearly under pressure from capital outflows. Of course, it cannot be directly concluded that all the money selling tech stocks flowed into other sectors. Fund flows cannot fully reflect institutional holdings changes, but at least it shows that different types of stocks are receiving completely different capital treatment. In recent years, AI has been one of the most crowded trades in U.S. stocks. From $NVDA to $MU, from storage to optical communications, many institutions hold similar growth assets. The market is willing to pay high valuations for demand growth over the next few years. But since October, with renewed concerns about AI financing risks, interest rates, and capital expenditure returns, institutions have begun reassessing the risks of these assets. Meanwhile, sectors like healthcare, consumer staples, and utilities, which were not as popular before, have shown clear relative strength. This is why I previously recommended studying $XLV, $XLP, and $XLU. However, I believe what is truly worth preparing for in advance is the next phase of rotation that may occur after the midterm elections. Citigroup once analyzed eight midterm elections from 1990 to 2018. From May to October before the elections, in seven out of eight cases, defensive sectors outperformed cyclical sectors by an average of about 7 percentage points. But from November to April after the elections, in seven out of eight cases, cyclical sectors outperformed defensive sectors by an average of about 6.5 percentage points. Historically, this shift from defense to cyclical sectors has indeed occurred. Why? Because before the elections, investors need to consider uncertainties in fiscal policy, taxation, regulation, and industrial policy, so some funds tend to reduce risk. Once the elections conclude and policy expectations become clearer, the market may refocus on economic growth, corporate investment, and profit expansion. If this change happens again, I will focus on three directions. $XLI Industrial sector. If corporate capital expenditures remain resilient and manufacturing orders improve, companies related to industrial equipment, automation, and infrastructure may see better profit expectations. $XLF Financial sector. If the economy can avoid a significant recession, bank credit quality remains stable, and capital market trading and corporate financing activities recover, some financial firms may benefit. But financial stocks are not better with higher interest rates per se. If high rates expand credit risk, banks may also face pressure. $XLB Materials sector. The materials industry is sensitive to industrial production, construction activity, and global economic demand. If the manufacturing cycle improves, some chemical, metal, and construction materials companies deserve reevaluation. Of course, these three directions do not guarantee gains after the elections. The biggest issue remains the macro environment. If oil prices stay high, inflation is not alleviated, and the Fed continues tightening, then even after the midterms, cyclical stocks may face significant pressure. Conversely, if energy inflation declines, U.S. Treasury yields fall, credit markets stabilize, and corporate profits continue to grow, the market may have more favorable conditions for gains. One more point I always believe should not be ignored. Even if sector rotation occurs, it does not mean AI loses its investment value. I remain long-term bullish on $NVDA, $MU, and the entire semiconductor supply chain, but the current market structure has not yet given a clear stabilization signal. Controlling tech exposure in the short term and continuing to research AI investment opportunities in the long term are not contradictory. I will not assume that because defensive sectors have recently risen, one should hold healthcare and consumer staples forever; nor will I rush to bottom-fish semiconductors just because they have crashed. What is truly worth observing next is whether the tech sector's outflows can ease, whether industrial and financial sectors can continue to attract buying, and whether market risk appetite improves after the elections. Historical patterns are only references and cannot replace current capital and fundamental judgments. What I care about most now is not which sector has risen how much, but who has the ability to continuously generate profits in the next phase. A bull market will not always have AI as the only direction. What really matters is whether you can timely identify changes when the market starts to shift direction, while retaining enough cash and position flexibility.

Semiconductors have fallen so much it makes you question everything, while healthcare and consumer sectors are starting to rise against the trend. With just half a month left until the U.S. midterm elections on November 3, Wall Street may be undergoing a very important sector rotation. $XLV $XLP $XLU $SOXX $QQQ I've been reminding everyone to pay attention to the healthcare sector recently, but this time I want to share a set of historical data that's even more worth studying. Citibank once analyzed 8 U.S. midterm elections from 1990 to 2018 and found a very interesting pattern: From May to October before the election, in 7 out of 8 cases, defensive sectors outperformed cyclical sectors, with an average excess return of about 7 percentage points. From November to April of the following year after the election, the situation reversed. In 7 out of 8 cases, cyclical sectors outperformed defensive sectors, with an average excess return of about 6.5 percentage points. Note, this refers to relative performance, not that defensive stocks must rise 7% before the election or cyclical stocks must rise 6.5% after the election. Why does this phenomenon occur? I believe the key lies in the market's pricing of uncertainty. As the midterm election approaches, investors need to consider possible changes in fiscal policy, taxation, regulation, and government spending. At the same time, if economic growth slows and interest rates remain high, institutions naturally reassess the risks of high-valuation assets. Healthcare, consumer staples, and utilities happen to have some profit drivers different from AI tech companies. $XLV represents healthcare, including large companies like $LLY, $JNJ, $ABBV. People don't stop buying medicine and medical services just because the economy slows down. $XLP represents consumer staples, including companies like $WMT, $PG, $KO. Even in a downturn, there is sustained demand for daily food and household products. $XLU represents utilities, including companies like $NEE, $SO, $DUK. Demand for electricity and basic public services is relatively stable, but their valuations are very sensitive to interest rates and financing costs. These three asset categories are not immune to declines, but their risk sources differ from those of high-valuation AI semiconductors. Recent market performance is also very noteworthy. Last week, the S&P 500 fell nearly 2%, Nasdaq dropped 4.5%, but about 70% of S&P 500 components actually rose. The tech sector faced a sharp sell-off, while healthcare, consumer staples, and utilities showed clear relative strength. This indicates that the current market is not crashing all stocks together, but risk preferences are shifting across different sectors. So what happens after the election? Historically, as political uncertainty fades, the market sometimes refocuses on economic growth and cyclical assets. If by then U.S. Treasury yields fall, credit conditions improve, and corporate earnings expectations stabilize, I will focus on industrials $XLI, financials $XLF, and the previously deeply corrected tech and semiconductor sectors. But it is crucial to distinguish that the end of the election does not mean cyclical stocks must rise, nor does the winning or losing party of Trump directly determine the U.S. stock market trend. In 2018, when Trump experienced his first midterm election, the Republicans lost control of the House, but the U.S. stock market still experienced a sharp decline in December that year. What truly affects the market are Federal Reserve policies, economic growth, and corporate earnings. So I won't mechanically buy or sell stocks based on the calendar. I previously recommended healthcare because AI semiconductors had already shown obvious signs of crowded capital and continuous breakdown risk, while some healthcare companies' profit models offer different risk exposures. Now that healthcare and consumer sectors have clearly risen, there is no need to blindly chase highs. I remain long-term bullish on AI and also believe that a deep correction in semiconductors in October during the bull market is still possible. But being long-term bullish on tech does not mean you can only hold tech in the short term. What’s truly worth thinking about is when one sector keeps falling while another keeps rising, do you realize the market is reallocating risk? The midterm election is just one window to observe market rotation. What ultimately determines how far the next market rally can go are earnings, interest rates, valuations, and liquidity. I will continue to watch the relative strength of $XLV, $XLP, $XLU, while waiting for $SOXX and $QQQ to show real signs of stabilization. The market never lacks opportunities; the key is not to put all your funds on the same risk.