#AIMemorySelloffEases

The AI Race Has Entered a New Era The AI industry is entering a completely new phase. If GPUs were once the defining factor, memory has now become the biggest bottleneck for the entire AI ecosystem. Exploding demand for AI is turning High Bandwidth Memory (HBM) and high-performance NAND into strategic resources. This is why the industry is calling the current phase the "AI Memory Stress Test"—a real-world test of whether the global memory supply chain can keep up with the next wave of AI innovation. One of the biggest recent developments is that $SKHYNIX and $SNDK have introduced the world's first High Bandwidth Flash (HBF) standard. By combining HBM-class bandwidth with the massive capacity of NAND, HBF delivers up to 3TB/s of bandwidth and 512GB per package, offering a potential breakthrough for AI inference workloads, one of the fastest-growing segments of the AI market. Meanwhile, $SAMSUNG is accelerating its AI memory roadmap with next-generation zHBM and 400+ layer BV-NAND technology, aiming to deliver higher density, better performance, and improved power efficiency as competition in AI infrastructure intensifies. At the same time, hyperscalers including Microsoft, Google, Meta, and OpenAI continue to increase AI infrastructure spending, with demand for HBM expected to outpace supply for the foreseeable future. That imbalance could strengthen the revenue growth and profit margins of leading memory manufacturers. The message is becoming increasingly clear: the AI race is no longer just about GPUs—it is now a race for memory leadership. Companies that dominate HBM, NAND, and next-generation technologies such as HBF, including $SKHYNIX, $SAMSUNG, and $SNDK, could emerge as some of the biggest beneficiaries as the AI supercycle continues to accelerate. #AIMemoryStressTest #Alphabet25BBond #SP500Eyes8000 $SNDK $SKHYNIX

SK Hynix has finally started buybacks, but the market might still be unsatisfied. One of the biggest winners in AI storage, SK Hynix is ready to spend: 💰 100 trillion KRW (about $71 billion) in shareholder returns Among which: 🔥 40 trillion KRW (about $28.4 billion) in stock buybacks This is roughly equivalent to repurchasing 2% of shares, just enough to offset dilution caused by ADR listing. Sounds impressive. But looking at the global storage war, the problem arises: $MU Micron: Directly promises to use 100% of free cash flow for buybacks, with market expectations that by 2028 it may repurchase over 40% of shares. $SNDK SanDisk: $15.5 billion buyback, about 8.7% of market cap. Kioxia: $5.5 billion buyback, about 3.4% of market cap. And Hynix? More like saying: "I will take care of shareholders." But not yet: "I will wildly return AI dividends to shareholders." In the AI era, HBM is the moat. Cash flow buybacks are the answer to investors. The ultimate competition among the future storage big three: Is not just about who sells more chips, But who better converts profits into shareholder returns. The storage war has just entered a more exciting phase. 🚀#存储股财报后续跌，AI内存牛市还稳吗？

Let's talk about the non-farm payroll data. I was stunned after reading it and couldn't say a word for a while. -23,000, the expectation was +80,000, a difference of 100,000. The data for May and June was also revised down by a total of 103,000. The previous two months were revised down by 100,000, and this month it directly turned negative. Honestly, the cooling speed of the job market is quite fierce. (The power of capital is still too strong) But what's even more contradictory is the unemployment rate, which dropped from 4.2% to 4.1%. Employment is contracting, but the unemployment rate is falling. These two data points together indicate mixed signals. Wage growth also slowed, with a month-on-month increase of only 0.1%. After the data release, the probability of a rate hike in September dropped from over 50% to about 44%. The market thinks the Fed can't raise rates anymore. Then the market reaction was very interesting—not a broad rally, but a split. $XAU broke through $4370, futures closed at $4399.7, standing above the $4400 mark. Weak employment → rate hike cooling → weak dollar → gold rises, this chain makes perfect sense. I've been watching $SPCX these past two days. It rose 6% on the unlock day, then surged 15.83% after the non-farm data, closing at $133.11. It climbed from around $105 to $133, a cumulative increase of about 23% over two days. The unlock bearishness has been digested, shorts are covering, and rate cut expectations are pushing it up. The rise is too strong, and I'm the happiest 😂 SanDisk $SNDK plunged from 1326 to around 1200 last night, closing down 3.68%. Weak non-farm data → lower rate hike expectations → high valuation growth stocks should benefit, but SanDisk was hit instead. Previously, despite earnings beating expectations, it fell 7%. This time, even with the non-farm data being favorable, it didn't recover, indicating that the valuation adjustment for AI storage is not over yet. Seagate fell over 10%, Western Digital dropped over 5%, the whole sector is under pressure. Let's expect something. #PayrollsDropCPIFocus

AI memory isn’t dying. But the “AI can only go up” story might be. 👀 🤗 Extra: Is the AI memory bull market still stable? Honestly, I’m not sure. But I think we need to separate two things: AI memory demand is still strong. AI growth expectations are what’s starting to crack. After the August 6 market close, both Western Digital and SanDisk reported numbers that looked strong on paper. SanDisk beat expectations with $8.97B in revenue vs. $8.48B expected, and EPS of $39.25 vs. $34.96 expected. Sounds bullish, right? Except the next-quarter revenue guidance came in at $10.55B, below Wall Street’s $10.82B expectation. And the market absolutely punished it. SanDisk dropped around 8% after hours and opened the next day down as much as 13%. Western Digital got hit even harder, swinging roughly between -11% and -19%. Then South Korea got messy too. SK Hynix briefly showed a bizarre 30% plunge in pre-market trading on Nextrade with just 11 shares traded — basically a liquidity ghost story rather than a real market move. But regular trading still saw real weakness: SK Hynix fell around 10% intraday, closed down 4.97%, and dropped another 3.9% on August 7. Samsung was under pressure too. And then came rumors that Nvidia may be evaluating lower video-memory configurations for Rubin Ultra because of HBM supply constraints. Whether that rumor becomes reality or not, the market clearly isn't treating “AI + memory” as an automatic buy anymore. And that’s the important part. SanDisk had already risen roughly 470% this year. Western Digital was up around 200%. At these valuations, “good results” are no longer enough. The market wants better than expected. That changes everything. Because if companies with real factories, real customers, real products and real revenue can get destroyed simply because guidance isn't amazing enough… What happens to the endless wave of AI + storage, AI + DePIN, AI + GPU tokens in crypto? Some of them have barely any revenue. Some have almost no users. #DailyOrbit

Let's talk about the non-agricultural data. After reading it, I was shocked and couldn't speak for a long time. - 23,000, expected to be + 80,000, a difference of 100,000. 5 The figures for June and June were revised down by 103,000. In the first two months, I made 100,000 yuan, but this month it turned negative. The cooling speed of the job market is quite rapid. (The power of capital is still too strong) But the unemployment rate, which fell from 4.2% to 4.1%. Jobs are shrinking and unemployment is falling, and these two numbers are a confusing signal. Wage growth has also declined, with a month-on-month growth rate of only 0.1%. After the data came out, the probability of a rate hike in September increased from over 50% to around 44%. The market feels that the Fed can't increase. Then the reaction on the market is very interesting, not a general rise, but a split. $XAU broke through $4370, and futures closed at $4399.7, surpassing the 4400 level. Weak employment → interest rate hikes and cooling → weak US dollar → rising gold, this chain runs through. $SPCX I've been watching for the past two days. On the day of the lifting of the lockdown, it rose by 6%, and the non-farm payrolls surged by 15.83%, closing at $133.11. From near 105 to 133, the cumulative increase in two days is about 23%. The negative impact of the lifting of the lock-up has been digested, the shorts are covering, the expectation of interest rate cut is also pushing, the rise is too strong, I am the happiest 😂 SanDisk $SNDK plummeted from 1326 to around 1200 last night, closing down 3.68%. Weak non-farm data → Decreased expectations of interest rate hikes → High-valued growth stocks should benefit, but SanDisk has been hit. The previous financial report exceeded expectations and fell by 7%. This time, the non-farm benefits did not recover, indicating that the valuation of AI storage has not been fully digested. Seagate fell more than 10%, Western Digital fell more than 5%,$BTC

Gold breaks out. Bitcoin holds. AI storage enters the “prove the returns” phase. A fresh market view from 1011 Insider Whale agent Garrett Jin highlights three important shifts: GOLD Gold has broken out of its previous consolidation range and is now viewed as a potential long-term allocation opportunity. The core thesis remains unchanged: a long-term weakening trend in the U.S. dollar could continue supporting gold. BITCOIN Since BTC approached the $57,700 low in July, the market has continued to satisfy key bottom-building conditions. No change in stance: BTC positions established around $60K are still being held. AI STORAGE The long-term demand story remains strong, but the short-term setup has changed. Garrett Jin previously favored accumulating storage chips on pullbacks. After the market rallied without giving that pullback, he sold half of the rebound positions. Not because the thesis is broken—but because the latest upside appears increasingly driven by capital structure and positioning. The bigger issue is leveraged Korean storage ETFs. JPMorgan estimates exposure has fallen 66% from the peak, while another calculation puts the decline closer to 38%. Either way, leverage has weakened—but has not disappeared. And the demand story remains powerful: • SK Hynix 2026 capacity reportedly sold out • Micron orders extend toward 2028 • AI storage demand could remain strong into H2 2027 But storage is still a cyclical industry. After several hundred-percent rallies, valuation expansion alone becomes harder to sustain. The AI investment cycle may now be moving from: CAPEX → EXPECTATIONS → REVENUE VERIFICATION The market is no longer simply asking: “Who is spending the most on AI?” It is starting to ask: “Who can actually turn that spending into revenue?” That shift could define the next phase of the AI trade. $BTC $XAU #PayrollsDropCPIFocus #Gold4300EasingOrHedge

📊 AI Demand Is Strong—But That Doesn't Guarantee Higher Stock Prices This week's earnings from memory and storage companies highlighted an important lesson: strong AI demand doesn't automatically translate into rising AI-related stocks. Western Digital reported solid results, posting around $3.75B in quarterly revenue and $3.56 in adjusted EPS, yet the stock still came under pressure. SanDisk also delivered a strong quarter with approximately $8.97B in revenue, but investors were disappointed because future guidance didn't exceed the high expectations already priced into the shares. The market is no longer questioning whether AI needs more storage and memory—it largely accepts that demand will continue growing. The bigger question is whether companies can maintain pricing power, supply shortages, and profit margins strongly enough to justify valuations that have already anticipated years of future growth. That's a much tougher benchmark. Both SanDisk and Western Digital rallied significantly during the AI infrastructure boom, so simply beating earnings estimates was no longer enough. Investors wanted clear signs of another phase of accelerated growth. At the same time, supply conditions remain tight. Industry developments—including the expanding partnership between NVIDIA and SK Group on next-generation AI memory and ongoing constraints in HBM supply negotiations—suggest that demand continues to outpace available capacity. There are also discussions that future AI chip designs could adjust memory configurations due to packaging and supply limitations. If that happens, it shouldn't automatically be viewed as weaker AI demand—it may simply reflect engineering decisions based on supply-chain realities. For me, the key question is no longer: "Will AI require more memory?" Instead, it's: "Can memory manufacturers turn ongoing supply constraints into sustainable earnings growth before market expectations become too optimistic?" That's the metric worth watching. #PayrollsDropCPIFocus #AIMemoryStressTest #SpaceXUnlockRebound

$SNDK — I’m positioned short here, expecting the pullback to continue. SanDisk sold off after earnings, and the market reaction was pretty clear: the results themselves weren’t bad, but expectations had simply become too high. The bigger problem was the guidance. Next quarter’s revenue midpoint came in at $10.55B, roughly $600M below the market expectation of $11.16B. That gap was enough to trigger an around 8% after-hours decline. Looking at the price action, $SNDK dropped sharply from around 2,300 and has struggled to recover despite multiple attempts by dip buyers. The lack of a meaningful rebound suggests selling pressure is still present. The broader storage sector is also showing weakness. SanDisk, Western Digital, and Micron have all come under pressure recently, which stands out given the broader rebound across U.S. technology stocks. For now, the sector weakness and failed rebounds keep me leaning bearish on $SNDK. #PayrollsDropCPIFocus #AIMemoryStressTest #SpaceXUnlockRebound

#Nonfarm unexpectedly turns negative, CPI becomes the key to rate hikes The nonfarm data came out at -23,000, with an expectation of 80,000, directly turning negative. But the unemployment rate dropped to 4.1% because the labor force participation rate fell, just like in previous instances, the data is conflicting with itself. The market reacted quickly, with the probability of a rate hike in September dropping to 44%, while Kalshi's probability of maintaining the current rate actually rose to 65%. Short-term employment data has indeed loosened rate hike expectations, but the real issue is next week's CPI. If CPI rises again, rate hike expectations will bounce back. As always, employment data and inflation data are pulling against each other, with both bulls and bears waiting. $SNDK SanDisk is currently trading in the 1200-1300 range. I estimate that in about a week, SanDisk will likely oscillate between 1200 and 1300. If it breaks below, there will be buyers; if it rises, there will be selling pressure, so both upside and downside are limited. Previously it hit a low of 1186 but did not break 1100, which is better than I expected. This indicates there is still support at this level, and the market has not completely given up on the medium- to long-term logic of storage. The grid strategy has paused because it stopped below the lower boundary of the range at 1219, with the price around 1216. The strategy is paused, but the base position remains. When the price returns above 1219, the grid will automatically resume. If SanDisk really trades in the 1200-1300 range for a week, the grid can resume and continue running, making this range sufficient for grid arbitrage. #PayrollsDropCPIFocus #SpaceXUnlockRebound #AIMemoryStressTest

The AI Race Has Entered a New Era The AI industry is entering a completely new phase. If GPUs were once the defining factor, memory has now become the biggest bottleneck for the entire AI ecosystem. Exploding demand for AI is turning High Bandwidth Memory (HBM) and high-performance NAND into strategic resources. This is why the industry is calling the current phase the "AI Memory Stress Test"—a real-world test of whether the global memory supply chain can keep up with the next wave of AI innovation. One of the biggest recent developments is that $SKHYNIX and $SNDK have introduced the world's first High Bandwidth Flash (HBF) standard. By combining HBM-class bandwidth with the massive capacity of NAND, HBF delivers up to 3TB/s of bandwidth and 512GB per package, offering a potential breakthrough for AI inference workloads, one of the fastest-growing segments of the AI market. Meanwhile, $SAMSUNG is accelerating its AI memory roadmap with next-generation zHBM and 400+ layer BV-NAND technology, aiming to deliver higher density, better performance, and improved power efficiency as competition in AI infrastructure intensifies. At the same time, hyperscalers including Microsoft, Google, Meta, and OpenAI continue to increase AI infrastructure spending, with demand for HBM expected to outpace supply for the foreseeable future. That imbalance could strengthen the revenue growth and profit margins of leading memory manufacturers. The message is becoming increasingly clear: the AI race is no longer just about GPUs—it is now a race for memory leadership. Companies that dominate HBM, NAND, and next-generation technologies such as HBF, including $SKHYNIX, $SAMSUNG, and $SNDK, could emerge as some of the biggest beneficiaries as the AI supercycle continues to accelerate. #AIMemoryStressTest #Alphabet25BBond #SP500Eyes8000 $SNDK $SKHYNIX