Brothers losing money, don't be discouraged This handsome guy lost 10 billion USD at 25 years old #PCE环比转负，GDP增速放缓至1.5%

This is what a genius trader looks like #PCE环比转负，GDP增速放缓至1.5%

How do experts in the English-speaking community view storage? I've compiled public opinions from 50+ KOLs I've organized public discussions from over 50 English-speaking KOLs (macro / US stocks) from roughly the past two weeks, covering $MU, $SNDK, $SKHY, and storage sector linkages. Below, I'll walk you through the market sentiment and thought processes in the English-speaking community so you can clearly understand how experts view this round of brutal market conditions. ─── I. Bullish Logic 1. Big tech capital expenditures haven't been cut; claims that AI will reduce spending don't hold up @aleabitoreddit believes: The narrative of "excess computing power and cloud providers cutting spending" has been disproven in recent weeks; Amazon and others are still confirming computing power shortages, which is bullish for upstream semiconductors. @DavidMorganInv states: Amazon's earnings report directly shatters the "AI computing power surplus / big tech cutting spending" narrative; Google, Meta, Amazon, and Microsoft all confirm shortages. @danadams_46 says: After this quarter's earnings, AI infrastructure hasn't weakened but strengthened; Google, Microsoft, and others continue to increase AI capital expenditures, with money flowing to the hardware chain, so rebounds in $MU and others are unsurprising. 2. Amazon raises capital expenditure and specifically mentions memory becoming more expensive @aleabitoreddit notes: Amazon's 2026 capital expenditure increased from about 200 billion to about 220 billion, partly due to higher memory costs, which is bullish for $MU and Samsung. @mahvid123 says: The extra spending mainly comes from rising memory prices, not just building more data centers—storage costs are rising first even with the same construction. @DavidMorganInv adds: Even with the 220 billion figure, Amazon's spending still doesn't cover 2026 demand, 2027 is tight as well, and capacity is already booked further out; upstream supply chains including storage remain constrained. 3. Industry shortages persist; memory prices are still rising @PhotonCap believes: A big part of this rebound comes from Samsung's warning—DRAM, NAND, and HBM shortages may last until 2028; this weakens short-term worries that ChangXin's expansion will cause oversupply. @ren_stocks notes: While DDR5 spot prices are still rising, storage stock prices fell first; for MU and SKHY, rising DDR5 prices for servers almost directly translate to profits, so they are bullish on storage. @SemiconductorsX retweets Morgan Stanley's view: The big drop in US storage stocks is actually a buying opportunity; data center/AI shortages persist, while PC/consumer noise exists; DRAM price increases and shortage timelines remain expected. @PolymarketMoney also emphasized Morgan Stanley analysts calling SNDK and MU pullbacks strong entry points. 4. HBM and product cycles indicate this is not just sentiment-driven @PhotonCap says: $MU has started volume shipments of HBM4 for Nvidia's Vera Rubin and looks ahead to HBM4E; the increase in memory usage per AI server remains a structural driver. @mrcooler6868 believes: The market seems to be selling $MU and $SNDK on the assumption that "AI memory/storage demand has peaked," but this is a mistaken sell-off—HBM volume growth and SanDisk's high data center revenue growth continue unless supply catches up with demand first. 5. The crash involved forced liquidations; after selling pressure subsides, a rebound follows @joveg8 summarizes: The recent crash was related to forced selling; after related funds publicly finished liquidating positions, manual selling pressure decreased; Microsoft, Meta, and others' earnings reassure that "AI spending continues"; Samsung's shortage narrative supports pricing power; when SNDK and MU surged, WDC, STX, and storage ETFs moved together. @MacroAlphaHQ believes: The simultaneous surge in SNDK and MU indicates bottom-fishing inflows; storage is a highly elastic semiconductor segment, with under-allocated positions chasing and shorts covering, so buying can self-sustain—but it remains to be seen if this spreads to broader chips. 6. After a deep price drop, a second bottom-fishing opportunity appears @ianmirikwa uses price trends to say: SNDK, MU, and $SKHY dropped from highs then rebounded, giving the market a second chance; most say it's over, but some buy at discounted prices amid panic. @lastman3344 says: The pullback involved profit-taking and leverage washout; HBM sold out, prices rose, data center demand remains, and capital is flowing back; this is a structurally driven rebound, not a confirmed full trend reversal, so it needs to be defended. 7. For SK Hynix, watch HBM positioning and insider moves @danadams_46 singles out $SKHY: Still at the center of the AI memory cycle; the drop is sentiment-driven. In research windows, @StockSavvyShay and others reported SK chairman increasing $SKHY holdings; @WhaleInsider and others reported UBS and others having a positive coverage and target price on SK Hynix—English-speaking community interprets this as sell-side and insiders not accepting that fundamentals are dead. 8. The sector moves together; it's not a single-stock story @MacroAlphaHQ and @joveg8 both emphasize: It's not $MU or $SNDK each telling their own story; the entire storage chain is moving together. Trading implication: You're playing storage sector elasticity, not suddenly loving only SanDisk today. ─── II. Bearish Logic 1. Bubble and peak @FinanceLancelot wrote during the big drop: SNDK, MU, and $SKHY collapsing together may signal the end of the memory and AI bubbles; later framed it as a "bubble returning to normal"—retail investors tend to go all-in at the wrong stage, with big money unloading the last leg. 2. Bears have real market representatives: Burry increasing shorts on $MU @StockSavvyShay, @EhrmantrautCap_, and others discussed Michael Burry increasing shorts on $MU. @EhrmantrautCap_ bullishly retorted that shorts will be squeezed; some accounts agreed with Burry, unwilling to hold Samsung, SK Hynix, Micron, and SanDisk anymore. Bear logic: Not based on sentiment but on real money shorting at cycle tops. 3. The rebound is for reducing positions; if it doesn't hold, it's over @MacroAlphaHQ: If it doesn't spread to broader semiconductors, this is pure position behavior; today's bottom-fishing is tomorrow's selling. @lastman3344 also noted failure conditions: If the close doesn't hold and it's only sentiment-driven, don't treat it as a trend. 4. High-level pullbacks indicate large volatility damage @Mr_Derivatives and others emphasized $SNDK's significant pullback from historical highs. Bear interpretation: Elasticity is two-way; a 20% one-day rise doesn't prove risk is gone; poor position management can cause losses even if logic is correct. 5. Supply and competition will eventually pressure prices @PhotonCap's question itself reflects bear concerns: Will ChangXin impact MU, SKHY, Samsung? Bulls counter with "shortages lasting until 2028," bears focus on mid-term supply. Many English KOLs also mention China's capacity expansion, customer self-development, and DRAM commoditization—profit margins are unlikely to stay at extreme highs long-term. ─── There are also many neutral articles from KOLs, which I won't include here. If anyone is interested, feel free to leave a comment 🫡

Recent Major Events in the AI Circle (Avoiding Mindless Hype) 1. DeepSeek: V4-Flash Official Version API Public Beta • DeepSeek announced the public beta of the DeepSeek-V4-Flash official version API • Officially stated that Agent capabilities have significantly improved, with multiple metrics surpassing the previous V4-Pro preview version • Community/benchmark comparisons: multiple benchmarks exceed GLM-5.2; some long-range programming tasks approach Claude Opus 4.8 (e.g., DeepSWE scores rose from about 7 to just over 54, Opus about 59) • Pricing approximately $0.14 input / $0.28 output (per million tokens); cache hit rate is lower • Architecture and parameter count remain basically unchanged; this upgrade is mainly post-training; V4-Pro official version not yet released Community testing: Independent organization Artificial Analysis comprehensive index about 50, close to GPT-5.6 Luna and GLM-5.2 (about 51), not a clear leader; open-source overall ranking still behind Kimi K3. Some high scores may include official Agent scaffolding, awaiting third-party retesting. Still pure text; peak time prices may double 2. OpenAI: GPT-5.6 Price Reduction • GPT-5.6 Luna price cut about 80%, approximately $0.20 / $1.20 (million input/output) • GPT-5.6 Terra price cut about 20%, approximately $2 / $12 • Sol price unchanged, added faster Fast mode (about 2.5x speed, 2x price) Note: Price cuts apply to mid- and low-tier models; flagship models remain unchanged; still relatively more expensive than some Chinese models 3. MiniMax: H3 Video Model Release • Released H3, supports text, image, video, and audio input • Can generate about 15-second, 2K video with native sound • Officially stated weights will be open-sourced soon; cost claimed to be about one-third lower than mainstream competitors for 2K video 4. ByteDance: Seedance 2.5 Release • Released Seedance 2.5 around the same time as MiniMax • Can generate about 30-second videos per session, emphasizing long narratives and multiple reference materials • Will launch products like Jimeng and Doubao, with APIs gradually opening 5. Zhipu GLM: Coding Package Price Increase • GLM Coding / Token related packages increased in price; subscriptions reopened • Community feedback: entry-level tier roughly doubled or more; high-tier reportedly increased from about $22 to $79 This is really a big mess ───