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

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.

Many people are now worried about the U.S. midterm elections on November 3rd, thinking that if Trump loses and the Republican Party loses control of Congress, the U.S. stock market will face a sharp decline. But if you carefully study the market performance over the past few decades, you will find that Wall Street does not decide stock rises and falls based on the winning or losing of the president's party. $SPY $QQQ $SOXX $NVDA Let's first look at historical data. Capital Group statistics show that since 1950, in the 12 months following the U.S. midterm elections, the S&P 500 has risen by an average of about 15.1%. BlackRock studied midterm elections since 1970 and found that even if the ruling party loses unified control of the White House and Congress, the S&P 500 still rose by an average of about 10.4% in the six months after the election. These two data points already indicate one thing: the president's party losing in the midterm elections can coincide with a rise in the U.S. stock market. Why? Because midterm elections are not presidential elections. Even if the Republicans lose control of the House or Senate, Trump is still president. The changes mainly occur in the legislative landscape over the next two years, not a sudden replacement of the entire U.S. government. If Congress becomes divided, major tax, fiscal spending, and regulatory legislation may face more negotiation and checks and balances. For Wall Street, this situation is not necessarily all bad. Some investors may think that the space for sudden major policy changes is limited, making the future business environment for companies easier to predict. But political deadlock can also cause risks such as budget negotiation difficulties and government shutdowns, so it cannot be simply assumed that a divided Congress is necessarily good for the stock market. What I think is more worth studying is the change in market risk premiums before and after the election. Every time the midterm elections approach, the market needs to consider tax, regulation, fiscal, and industrial policies under different political outcomes. These uncertainties affect the valuations investors are willing to pay. Once the election is over and some political risks are repriced, the market may gradually shift its focus back to corporate earnings, economic growth, and interest rates. 2018 is a very representative example. At that time, Trump was also president, and the Republicans lost control of the House in the midterm elections. But the market did not enter a multi-year decline because of this. Although the U.S. stock market continued to plunge in December 2018, the S&P 500 still rose about 28.9% in 2019. This case also shows that the end of the election does not mean a surge the next day; interest rates and the economic environment can still dominate the market. Back to 2026. The pressure facing the U.S. stock market now is far more than just the midterm elections. U.S. Treasury yields, energy prices, inflation expectations, AI infrastructure financing, and the previously crowded positions in tech stocks are all affecting the entire market. Assuming that after the election, political uncertainty decreases, oil prices fall, inflation eases, U.S. Treasury yields decline, and corporate earnings continue to grow, then multiple risk factors improving simultaneously could create conditions for a valuation recovery in the U.S. stock market. Especially for the tech and semiconductor sectors, which have previously experienced large declines. But conversely, if after the election oil prices continue to rise, U.S. Treasury yields remain high, and corporate earnings forecasts are revised downward, then relying solely on historical midterm election patterns cannot guarantee a stock market rebound. Therefore, I have always emphasized that political news should not be directly treated as a trading signal. Whether Trump wins or loses the midterm elections does not have a simple, inevitable relationship with whether the U.S. stock market will rise in the future. I remain long-term bullish on AI and also believe the current semiconductor adjustment has the characteristics of a deep bull market correction. But I will not assume that because the U.S. stock market generally performed well after midterm elections historically, it is now safe to ignore risks and go all-in with leverage. Short-term position control still needs to be maintained. What is truly worth looking forward to is that after election uncertainty gradually fades, if inflation, interest rates, credit environment, and corporate earnings all improve simultaneously, the market may regain a foundation for sustained growth. Ultimately, Wall Street trades profits, cash flow, interest rates, and valuations. Election results can change market expectations but cannot replace these fundamental factors. So even if Trump loses in the midterm elections, the U.S. stock market still has a very real possibility of a significant rise. But the rise requires conditions; it is not guaranteed just because a certain party loses.

The semiconductor sector hasn't finished falling, and software stocks have started to collectively plunge again. Wall Street may now be selling off not just AI hardware, but the entire overvalued tech sector. $SOXX $QQQ $IGV $NVDA $MSFT On October 16, the $IGV software ETF dropped 4.8% in a single day, $SOXX fell about 4.1%, and software companies like $ORCL, $NOW, and $CRM also came under pressure. Interestingly, just the day before, some software stocks showed a clear rebound and were considered a possible new direction for funds leaving semiconductors. But the next day, software stocks also started to be hammered. This situation is more concerning than a simple semiconductor decline. Previously, the market worried about high AI capital expenditures, data center financing pressures, GPU demand, and storage cycles, which at least partially explained the valuation compression of some semiconductor companies. But companies like $CRM and $NOW do not rely entirely on building data centers to make money. They have mature enterprise customers and subscription revenues, so why are they starting to fall now? I believe the problem is shifting from the fundamentals of individual companies to the valuation and capital structure of the entire tech sector. In recent years, AI-related assets have accumulated a large amount of profit-taking, and the market was willing to pay a high premium for future growth. Now, with changes in the macro environment, institutions are beginning to reduce risk exposure, making crowded positions prone to continuous sell-offs. There is another set of data worth noting. This week, the Nasdaq fell 4.5%, the S&P 500 dropped 1.9%, while the Dow Jones rose 0.7%. This indicates that the internal divergence within the U.S. stock market remains very clear. Some funds are still willing to hold large traditional companies, but their risk tolerance for overvalued tech stocks is declining. I have always been a long-term AI bull and still tend to believe that October is a deep pullback within a bull market. But I repeatedly emphasize that when it's time to reduce positions, you really need to reduce positions. Because the market will not guarantee you won't lose 30% or more in the short term just because you have been right about the industry direction in the long term. Especially now, with semiconductors continuously breaking down and software starting to catch down. Even if a strong rebound occurs next, I will not immediately conclude that the correction is over. I am more focused on whether the rebound can sustain, whether market breadth improves, and whether tech stocks can form effective buying support again. Before these signals appear, controlling position sizes and holding cash remain more important than rushing to bottom-fish. The long-term development of AI has not ended because of a few weeks of decline, but a long-term bull market never means every pullback is worth holding full positions through.