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

693followers

Feed

$NVDA Nvidia's quarterly revenue has nearly reached $100 billion, with a gross margin still as high as 75%, but the stock price continues to face significant adjustments due to AI investment returns and financing risks. What kind of company is Nvidia exactly, and why can it achieve such high profit margins in the semiconductor industry? Many people still think of Nvidia as just a GPU seller, but today's Nvidia is far more than a graphics card company. Its core businesses are divided into data center, gaming, professional visualization, automotive, etc., with the data center already holding an absolute dominant position. According to the Q2 fiscal 2027 report, Nvidia's single-quarter revenue was $96.2 billion, a year-over-year increase of 106%, with data center revenue at $89 billion, up 117% year-over-year, accounting for over 92% of total revenue. The GAAP gross margin was 75%, and net profit was about $59.7 billion. The company expects next quarter's revenue to reach $108 billion, indicating management still believes AI infrastructure demand will continue to expand. Nvidia's ways of making money mainly fall into three levels. At the hardware level, Nvidia provides customers with computing platforms like Blackwell and Vera Rubin, including GPUs, CPUs, network chips, and supporting systems. Cloud computing companies such as $MSFT, $GOOG, $AMZN, and many AI computing power operators are important customer groups. At the software level, CUDA is one of Nvidia's most important competitive advantages. Developers have long built software tools, computing frameworks, and application ecosystems around CUDA. Even if other manufacturers launch GPUs with comparable performance, it doesn't mean customers can immediately complete migration. Hardware can compete, but the software ecosystem accumulated over many years is difficult to replicate in the short term. At the system level, Nvidia uses technologies like NVLink, InfiniBand, and Spectrum-X to expand a single GPU into a large-scale AI computing cluster. AI training and inference require not only chip performance but also communication efficiency between GPUs, network bandwidth, memory capacity, and overall power consumption. Nvidia's sales are shifting from single chips to complete computing infrastructure. However, Nvidia is not without competitors. $AMD is improving AI GPUs and software ecosystems, $AVGO participates in custom ASIC development, and Google's TPU and Amazon's Trainium are also competing for specific workloads. In the long term, custom chips may divert some general GPU demand, but in the short term, fully replacing Nvidia's system ecosystem remains difficult. Nvidia also has a frequently overlooked risk: a large portion of its revenue ultimately depends on the continuous expansion of AI investment. Although direct customers may be large cloud providers, some end computing power demand comes from AI companies relying on external financing. If these companies face financing difficulties or data center delays, the impact may gradually transmit to Nvidia through cloud providers and equipment procurement. Regarding valuation, as of the trading session on October 20, Nvidia's forward P/E ratio is about 18 times, which on the surface no longer seems as expensive as during the early AI boom. But this valuation is based on forecasts of continued rapid profit growth. If growth expectations are revised downward, the actual valuation is not as cheap as it appears. In the long run, Nvidia remains one of the companies with the strongest business model and most outstanding profitability in the AI infrastructure industry chain. However, judging whether the stock price can continue to rise requires not only looking at strong GPU demand but also the data center growth, gross margin, free cash flow, and major customer capital expenditures in the coming quarters. A good company and a good price are two different things. Nvidia's moat is worth long-term study, but no matter how strong the fundamentals, it cannot guarantee the stock price will never experience valuation compression.

The real turning window for the storage sector may be within the next 8 days. $MU has retraced nearly 20% since October 7, and $SNDK, SK Hynix, and Samsung have also experienced sharp adjustments. However, the next four earnings reports may directly determine whether this decline is a golden opportunity in a bull market or a sign that the storage cycle is about to peak. $MU $SNDK $SOXX $NVDA Let's first look at the latest industry data. On October 21, South Korea announced that semiconductor exports in the first 20 days of October reached $27.5 billion, a year-on-year increase of 222.5%, setting a historical record for the same period. TrendForce expects traditional DRAM contract prices to rise 10%–15% quarter-over-quarter in Q4, NAND to increase 15%–20%, and enterprise SSDs to possibly rise 23%–28%. This means the actual operating environment for storage companies is diverging significantly from stock price trends. Stock prices have plunged continuously, yet storage contract prices are still rising. The market is now trading not just on how much companies can earn, but on concerns about how long this ultra-high prosperity can last. The upcoming earnings season is very intense. On October 22, $WDC will be the first to report earnings. Its main business is mechanical hard drives, focusing on verifying the data storage demand and order continuity of large cloud providers. On October 28, SK Hynix will announce its Q3 earnings. Key points include HBM orders, DRAM prices, 2027 capacity allocation, and capital expenditures. It is especially important to see if management continues to confirm tight supply and demand, rather than just delivering a high-profit report already expected by the market. On October 29, Samsung and $SNDK will report earnings consecutively. Samsung will further verify supply and demand changes across the DRAM, HBM, and NAND markets, while $SNDK is more directly related to whether enterprise SSD and NAND price increases can translate into sustained improvements in revenue, gross margin, and free cash flow. There is also an unignorable variable: Chinese storage manufacturers are accelerating their catch-up. The technological progress of Changxin Memory, the NAND competition from Yangtze Memory, and Samsung and SK Hynix upgrading existing NAND production lines all remind the market that today's supply tightness does not mean perpetual undersupply in the future. Therefore, my judgment on storage remains long-term bullish but short-term cautious. The violent rebound in the past two days is not enough to confirm a trend reversal. What really deserves attention is whether management can continue to raise future profit forecasts after these earnings reports, and whether stock prices can consecutively reclaim the key moving averages previously broken. If SK Hynix and Samsung continue to confirm strong server storage demand in 2027, $SNDK proves that enterprise SSD price increases can sustainably improve profits, and $MU re-establishes higher highs and higher lows, then the probability of the storage sector re-entering an uptrend will significantly increase. Conversely, if earnings reports exceed expectations across the board but stock prices continue to open high and close low, it indicates that capital is still unwilling to pay higher valuations for the current prosperity, and the correction may not be over. The next round of storage sector gains may not require new earth-shattering positive news, just the market's renewed belief that 2027 profits will not be worse than currently expected. The October plunge has already released some valuation risk. What will truly determine the direction next is how these giants respond regarding orders, prices, and profits for 2027.

SK Hynix's latest buyback really turned out to be a lonely one. They spent 40 trillion KRW, but the stock price actually dropped 23.5%, almost entirely helping foreign investors offload their shares 😅 $SKHY On August 19, SK Hynix announced the largest stock buyback plan in its history, preparing to spend 40 trillion KRW to repurchase and cancel shares, accounting for about 3.3% of the total shares. The buyback officially started on August 20, with the closing price that day at 1.691 million KRW. By October 21, the company announced that the 40 trillion KRW buyback was fully completed, nearly a month ahead of the original November 19 schedule. Although SK Hynix surged 5.81% today, it closed at only 1.294 million KRW. In other words, during these two months of buybacks, the stock price fell from 1.691 million to 1.294 million KRW, a cumulative drop of about 23.5%. What’s even more outrageous is the capital flow. From August 20 to October 20, foreign investors net sold about 27.68 trillion KRW, Korean individual investors net sold 8.31 trillion KRW, and institutions net sold 3.13 trillion KRW, totaling a net sell of about 39.12 trillion KRW among these three investor groups. Meanwhile, SK Hynix repurchased about 39.43 trillion KRW worth of shares during the same period. Putting these numbers together is quite interesting. Foreign investors, retail investors, and institutions took turns selling, while the company kept buying tens of trillions of KRW, yet the stock price still plummeted. Of course, the buyback is not a total loss. After shares are canceled, the total share count decreases, which is beneficial in the long term for increasing the equity per remaining share. But from a short-term stock price perspective, this 40 trillion KRW clearly couldn’t withstand the market selling pressure. Moreover, Samsung Electronics’ previous buyback of about 15 trillion KRW has also ended. The combined buybacks of these two Korean semiconductor giants, totaling about 55 trillion KRW, have now exited the market. If foreign investors continue to sell, who will provide new buying support? Now all eyes are on SK Hynix’s Q3 earnings report on October 28. The company has previously promised that from 2025 to 2027, over 50% of cumulative free cash flow will be used for shareholder returns including buybacks, cancellations, and dividends, and said it will disclose further plans in the Q3 report. Whether the next step will be to increase stock buybacks or raise cash dividends is worth close attention. SK Hynix’s competitive advantage in HBM and the long-term demand logic for storage haven’t disappeared despite the stock price plunge, but this 40 trillion KRW buyback at least shows one thing: No matter how much money the company has, it may not be able to absorb the entire market sell-off. On October 28, we’ll see if management is still willing to put real money on the line to support the stock.

Before the U.S. stock market opened on October 21, semiconductors began a collective counterattack again. SK Hynix rose 5.81%, Samsung rose 5.67%, and South Korea's KOSPI surged 3.29%. Overnight, $SMCI suddenly raised its quarterly revenue forecast significantly, driving $QQQ to rise about 0.7% in after-hours trading. Market sentiment is rapidly recovering, but I still tend to believe this is just a continuation of the oversold rebound on the 19th, rather than the establishment of a new upward trend. $SOXX $NVDA $MU $SNDK $AVGO From the candlestick perspective, on October 20, $QQQ's intraday low fell to around $710, then quickly rebounded in a V-shape, reaching a high of $722.74, but ultimately closed at about $717.6, even slightly below the closing price on the 19th. Although the lower shadow indicates support at the low, the high point of $724.55 on the 19th was not broken, let alone the dense trading area between $740–$745 previously. The short-term downtrend formed earlier has not been broken for now. The Nasdaq Composite Index closed near 26,424 points on October 20, still below the 50-day moving average of about 26,624 points. The Philadelphia Semiconductor Index had eight consecutive declines, with the RSI dropping to as low as 21.5 on the 9th. Such an extreme oversold condition naturally tends to trigger a sharp rebound. A V-shaped counterattack after a rapid decline does not mean the trend has changed. A true reversal requires the continuous formation of higher highs and higher lows, not just a single-day sharp rally. Looking at the options structure, according to third-party options statistics after the close on October 20, the $QQQ Put/Call open interest ratio was about 1.4, with Put open interest significantly higher than Call. However, this cannot be directly interpreted as large-scale short positioning because Puts may come from protective hedging, and open interest alone cannot tell us who is the buyer or seller. There is another easily overlooked risk in the options market: if market makers are in a net negative Gamma position, price declines may trigger more sell hedging, while price rises may trigger chasing buy orders, amplifying volatility in both directions. But public open interest alone cannot confirm the actual net Gamma exposure of traders. Therefore, the V-shaped rebound on the 20th may include genuine buying but could also be driven by short covering and options hedging adjustments. It cannot be simply understood as institutions reestablishing large-scale positions. As for why there were consecutive sharp declines in the previous days, I still tend to believe the main pressure likely came from existing long holders exiting, profit-taking, and deleveraging. In recent months, AI, memory, and optical communications accumulated substantial gains. When U.S. Treasury yields, AI financing costs, and valuation expectations all faced pressure simultaneously, existing positions actively reducing risk was enough to cause severe sell-offs. However, existing capital withdrawal and new short entries are not mutually exclusive. There is currently insufficient data to prove that shorts have fully exerted force, nor to prove they have not acted yet. If this rebound fails to break through resistance above and then falls below previous lows again, new shorts and remaining longs selling could jointly amplify the next decline. However, my long-term bullish view remains unchanged. $SMCI now expects quarterly revenue to reach $18.5–19.5 billion, significantly exceeding previous forecasts. The sharp rebound in South Korean semiconductors also indicates that the market's previous pessimistic pricing of AI demand is being challenged by new operational data. The long-term industry logic of companies like $NVDA, $MU, and $AVGO will not automatically disappear due to a sharp pullback in October. I still regard October as a relatively large correction within a bull market, not a confirmed entry into a bear market. After valuations are fully digested, leverage risks released, and earnings expectations stabilized again, AI infrastructure and semiconductors have every chance to enter a stronger upward phase. But being long-term bullish does not mean going all-in to bottom-fish now. In the short term, $QQQ needs to be watched to see if it can effectively break through the $724–730 range and then further hold above the $740–745 area; semiconductors need to reclaim key moving averages and continue to outperform the broader market. If the upward structure is confirmed, I will gradually revise my currently cautious stance. Conversely, if the rebound ends and $QQQ falls below the $707–710 range again, the risk of a secondary pullback needs to be reassessed. The bull market has not ended, but that does not mean the October correction is over. What is truly worth waiting for is the market turning from an oversold rebound back to sustained gains, rather than rushing to declare a new major uptrend after every V-shaped counterattack.

SK Hynix's performance today really feels like a rare occurrence in this lifetime. Yesterday, Korean semiconductors were being frantically sold off, but today they launched a violent counterattack, with intraday gains exceeding 8% at one point, and Samsung Electronics also approaching 8% at its peak. At the close of the Korean market on October 21st, Eastern US time, SK Hynix was still up 5.81%, Samsung up 5.67%, and the KOSPI surged 3.29%. The worse the drop yesterday, the fiercer the counterattack today. $MU $SNDK $NVDA $SOXX Even more astonishing, the Korean market triggered the buy-side Sidecar mechanism intraday. Because KOSPI 200 futures rose more than 5% in a short time, the exchange suspended programmatic buy orders for 5 minutes. This was not a typical low-open, high-close scenario, but an extremely fierce rebound across the entire semiconductor sector. The catalyst came from the US stock market after hours. $SMCI suddenly raised its quarterly revenue forecast from $14.5–15.5 billion to $18.5–19.5 billion, with gross margin expectations also upgraded accordingly. Previously, the market was worried about AI capital expenditure slowing and data center financing difficulties, but now the AI server supplier has delivered revenue guidance that clearly exceeds expectations. Today's export data released by Korea is even more noteworthy. In the first 20 days of October, Korean semiconductor exports reached $27.5 billion, a year-on-year increase of 222.5%, setting a record for the same period, accounting for 44.7% of Korea's total exports. This shows that even though stock prices experienced severe adjustments in the past two weeks, the actual semiconductor export value remains very strong. Of course, the export growth is also driven by rising memory prices and should not be entirely interpreted as shipment volume growth. This is where SK Hynix deserves praise. As one of the world's core suppliers of HBM and DRAM, it occupies an important position in the AI computing power industry chain while bearing valuation pressure from the memory cycle and high volatility. Yesterday, the market was still worried about a collapse in AI demand, but today industry data and server orders have sent a completely different signal. However, an intraday 8% rise does not mean the previous adjustment is completely over. The closing gain today fell back to 5.81%, indicating that profit-taking and selling pressure still exist at high levels. Although the Korean market surged, the number of advancing stocks was still fewer than declining ones, indicating the rebound was clearly concentrated in heavyweight stocks like semiconductors. For the US market, SK Hynix's DRAM and HBM business is most directly related to $MU, and the renewed strength in Korean memory stocks helps improve risk appetite in the US memory sector. $SNDK requires judgment in combination with NAND prices and the performance of Japan's Kioxia, and cannot simply apply HBM logic. It has been emphasized before that the US stock market's rise on the 19th and 20th can only be defined as an oversold rebound for now. But today's violent counterattack by Korean semiconductors has indeed brought positive short-term structural changes to the market. After continuous sharp declines, what really deserves observation is not how much it can rebound in one day, but whether these rebounds can form a continuous upward trend. If Korean memory stocks can hold most of today's gains and US semiconductors regain relative strength against $QQQ, then the previously cautious judgment will need to be gradually revised. SK Hynix's big bullish candle today is indeed worthy of praise, but whether it can continue to strengthen afterward is more important than the intraday surge past 8% today.