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

300Following

687followers

Feed

With only half a month left until the U.S. midterm elections on November 3, which sectors might the next round of declines in the U.S. stock market focus on? $QQQ, $SOXX, $IWM, $XLF, $XLI—these areas deserve close attention. Especially AI infrastructure and semiconductors, which have already seen significant pullbacks in October, but this does not mean the risks have been fully released. The most vulnerable may be highly leveraged AI infrastructure companies. Companies like $CRWV, $NBIS, and $APLD heavily rely on external financing and continuous expansion, with future profit expectations based on data centers being delivered as planned. Recently, CoreWeave bonds have come under pressure, reflecting credit market concerns about AI financing models. With U.S. Treasury yields remaining high, if financing costs continue to rise, both company valuations and project investment returns will face pressure. These companies will not only be affected by a decline in stock market risk appetite but may also face a double hit from credit and equity markets. Semiconductors are another area to watch. $MU, $SNDK, $AMD, $LRCX, as well as the optical communication sector’s $LITE and $COHR, have benefited from AI capital expenditure expansion, with the market assigning high growth expectations. Although the Philadelphia Semiconductor Index rebounded significantly on October 19, it had already fallen for eight consecutive trading days, with a cumulative pullback of about 24% from the June peak. If the market continues to revise down future AI investment growth rates, both profit expectations and valuations may come under pressure. There is also a special risk related to the midterm elections this year: AI data centers have become a local political issue. According to early October statistics, 34 data center-related voting proposals are involved in the U.S. November elections, covering development restrictions, land use, and other regulatory issues. Electricity prices, water resources, and infrastructure costs are affecting local approvals. If project delays expand, risks may gradually spread from data center operators to GPU, HBM, enterprise-grade SSD, and optical module suppliers. However, project delays do not necessarily mean orders will be canceled. The financial and industrial sectors should not be overlooked either. Historical research by BlackRock’s iShares shows that since 1990, in midterm election years, the financial sector’s average return is about -0.6%, and the industrial sector only +0.6%, significantly weaker than the healthcare and energy sectors during the same period. $XLF and $XLI face not only policy uncertainty but also changes in financing needs, interest rates, corporate investment, and economic growth expectations. Small-cap stocks represented by $IWM are more susceptible to high financing costs. If U.S. Treasury yields continue to rise, AI company financing pressures expand, and tech giants lower capital expenditure expectations, then highly leveraged AI infrastructure and highly elastic semiconductors may continue to face significant selling pressure, with risks potentially spreading to broader large-cap stocks afterward. However, the midterm elections themselves are not necessarily a reason for a decline. Historically, as elections approach and political uncertainty fades, the U.S. stock market often rebounds in advance. To judge whether the risk has ended, one should not only look at $SOXX rebounding a few percentage points but also consider credit spreads, corporate earnings guidance, data center project progress, and whether $SPY experiences a noticeable catch-down. The market has already begun repricing the most vulnerable links in the AI industry chain. If even cash-flow-stable tech giants start to weaken continuously, the nature of the adjustment may undergo further changes.

The semiconductor sector has just experienced its most severe sell-off since October. What truly matters this week is not how much the US stock market can rebound, but whether the upcoming earnings reports can prove that the profit expectations for AI and semiconductors still hold. $NVDA, $MU, $SNDK, $AVGO, $AMD, $LRCX could all be influenced by their earnings results in the coming days. On October 20, $VICR will release its earnings. This company mainly provides high-performance power conversion and supply solutions. As AI server power density continues to increase, the importance of power management is also rising. Key points to watch are order growth related to AI business and management’s outlook on future demand, though its representativeness for the entire GPU industry is limited. October 21 is the most important day this week. $LRCX Lam Research: Directly impacts the semiconductor equipment sector and is an important window to assess future capital expenditures in the memory industry. Focus on demand for DRAM and NAND-related equipment, customer expansion plans, and guidance for the next quarter. If memory manufacturers continue to increase investment in advanced processes, the long-term capacity expansion expectations for companies like $MU and $SKHY may be validated, which would also benefit equipment companies like $AMAT and $KLAC. However, increased capacity expansion also means potential future supply growth, which does not unconditionally support memory prices. $TXN Texas Instruments: Mainly observe demand for industrial, automotive, and analog chips. The market has focused too much on AI chips in the past, causing traditional semiconductor cycles to be overlooked. If industrial and automotive orders continue to improve, it indicates that the industry recovery may no longer be limited to AI data centers, which is meaningful for companies like $ADI, $NXPI, and $ON. $META: This may be the earnings report with the greatest impact on the overall valuation of AI semiconductors. Meta previously projected capital expenditures of $130 billion to $145 billion for 2026. The market is no longer just concerned about how much is invested, but whether these investments can translate into revenue, profit, and cash flow. If management maintains an aggressive AI capital expenditure plan while demonstrating improving commercial returns, $NVDA, $AVGO, $TSM, and the memory supply chain could all gain sentiment support. Conversely, if capital expenditures continue to grow but free cash flow and profit margins further deteriorate, the market may continue to reassess AI investment returns. On the same day, $TSLA and $IBM will also report. The former mainly focuses on investments related to autonomous driving, robotics, and automotive business cash flow; the latter focuses on enterprise AI demand, server business, and infrastructure revenue. Both have a more indirect impact on semiconductors. On October 22, $WDC, $MXL, $NOK, and $BE are worth watching. $WDC mainly reflects demand for large-capacity HDDs in data centers, focusing on cloud vendor procurement, Nearline hard drive shipments, and order visibility. It can help assess AI data storage demand but cannot be directly used to predict $SNDK’s NAND prices because the two companies have completed business spin-offs. $MXL focuses on communication and connectivity chip demand; $NOK focuses on network infrastructure and optical network business. Both have some reference value for optical communication companies like $LITE and $COHR, but their product structures and customers are not entirely the same. $BE requires attention to on-site power supply demand in data centers; whether power supply can keep up with AI infrastructure expansion also affects the order fulfillment speed across the entire industry chain. The key issues this week are threefold: whether semiconductor equipment orders have declined, whether AI giants have cut capital expenditures, and whether traditional chip demand continues to recover. If $LRCX’s orders and guidance remain robust, $TXN confirms industrial demand improvement, and $META continues to expand AI investment without significant deterioration in profitability, then the continuous decline in semiconductors in October is more likely to include a considerable degree of valuation compression and deleveraging factors. But if orders, capital expenditures, and profit guidance all weaken simultaneously, even if the stock price has already dropped 20%, it cannot be simply concluded that the valuation is cheap just because of the large decline. This week’s earnings reports will help determine whether the fundamentals have worsened, but whether the adjustment truly ends depends on the stock price reaction after the reports and whether the entire semiconductor sector can sustain a recovery.

After the semiconductor sector's sharp decline in October, who should be bought in the next market cycle? $NVDA, $TSM, $AVGO, $MU, $AMD, $ASML — although all these companies belong to the semiconductor industry chain, their future growth potential, industry status, and earnings elasticity differ significantly. The industry leader is not necessarily the stock with the largest gains, and the one that fell the most is not always the best bottom-fishing opportunity. First, let's look at companies with true core moats. $NVDA is the AI GPU and computing power ecosystem leader, with CUDA, network interconnect, and hardware-software systems forming competitive barriers. $TSM is the leading advanced wafer foundry, with almost all top AI chip design companies highly dependent on its manufacturing capabilities. $ASML controls the critical equipment segment of EUV lithography machines. The biggest advantage of these three companies is not short-term earnings growth but that competitors will find it difficult to replicate their industry positions within a few years. If AI capital expenditures continue to expand, they will remain at the core of the industry chain. However, if the focus is on earnings growth and stock price elasticity over the next two to three years, $MU, $AVGO, $AMD, and $LRCX may be more worth studying. $MU benefits simultaneously from HBM, server DRAM, and storage price cycles, with earnings highly sensitive to supply and demand changes. $AVGO is expanding its market space through custom AI accelerators and high-speed network chips. $AMD's opportunities come from increasing server CPU market share and expanding AI GPU business. $LRCX benefits from storage companies upgrading to advanced processes and increasing equipment investments. In the storage sector, attention should also be paid to $SKHY and $SNDK. The former is a leading HBM supplier, while the latter is mainly influenced by NAND, enterprise SSD demand, and flash memory prices. Both have cyclical elasticity, but the factors driving their earnings differ. In semiconductor equipment, $AMAT and $KLAC hold strong competitiveness in materials engineering equipment, process control, and inspection segments. If only three companies are chosen as core observation targets for the next semiconductor market cycle, $TSM, $AVGO, and $MU deserve to be prioritized. $TSM corresponds to the certainty of advanced manufacturing, $AVGO to the growth potential of custom AI chips, and $MU to the earnings elasticity of the storage industry. Meanwhile, $AMD, $LRCX, and $SNDK are better suited for opportunities driven by cycles and expectation changes with high elasticity. But it is essential to distinguish between good companies and good prices. No matter how strong $NVDA is, if its valuation is too high, future stock returns may be limited; even if $MU's valuation looks cheap, if the storage price cycle reverses, profit expectations may still be revised downward. After the October decline, what is worth studying is not which company is furthest from its historical high, but whose earnings growth can sustain, how much risk the valuation has already priced in, and which companies have the strongest ability to regain market pricing power when the next round of capital flows back.

Samsung itself owns the world's leading DRAM business and advanced wafer foundries, yet it has now been revealed that it is seeking competitor TSMC to produce key chips for the next-generation HBM. What is even more noteworthy is that the request is not coming from Samsung, but from Nvidia $NVDA. The rules of global HBM competition may be undergoing a significant change. $TSM $MU $SOXX On October 18, the Seoul Economic Daily reported that Nvidia has requested Samsung, SK Hynix, and Micron $MU to uniformly use TSMC's 3nm process to produce the base logic chips (Base Die) for the next-generation custom high-bandwidth memory NVHBM, while also requiring the three manufacturers to use the sixth-generation 1c DRAM process to manufacture the memory core chips. Samsung is said to have accepted this request. It should be emphasized that these supply chain arrangements currently come mainly from industry sources cited by Korean media and cannot yet be regarded as fully confirmed mass production plans by the three companies. Why is this matter worth attention? Past HBM competition mainly focused on DRAM process technology, stacking layers, yield, and thermal performance. Samsung has always hoped to leverage its advantage of owning both memory manufacturing and wafer foundry businesses to complete internal production from DRAM to logic chips. However, with the development of HBM4, HBM4E, and custom HBM, the tasks of the base logic chip have become increasingly complex, and the importance of advanced logic processes has clearly increased. Nvidia's NVHBM technology announced in August is a typical example. Some control functions of traditional HBM are located inside the compute chip, whereas NVHBM integrates the custom memory controller into the base logic chip. According to Nvidia's official data, compared to standard HBM4E, this design can increase memory bandwidth by up to 30%, reduce HBM power consumption by 15%, and free up to 25% additional area on the compute chip. This means the next-generation HBM is no longer just about stacking more DRAM, but requires simultaneous optimization of memory, logic chips, advanced packaging, and compute architecture. Whoever can provide stable advanced logic manufacturing capabilities may gain a more important position in the industry chain. This is also TSMC's competitive advantage. On October 15, $TSM announced its Q3 financial report, with revenue reaching $46.72 billion, net profit increasing 68.7% year-over-year, and a gross margin of 66.4%. Among this, the 3nm process has contributed 33% of wafer sales revenue, becoming the company's largest single process revenue source. When advanced process capacity is already needed to supply GPU, CPU, and smartphone chips, if in the future HBM base logic chips also begin to widely adopt 3nm, TSMC will further enter the value chain originally mainly controlled by memory manufacturers. However, the scale of related orders, pricing, and mass production timing remain unclear, so it cannot be directly inferred that the company's future profits will increase significantly because of this. Samsung's situation is even more complex. It has not completely lost HBM manufacturing capability, nor has it abandoned its wafer foundry business. According to Korean media reports, Samsung's standard HBM4 and HBM4E still use its own 4nm process to manufacture base logic chips, and it is also verifying the possibility of producing NVHBM with its own 2nm process. But if Nvidia insists on uniformly using TSMC's 3nm, Samsung will need to make trade-offs between customer demands and internal manufacturing capabilities. This at least indicates that owning its own foundry does not necessarily mean full control over the supply chain for the highest-end AI memory products. For Micron $MU, this matter is worth observing from another perspective. Although Micron does not have a large advanced logic foundry business like Samsung, if all three memory manufacturers adopt TSMC to produce custom HBM base logic chips, the competition focus may further shift to DRAM core chip yield, stacking capability, product energy efficiency, and actual delivery capability. Micron has previously revealed that future HBM4E and NVHBM base logic chips will rely more on external foundries. A unified manufacturing platform helps reduce process differences between different suppliers but does not mean the three companies' costs, yields, and final product performance will be exactly the same. SK Hynix also cannot be ignored. As an important supplier in the HBM field, it has already cooperated with TSMC on advanced logic chips. If the industry standard for custom HBM further concentrates on TSMC, SK Hynix's accumulated manufacturing and packaging experience still holds value. From Nvidia's perspective, unifying the base logic chip process also helps reduce verification complexity between different memory suppliers, improve product compatibility, and reduce dependence on a single HBM manufacturer. This reflects a long-term trend: Nvidia's competitive advantage may no longer be limited to the GPU itself but is gradually extending to the design and supply chain coordination of the entire AI computing system. However, this standardization should not be directly interpreted as Nvidia having control over all memory manufacturers. HBM capacity, DRAM technology, advanced packaging, and each company's commercial negotiation capabilities will still affect the final supply pattern. For the memory sector, which has recently experienced a significant pullback, this news at least indicates that high-end HBM technology continues to upgrade and industry competition has not stopped. But technology upgrades do not necessarily mean related stocks will rise in the short term; subsequent observation of product certification, mass production progress, order pricing, and 2027 profit expectations is still needed. Additionally, $SNDK mainly operates in NAND flash memory business and is not a direct producer of HBM, so this news impacts it more through overall AI memory sentiment rather than direct HBM order changes. Future competition in the AI memory industry may increasingly depend on the synergy among DRAM manufacturers, advanced wafer foundries, and AI chip design companies. Samsung owns its own advanced foundry but may still need to use TSMC's 3nm. This situation reflects not that Samsung has lost competitiveness, but that the technical threshold for next-generation HBM is extending from pure memory manufacturing to the entire advanced semiconductor manufacturing system. For investors who have long studied AI infrastructure, the future profit distribution of $TSM, $MU, and the global HBM industry chain deserves reevaluation.

Earned 439% in half a year, lost 67% in one month, and was eventually forced to sell most of the stocks to Citadel. This is what happened to the fund Situational Awareness, managed by former OpenAI researcher Leopold Aschenbrenner, in 2026. Leopold is definitely not someone lacking investment ability. He graduated from Columbia University at 19 as a student representative and later joined OpenAI to conduct AI research. He was an early believer in AI infrastructure, heavily invested in companies like $SNDK, $MU, $BE, $CRWV, and at one point achieved extremely impressive investment returns. According to public reports, the fund rose 439% cumulatively in the first half of 2026, but in July it faced a concentrated decline in AI stocks, losing about 67% in a single month. Due to margin calls on leveraged positions, it was ultimately forced to sell most of its publicly traded stocks to Citadel and voluntarily deleverage. The most valuable lesson here is not that he misjudged AI, but that even with a correct long-term view, leverage can cause you to lose the qualification to continue holding. Assuming an initial $1 million, a 439% increase in the first half would grow assets to $5.39 million. Then a 67% loss leaves about $1.78 million. Although still up about 78% for the year, most of the previously accumulated profits have disappeared. Even harsher, to get back from $1.78 million to $5.39 million requires about a 203% gain, not 67%. Leverage further amplifies this asymmetry. Assuming 3x leverage to buy stocks, a 10% drop in the underlying would theoretically cause about a 30% loss of principal; a 20% drop would cause about a 60% loss. If the decline continues and margin requirements are not met, forced liquidation may occur. Actual results are also affected by borrowing interest, margin rules, and price gaps. Investors without financing leverage can decide when to sell on their own as long as they have no other financial pressures. Even if $MU and $SNDK drop 30% in the short term, as long as the company's long-term fundamentals remain unchanged, they can choose to hold on and wait for future valuation recovery. Of course, waiting does not guarantee a return. But at least you won’t be forced to liquidate at the worst possible time due to broker margin requirements. Options work similarly. In September, reports indicated that Leopold’s fund re-engaged in options trading on $AMD, $BE, $CRWV, $SNDK, with related large option purchases involving about $315 million in premiums. Buying options can limit the maximum loss per trade to the premium paid, but if the amount invested is too large, time decay and changes in implied volatility can still cause severe losses. Buying the right direction but the wrong expiration date can also result in losing the entire premium. Especially for short-term options, the stock ultimately rising does not mean your contract will still be valid by then. After experiencing significant market volatility, I increasingly prefer to separate stock holdings from leveraged trading. For companies with genuine long-term potential, control position size and hold the underlying stock; for high-risk options, determine the maximum tolerable loss in advance rather than relying on adding positions to wait for a rebound. The market never lacks opportunities, and AI, semiconductor, and storage industries may continue to grow in the future. But if a single extreme event can cause unbearable principal losses, no subsequent big rally matters. The most important thing in investing is not how many times you make money in a single year, but whether you can stay in the market after several bull and bear cycles. Even fund managers managing tens of billions of dollars with professional research capabilities can be forced by leverage to sell stocks they believe in long-term. Ordinary investors have even less need to prove their judgment through high leverage. If the investment logic can be achieved with the underlying stock, there is no need to use leverage to do so.

On October 19, the US semiconductor market finally saw a significant rebound, with $MU and $SNDK leading the storage sector, and the Philadelphia Semiconductor Index rising more than 2% intraday. So, will the Asia-Pacific market follow suit on October 20? Especially for Samsung, SK Hynix, and Japan's Kioxia, which previously experienced historic-level crashes—can they reignite the global storage market? $SOXX $QQQ $MU $SNDK Let's first look at the South Korean market. On October 19, the Korean KOSPI fell more than 3% in early trading but ultimately reversed to close up 1.76%. Samsung Electronics closed up 1.16%, and SK Hynix rose 0.39%. This rebound indeed sent a positive signal, but it is important to note that the two major Korean storage leaders actually underperformed the KOSPI index that day and did not show a very strong leading rally. Why is the rebound strength of Korean storage stocks still limited? Because the prior sell-off was extremely severe. From October 12 to 16, Samsung Electronics dropped a cumulative 17.8%, and SK Hynix plummeted 24.3%, both marking the worst weekly declines since the 2008 financial crisis. Foreign investors net sold about 11.2 trillion KRW worth of Korean stocks in one week, with Samsung and SK Hynix accounting for approximately 8.8 trillion KRW of net sales, nearly 79% of the total foreign net selling. More notably, although SK Hynix closed up on October 19, foreign investors still net sold about 217.6 billion KRW. Samsung, on the other hand, saw about 172 billion KRW in foreign net buying that day. This indicates that bottom-fishing funds have appeared in the Korean storage sector, but it cannot yet be confirmed that the large-scale foreign capital outflow has fully reversed. Now, looking at October 20. If the US stock market maintains its current rebound on October 19, with $MU, $SNDK, and the Philadelphia Semiconductor Index closing strongly, the conditions for continued recovery in South Korea, Japan, and Taiwan on the 20th will significantly improve. In South Korea, SK Hynix and Samsung Electronics may benefit from the sentiment recovery brought by the US storage stock rebound. Especially SK Hynix, which fell over 24% last week, has considerable short-term oversold rebound potential. However, a larger drop does not necessarily mean a stronger rebound; whether foreign investors stop concentrated selling remains very important. In Japan, Kioxia's performance is also worth watching. Kioxia is a major global NAND flash manufacturer and operates in a market closer to $SNDK. However, the Japanese market currently faces an independent risk: as of October 15, Kioxia's margin trading buy balance reached about 920 billion JPY, the largest among Tokyo Stock Exchange-listed companies. Many leveraged long positions had previously bet on storage price increases, but with the stock price continuing to fall, some investors may face margin calls or forced liquidations. This means that even if the US semiconductor market rebounds on October 20, Kioxia may experience a pullback after a rally due to unresolved leveraged positions. Short-term gains need to be judged in conjunction with trading volume and margin position changes. In Taiwan, the weighted index closed down 0.61% on October 19, but intraday it plunged more than 940 points before significantly narrowing losses, led by TSMC and other tech heavyweights. If the US semiconductor rebound continues, Taiwan's advanced manufacturing and packaging/testing supply chains may also gain further support on the 20th. So, will the performance of Korean storage stocks, in turn, affect US stocks $MU and $SNDK? The answer is yes, and this linkage deserves close attention. SK Hynix, Samsung, and Micron are key competitors in the global DRAM and HBM markets. These three companies are influenced by similar AI demand, capacity investments, and storage price expectations, and global capital often allocates them as part of the same industry cycle. Therefore, if SK Hynix and Samsung continue to rebound strongly on the 20th, and foreign investors resume net buying, the market may interpret this as global DRAM and HBM sectors regaining capital recognition. This usually has more reference value for $MU's market sentiment. But $SNDK requires separate analysis. $SNDK mainly operates in NAND flash, with more direct industry ties to Japan's Kioxia, Samsung's NAND business, and SK Hynix's Solidigm. Therefore, judging $SNDK's short-term performance should focus not only on Korean DRAM and HBM stocks but also on Kioxia's trend, enterprise SSD demand, and NAND price expectations. Korean storage gains do not simply imply that $SNDK will rise in sync. Chronologically, the US market rebound on the 19th may drive a higher open in Asia-Pacific on the 20th, and the Asia-Pacific market's performance on the 20th will become an important reference for US pre-market trading on the same day. Both sides influence each other through shared industry expectations and global capital sentiment, but correlation alone cannot determine who is driving whom. One unresolved issue remains. Although US semiconductors rebounded clearly on October 19, Reuters afternoon data showed that 7 of the 11 major S&P 500 sectors still declined. Some large tech and semiconductor stocks are pulling the index up, but overall market risk appetite has not recovered in sync. If US tech stocks are only experiencing an oversold rebound after continuous crashes, and AI financing costs, US bond yields, and market breadth do not improve sustainably, then even if Asia-Pacific opens collectively higher on the 20th, it may again face profit-taking and selling pressure from previous trapped positions. For the market on the 20th, key points to watch are whether Korean storage stocks can continue to outperform the KOSPI, whether foreign investors switch from selling to buying, and whether Japan's Kioxia can remain stable during the rebound. If Samsung, SK Hynix, and Kioxia all strengthen simultaneously, Taiwan's semiconductor sector continues to recover, and US futures remain stable, then the global storage sector's rebound is more likely to extend into US trading on the 20th. Conversely, if Korean storage opens high but closes lower, and Kioxia continues to be pressured by leveraged selling, even if US stocks closed up on the 19th, caution is needed for weakening rebound momentum in US storage stocks on the 20th. Currently, the tendency is to believe there is a chance for continued semiconductor rebounds in Asia-Pacific on October 20, but a single US market rise is not enough to confirm that the global storage sector's adjustment has ended. Short-term trends in Korean and Japanese storage stocks can help gauge market sentiment for $MU and $SNDK on the next trading day; however, what truly determines the upward potential in the coming months remains the supply-demand balance, product prices of DRAM, HBM, and NAND, and whether corporate earnings expectations for 2027 can be continuously realized. The long-term growth logic of the storage industry has not disappeared, but after such severe concentrated sell-offs, rebuilding an upward trend requires time. What is more important to observe on the 20th is whether the rebound can be sustained and whether capital is genuinely starting to flow back, rather than simply predicting how much the market will rise at the open.

Today, the semiconductor sector collectively rebounded, but I still maintain my previous judgment: this round of rise can only be temporarily defined as an oversold rebound, not the establishment of a new upward trend. More notably, the market movement after the U.S. stock market opened on October 19 has actually begun to reveal the problem of insufficient rebound strength. $SOXX $QQQ $MU $SNDK $AVGO Let's first look at the most direct price changes. $SOXX surged to around $513 in the early session today, with gains approaching 4% at one point, but it fell back to around $502 in the afternoon in the U.S. East, narrowing gains to about 1.7%. In contrast, on October 6, $SOXX was still around $589, and on October 16, it had dropped to a low of $492.83. In other words, after such a significant decline, even a 3% or 4% rebound today does not change the previously formed downtrend. From a technical structure perspective, $SOXX still has not firmly reclaimed the previous important resistance zone. The $515–$520 area is a short-term position that needs to be recovered, followed by facing the dense trapped positions in the $530–$545 range. The continuous decline previously formed a large amount of trapped chips; when the stock price rebounds to these areas, some funds may choose to reduce positions rather than continue chasing highs. The intraday fall from around $513 today at least indicates that selling pressure above still exists, but a single-day surge and fall cannot independently prove that institutions are concentrating on selling out. The second issue is that the core risk of this decline has not disappeared. Last week, the market's real concern was not only that AI stock valuations were too high but also the financing ability of some AI infrastructure companies. The yield on $CRWV's bonds maturing in 2031 once exceeded 16% last Monday; although it fell back to about 15.13% today, it remains at a very high level. Goldman Sachs' rating upgrade can drive a short-term stock price rebound but cannot immediately change bond investors' pricing of its debt repayment ability and future financing costs. Third, U.S. Treasury yields have only fallen back from previously extreme highs and have not truly entered a loose environment. The 10-year U.S. Treasury yield remains above 5%, and the construction costs, debt financing, and long-term capital returns for AI data centers still face pressure. For cash-rich tech giants, the impact mainly reflects in valuations; for emerging computing power companies relying on external financing, the impact will directly affect project costs and financing conditions. Additionally, today's semiconductor rebound mainly occurred after eight consecutive trading days of decline. Previously, the Philadelphia Semiconductor Index lost about $2 trillion in market value cumulatively, with a single-week drop exceeding 11%. After such an extreme market, the temporary exhaustion of selling, short covering, and short-term funds bottom-fishing could all drive rapid stock price increases. The rise itself is real, but there is currently insufficient evidence to prove that new long-term funds have retaken market leadership. More importantly, this week $META, $LRCX, $TXN, and other companies will release earnings reports. The market needs to reconfirm 2027 AI capital expenditures, storage demand, wafer manufacturing equipment orders, and data center financing environments. If earnings continue to deliver growth and profit expectations remain stable, semiconductors will have more sufficient reasons to regain valuation premiums. Conversely, even if the current quarter's performance is good, as long as management turns cautious about future growth, the market may continue to downgrade valuations. Therefore, the focus should not only be on how much the price rose today but on whether $SOXX can firmly reclaim the $515–$520 level and further break through the $530–$545 resistance zone. On subsequent pullbacks, it must hold previous lows, and $QQQ, $SPY, and major semiconductor leaders need to form sustained synchronized rallies. If it only rises today and falls tomorrow, then retests $493 or even breaks below previous lows, the downtrend has not materially changed. The long-term bullish logic for AI, storage, and optical communications industries remains unchanged, but the market is now trading not only on demand growth but also financing costs, 2027 profit realization, and how many times valuation investors are willing to pay. I have previously considered the rebound as an opportunity to readjust positions, and there is currently no sufficient reason to change this judgment. For investors holding large amounts of highly elastic semiconductor stocks or short-term options, reducing excessive risk exposure during the rebound remains more reasonable than fully restoring positions. If the market continues to break through resistance and stabilizes on pullbacks over the next few trading days, and the credit market simultaneously recovers, then a reassessment of the previously cautious judgment will be necessary. But before these conditions appear, today's rise looks more like a normal rebound after continuous sharp declines. A second round of decline may not necessarily occur, but there is currently no evidence that the market has completely escaped the downtrend.