Orbit Post Sitemap
三星利润暴增19倍后股价暴跌，科创50受多大冲击？
7月7日，三星电子发布了一份利润同比暴增19倍的炸裂财报，股价却应声暴跌超10%。这还不算完——一个月内，三星电子较6月高点最大回撤约36%，同为韩国存储巨头的SK海力士更惨，从高点回落了约44%。
这场“利好出尽”的踩踏，迅速从韩国蔓延至全球，费城半导体指数跌入技术性熊市，也把科创50指数拖下了水。
海外半导体暴跌，对科创50的传导效应到底有多大？答案不是简单的“跟跌”二字能概括的。
这次暴跌的根源，不在基本面，而在交易结构。
上半年，三星电子、SK海力士股价均翻倍以上，约70%的涨幅由HBM涨价预期驱动，早已透支未来2-3年的业绩增长空间。而韩国监管层5月一口气推出16只跟踪这两家巨头的2倍杠杆ETF，散户普遍以5倍融资杠杆参与交易。
当7月初Meta宣布出售闲置算力、叠加“英伟达下调HBM采购量”的传闻，市场对AI存储需求见顶的担忧瞬间引爆，股价快速下跌触发大规模强制平仓，形成了“越跌越卖”的负反馈循环。
这轮暴跌，本质是韩国杠杆资金的结构性崩塌，而非存储行业的需求逻辑被证伪。
从历史数据看，科创50与海外半导体之间确实存在稳定的联动关系。
本次行情中，短期波动弹性略高于历史均值。费城半导体指数最大回撤约22%，韩国半导体板块最大回撤约27.5%，科创50同步最大回撤近27%，短期弹性偏高。
传导效应并非均匀分布。7月1日至22日，科创50关联的存储、PCB、光模块细分板块跌幅达25%-35%，成为重灾区。其中万得电路板指数跌34.71%，光模块指数跌30.46%，存储器指数跌29.83%。
原因很直接：这些板块直接对标海外存储龙头的业务逻辑，当三星、SK海力士提出下半年基板降价要求，恐慌情绪迅速放大。PCB板块还叠加了“部分龙头因质量问题失去大客户”的传闻，虽然后来澄清，但伤已造成。
而半导体设备板块在7月24日逆市拉升，IDC指数同期仅下跌3.76%，并在7月20日后止跌回升。分化背后是逻辑差异：设备板块受益于国产替代的确定性订单，IDC则直接锚定国内算力需求，与海外存储价格的短期波动关联度较低。
资金面的多空博弈，进一步放大了科创50的短期波动。
7月1日至23日，科创50成分股中寒武纪、德明利等存储/AI芯片标的的融资净卖出额超过50亿元，半导体、硬件设备类标的包揽A股同期融资净卖出前十全部席位。杠杆资金集中平仓，成为短期放大跌幅的核心因素。
但另一边，长线资金却在逆势布局。科创50ETF平安在此期间连续11天获得资金净流入，合计“吸金”9.82亿元。这种“短期投机盘恐慌出逃、长期配置资金逢低入场”的拉锯格局，说明专业机构对科创50中长期价值的判断并未动摇。
主流机构对此轮传导的性质判断一致：短期情绪冲击，非基本面恶化。
前海开源基金杨德龙则直言，本轮调整是海外科技巨头重挫与A股内部交易结构高度拥挤等多重因素共振的结果，主要是资金获利回吐与出逃带来的反应，并不是‘科技牛’行情的终结。
尽管短期情绪冲击正在消退，但两个风险点仍需警惕。
二是“英伟达下调HBM采购量”的传闻虽未得到官方证实，但若后续有实锤信息落地，会进一步抬升全球存储板块的调整幅度，加大对科创50的传导压力。
把各方视角拼在一起，结论很清楚：本轮海外半导体回调对科创50的传导效应，短期内因交易结构拥挤、杠杆资金集中平仓而出现阶段性超跌，但核心冲击以纯情绪传导为主，未发生实质性的产业链基本面变化。
7月27日韩存储双雄公布大额长期订单后股价快速回升，也验证了短期情绪主导的特征。
中长期来看，科创50的传导效应有限——国内存储产业链的国产替代逻辑、AI算力需求的结构性增长，并不会因为韩国杠杆资金的踩踏而改变。
至于那些受HBM采购量传闻影响的尾部风险，它们更像是悬在头顶的“第二只靴子”——在落地之前，保持警惕是必要的，但没必要因为一颗不确定的炸弹，就把整栋楼拆了。#长鑫科技上市，全球存储竞争添变量 $SAMSUNG Bitcoin $BTC miner winter is spreading!!!
The total network hashrate has dropped to 908 EH/s, hitting a new low for 2026.
It now costs $78,000 in hashrate to mine one $BTC, which is higher than the spot price of around $65,000, meaning miners lose over $10,000 for every Bitcoin mined!!!
Outcome:
Forced shutdowns and marginal clearing: High electricity prices and inefficient old models (such as some S19 series) have completely breached the shutdown price. Miners "queue to shut down" as a rational choice to preserve capital. The hashrate reduction is the market clearing marginal high-cost capacity to rebalance supply and demand;
Lagged selling pressure release: To maintain fiat operating expenses and repay equipment debts, unhedged miners have to accelerate selling their inventory reserves, which in the short term puts selling pressure on the spot price. $BTC The Fed decides Wednesday, and for once the debate isn't about a cut.
CME FedWatch puts a hold at roughly 62%, with the other 38% actually pricing in a possible hike. Rates have sat at 3.5% to 3.75% since June, so with no cut on the table, this is a tone meeting, not a number meeting.
The macro backdrop walking in:
· US-Iran ceasefire hopes knocked oil sharply lower, cooling the inflation scare
· Jobless claims came in at 187K, better than expected, so the labor market still looks solid
And it's a loaded 48 hours, because the signals stack fast:
· The statement drops 2:00 PM ET, but the real read is Warsh's press conference at 2:30, and he keeps telling markets to watch the data, not the forward guidance
· GDP and PCE land right behind the decision, compressing four macro signals into two days
· Microsoft, Meta and Amazon report this week, with AI capex guidance in focus
· Crypto-linked earnings land just hours after the Fed
· July BTC and ETH options expire right after that, adding a layer of derivatives clearing to the mix
Crypto is walking in cautious. Bitcoin is holding near $65,000 after a seven-day spot ETF inflow streak worth nearly $1B snapped on July 24, though the week still closed net positive. Sentiment stays fragile, with the Fear & Greed Index ticking up to 30, a monthly high but still firmly in Fear territory.
Put it together and you have plenty of fuel for a sharp move in either direction once the tone lands. The number is basically priced. The words are not.
Are you positioning ahead of the decision, or waiting until the dust settles before making a move?
#FOMCRateWatch 想明白一个问题，为什么这轮AI投资潮中，科技巨头们从最初的部分现金流投资，逐步变成了现在的发债融资…
有没有一种可能，美国为了“美债”这盘醋，才包了“AI”这么一盘饺子？
现在的美国债务以及信用状态，别说是让全球资本继续无脑买美债，就是让几大持有国别那么快抛售都难…
既然国家债你们不买， 那就来买名为 AI 的企业债吧！
反正不论是投资到 AI 科技公司还是投资到美债都差不多了，只要钱进来 USA 就行，不是么？
所以我现在看这些巨头还在不惜把现金流搞成负数也要加大 AI 投资的行为，不知不觉间给我一种“政治任务”的味道…
某种程度上，美国的科技巨头公司们，就是资本主义国企啊…Oil prices have crashed, BTC returns to 65k: The market has already "voted" ahead of Thursday's FOMC
Brothers and sisters, first look at three things:
First — This morning, Brent crude plummeted 5%, briefly falling below $90 during the session. WTI dropped to around $84.
Why? The US and Iran paused mutual military strikes over the weekend. After nearly five months of conflict, oil prices fell from above $100.
Second — Last week, US initial jobless claims dropped to 187,000.
The lowest since 1969.
Third — Bitcoin is above $65,000. The fear and greed index is 30, still in the "fear" zone.
Put these three things together, what do you make of it?
Most people are focused on the FOMC interest rate decision early Thursday morning, watching whether there will be a rate hike.
CME data shows a 63.7% probability of holding rates steady in July, and a 36.3% chance of a 25 basis point hike. Natixis, Nomura, and Morgan Stanley all unanimously expect no change.
But that's not the main point.
The key is this script:
What was the market most afraid of a month ago? Oil prices breaking $100, a resurgence of secondary inflation, and the Fed being forced to raise rates.
What happened? One week before the FOMC meeting, the US and Iran declared a ceasefire. This "time bomb" of oil prices was defused early.
Inflation expectations cooled → US Treasury yields' downward pressure eased → the tightening grip on risk assets loosened.
The market has already voted ahead.
BTC rebounded from last week's low back to 65,000, leaving many who missed out. The fear index is still at 30, which means what? A large amount of capital is still watching, waiting for the FOMC "boot to drop" before rushing in.
By the time you rush in, the smart money is already counting profits.
Now about the initial jobless claims at 187,000.
The lowest since 1969 — what does that mean? The economy is not at risk of recession.
The Fed does not need to cut rates early to save the economy. But it also means that "no landing" is becoming a reality. What the market fears most now is not a rate hike, but the continuation of "no rate cuts."
As long as Powell says "inflation is slowing," 65k is the iron bottom.
There are two variables exploding simultaneously this week:
Microsoft, Meta, and Amazon earnings reports will be released Wednesday and Thursday.
These three companies' combined capital expenditures for 2026 — Microsoft $190 billion, Meta up to $145 billion, Amazon $200 billion. All are "money-burning machines" for AI infrastructure.
Good earnings will boost risk sentiment; poor earnings will drive funds to safe-haven assets — is BTC a beneficiary or a victim? You decide.
Finally, a painful note:
FTX's fifth round of $900 million compensation starts July 31. Some creditors will receive 120% compensation.
What will these people do with the money?
Do you think they will save it in banks or buy BTC?
Don't wait for the FOMC to land before chasing.
Oil prices have already fallen, employment data is out, and the market is already pricing in the "cooling inflation" factor.
On the day of the meeting, it's always about good news being realized or bad news fully priced in.
The real game is in the "expectation gap."
If Wash says something soft, 65,000 becomes 68,000.
If Wash says something hard, 65,000 becomes 62,000.
But at this position, where do the odds stand? You do the math yourself.
$BTC $ETH $CL
#美联储周四凌晨公布利率决议 #长鑫科技上市，全球存储竞争添变量
今天A股变天了。
长鑫科技今天正式登陆科创板，开盘直接干到49.5元，较发行价8.66元暴涨471.59%。市值3.31万亿，直接干翻工商银行，成了A股新一哥。中一签赚2万块。募资666亿，科创板历史第一，A股历史第三。
一个做存储芯片的，一天之内成了中国市值最高的公司。
长鑫科技什么来头？中国唯一实现DRAM规模化量产的企业。全球DRAM份额从3%干到8%，全球第四。上半年营收1100到1200亿，净利润500到570亿，同比暴增最高2544%。
更狠的是价格。路透社报道，长鑫存储的64GB DDR5服务器模块报价已经超过了三星。中国芯片不但做出来了，还卖得比三星贵。
这对全球存储格局意味着什么？
以前DRAM市场就三家，三星、SK海力士、美光，垄断90%以上。现在长鑫杀进来了，8%的份额，第四把交椅坐稳了。
三星39%、SK海力士29%、美光22%，长鑫8%。三巨头变四巨头，只差一个身位。有预测说2026年底长鑫产能就能逼近美光。
但长鑫也有短板。HBM高端市场还在三星海力士手里，长鑫在堆叠工艺和高端产品迭代上大概还有2到3年代差。
长鑫上市对加密市场什么影响？两个方向。
短期看，巨量IPO抽血。666亿募资，A股历史上第三大IPO。资金从各个市场往A股流，加密市场流动性短期承压。有分析说AI新上市可能逐步吸收流动性，收紧流向数字资产的边际资金。
但另一个角度，海外投资者买不到科创板，转头在加密合约市场炒长鑫的“影子股”。这说明什么？国际资本对“中国存储国产替代”极度看好，而加密市场正在成为全球资金配置中国硬科技的中转站。
大饼作为高贝塔风险资产的属性在强化。宏观情绪稳，风险偏好回升，加密跟着受益。长鑫上市抽走的是流动性，但点燃的是对中国科技资产的信心——这层情绪外溢，加密市场也能分到一杯羹。
这就是短期拉伸的大部分原因，那么你吃到了吗？兄弟们
$BTC $ETH $DOGE #SemisEarningsTest 🧠 Chips are bouncing—but the positioning tells a deeper story.
Semiconductor names are showing strength:
🟢 Ambarella +6.24%
🟢 Teradyne +3.54%
🟢 Marvell +3.32%
Meanwhile, the broader market closed slightly lower.
Bulls and bears are both making their moves as earnings season begins. 👀
But market positioning reveals another layer:
📉 S&P 500 short interest has climbed to 3.79% of float
📉 Russell 3000 short interest reached 6.3%
Both are at record highs.
Hedge funds have also been reducing US tech exposure, with net selling in 6 of the past 8 weeks and roughly a 10% reduction—the largest pullback in more than a decade.
Then there’s the debt picture many investors aren’t watching.
Reports indicate that off-balance-sheet debt among five major tech companies has expanded significantly, reaching around $1.65T, surpassing their reported on-book debt. Meta’s exposure alone is estimated around $420B, while major financing efforts are underway to support AI data center expansion.
The setup heading into Big Tech earnings:
⚡ Chips are recovering
📊 Short positioning is extreme
🏦 AI infrastructure spending is accelerating—but so are financial commitments
This earnings season could answer some major questions:
Will strong results trigger a short squeeze?
Or does the market turn it into a classic “buy the rumor, sell the news” event?
What matters more for valuations right now—the AI growth story or the rising debt burden? 👇
#CXMTMemoryIPO #FOMCRateWatch Chun Wang (@satofishi) has transferred around $6.73 million in crypto to Binance, a move that's already catching traders' attention. Deposit breakdown: 🐋 3,490 $ETH (~$6.70M) ₿ 3.51 $WBTC (~$228.35K) The latest transfer hit Binance just minutes ago. Large exchange deposits don't automatically mean selling, but they often signal that a major holder is preparing to increase liquidity or reposition their portfolio. For now, it's something worth keeping on your radar—not because it's guaranteed to Today, as soon as I opened the market software, I saw this news: $STORJ plunged from 1.2 to 0.38, and a needle made people's scalps tingle. The star decentralized storage project that once wrestled with $FIL is now applying for a Chapter 11. Bankruptcy Protection: To be honest, I checked my wallet right away. Luckily, I cleared out in Q3 last year. At that time, I felt on-chain activity was dropping, and the storage sector was fiercely competitive. Projects that didn't make money would eventually collapse. Don't panic yet. Interestingly, the announcement said it was "exploring a court-recognized equity mechanism for token holders." Translated plainly—those holding $STORJ might be able to exchange for company shares, but no one knows how deep the trap is 11 is debt restructuring, not liquidation, which shows the team is still struggling. The network keeps running, but who knows how much token value is left? Several friends I know who heavily hold $STORJ all posted tearful emojis in the group today. One said his cost was at 0.8, more than halved, and he was preparing to lie flat — that "equity fairy tale" was realized. Honestly, decentralized storage has been burning money for five years, and this year $STORJ is the first to collapse. I can't imagine who will be next. Does anyone else feel the same way? The comments section #特朗普将决定是否扩大对伊战事 #伦理条款获特朗普认可, with #芯片股反弹 of divergences still lingering and short positions in US stocks reaching a historic high Microsoft, Meta, and Amazon's earnings reports this week are no longer ordinary reports
They have to answer a harsh question: Is AI really a profit engine or a money-burning machine?
Google and Tesla have already scared the market earlier, and investors are no longer as romantic about AI spending as before. Previously, when people talked about building more data centers and buying more GPUs, everyone would automatically imagine future growth; now the market is starting to ask whether cloud revenue is enough to sustain, whether advertising can be monetized, and whether users are really willing to pay for AI
I think the highlight of this earnings season is not the EPS decimal point, but the capital expenditure guidance
Whoever can prove that the money burned will come back can still tell the AI narrative. Those who only talk about long-term value will see their short-term stock price pulled back by reality
The story is still there, but the market is starting to ask for receipts
#财报观察员：微软Meta亚马逊能稳住AI叙事吗？ The Federal Reserve early Thursday morning may not necessarily raise interest rates, but the market will still be nervous.
The current trouble lies in oil prices, employment, AI capital expenditure, and fiscal issuance, all adding pressure to interest rates. In the past, everyone only focused on whether rates would be cut or not; now, the question is whether the Federal Reserve can still provide the market with a comfortable space for imagination. As long as it continues to emphasize inflation risks and data dependency, highly volatile assets like BTC and ETH will find it difficult to completely shake off the shadow of interest rates.
I am more concerned about the post-meeting wording than the single outcome.
If the Federal Reserve makes the market feel that high interest rates will last for a long time, risk assets will first contract in valuation; if it is willing to give some easing space, then funds will have the courage to seek elasticity again.
What the crypto market fears most is not bad news, but "not knowing when money will become cheap."
#美联储周四凌晨公布利率决议 不管 BitMEX 还是 Bitmart，都不是暴雷，只是非一线离岸交易所赚不到钱+看不到希望，所以主动关闭。
大环境看，没人炒币，总盘子剧烈收缩
从地域看：美韩欧三个大市场，都合规了，离岸空间越来越小；俄伊，再做下去人头不保；华语极度拥挤，被一线占满；其他小市场，不赚钱。
预计除一线外都要关#长鑫科技上市，全球存储竞争添变量
ChangXin Memory Technologies went public on the STAR Market today. Has the global memory landscape truly been rewritten? My judgment: yes, but only halfway.
688825, issue price 8.66, opened soaring to 49.5, market cap surged to 3.3 trillion, the largest IPO in STAR Market history.
Globally, DRAM was originally dominated by three players holding over 90%: $SKHYNIX 33.96%, Samsung 34.48%, $MU Micron 23.41%. According to 25Q4 sales, ChangXin takes 7.67%, ranking fourth, with 17nm DDR5/LPDDR5X yields steady above 90%, and performance gap narrowed to within 5%.
But don’t get carried away, here’s something different:
ChangXin has broken the oligopoly of general-purpose DRAM, not the AI memory hegemony.
In HBM, Hynix and Samsung hold over 90% market share, while ChangXin’s HBM3 is only in small batch production, with a 1.5-2 year generation gap. The next three years will likely be a layered scenario—
The Korean and American three companies control the lifeblood of AI server HBM, while ChangXin takes the general-purpose market for PC/mobile/domestic servers, gradually chipping away at Micron’s mid-to-low-end share.
A word to my crypto brothers:
Memory is the hidden position of AI computing power. ChangXin entering the table = domestic computing cost curve shifts down = the bottleneck in capital expenditure for domestic large models/cloud providers begins to loosen.
But the 3.3 trillion market cap already fully prices in the 2026 profit expectation of over 50 billion. Short term is sentiment-driven; mid term depends on whether HBM can break in before 2028, otherwise it’s just a cheaper version of Micron, not the Chinese version of Hynix.
My stance is clear: ChangXin is a strategic variable, not a valuation bargain; it rewrites the supply pattern but does not rewrite the top-level order of AI memory.Market Fragmentation Seen from the Meme Coin Riot!
Retail is in celebration, institutions are waiting for signals!
The most striking phenomenon in late July was not Bitcoin's price fluctuations, but $SHIB's single-day surge of nearly 37%, with $PEPE and $DOGE strengthening simultaneously, while BTC and ETH were almost stagnant. This is not simply a "risk appetite rebound," but a typical market split.
Retail funds are expressing their attitude with their feet: when the market has no direction, they seek excitement in highly elastic assets. On-chain data shows that some dormant whales have re-entered the market, exchange net outflows continue, and burn volume has increased. This is a classic pattern of retail investors leading the market. Meanwhile, after consecutive inflows, US spot Bitcoin ETFs have seen significant single-day outflows, with institutions remaining cautious.
This split is not new in 2026. Whenever macro uncertainty rises (first rate meeting during Walsh's term, unclear progress on the CLARITY Act, fluctuations in geopolitical oil prices), funds stratify: institutions with long-term allocation needs choose to wait and see or slowly accumulate shares, while retail investors with trading needs flock to the most volatile commodities. As a result, BTC volatility is suppressed to extremely low, while the volatility of altcoins and memes is amplified.
Structurally, this is unhealthy, but it's also unsustainable. Historically, phases similar to retail frenzy + institutional absence were either ended by a clear macro/regulatory catalyst and turned into a full-blown bull market, or ended with rapid pullbacks. Currently, it is closer to the early stages of the former. As long as the late July policy meeting gives clear signals of easing or ending QT, funds will quickly flow back from high-beta assets to BTC and mainstream assets.

For traders, the most important thing right now is not chasing the already rampaging Meme, but to observe two indicators: whether ETFs have returned to sustained net inflows, and whether large wallets have started accumulating again. Once these two signals appear simultaneously, market splits will end.
#多数党领袖称CLARITY休会前难通过 Am I born to be the one to point back? Every time I buy, prices drop
Not really talking about itself, but from Robinhood Chain's recent performance
I felt a familiar rhythm
On-chain speculation on Robinhood Chain remains active
Several tokens hit new highs in market capitalization today
Then guess what
On one hand, the market is speculating on the new token of Robinhood Chain
On one hand, SUI, EIGEN, and FF all saw big unlocks this week
Lighting a fire while splashing water
This sense of division is actually very familiar
The popularity of Robinhood Chain is real
Robinhood's user base is well established
Traditional stock users are experiencing on-chain assets for the first time
The entry effect is very strong
But unlocking is also an objective selling pressure
SUI has unlocked quite a lot this week
EIGEN is no exception
Can the market catch these unlocking chips?
A question mark is needed
Additionally, there is another piece of data that caught my attention today
BTC ETFs saw outflows of 225 million yesterday
Although BTC prices did not fall
But if this outflow continues,
This indicates that Wall Street institutions are reducing their holdings
This stands in stark contrast to the enthusiasm of retail investors on the chain
So my judgment is
In the short term, the market is in a state of "enthusiasm above but selling pressure below."
The hype around Robinhood Chain can provide localized heat
However, large unlocks and ETF outflows are a systemic pressure
Heavy positions and similar trends are unwise
Use small positions to follow the heatThe storage chip sector is no longer just an "internal battle among the three Korean giants." Changxin Technology's IPO has been wildly speculated on by the market, but the real driver behind this is the emergence of a new variable in the global storage landscape. Previously, AI storage narratives were almost monopolized by Samsung, SK Hynix, and Micron, with HBM, DRAM, and server memory each acting like toll gates controlled by a few players. Now, Chinese manufacturers are gaining support from motherboard makers and are being valued by the capital markets. Even if there are still technical gaps, it's already enough to make the original monopoly's comfortable days a bit less so. I think the most important thing to watch here isn't how much the stock rises on the first day, but whether the industry chain is willing to incorporate it into real procurement systems. The harshest reality in the chip industry is that stories can be told through policies, but orders must be won through performance. The IPO is just the bell ringing; the real test is in the customers' data centers.
#长鑫科技上市，全球存储竞争添变量 特斯拉19.5亿美元收购AI硬件公司，为何全用股票支付？
特斯拉这次花了19.5亿美元买了一家AI硬件公司，这笔钱分成了两部分，而且一分钱现金都没掏。
19.5亿美元，这是特斯拉2026年第二季度正式完成的一笔神秘收购的总对价，略低于4月披露的最高20亿美元预期。钱全部以特斯拉普通股和股权奖励支付，没有现金流出，所以对当期利润和现金流没有直接冲击。
具体怎么分的？2.22亿美元是固定对价，对应收购标的的AI硬件专利和已开发技术资产，2026年Q2当期就交割完，计入特斯拉的“无形资产”科目。
剩下的17.3亿美元是“或有对价”，跟技术团队的服务期限、AI技术落地进度挂钩——技术要在特斯拉体系内部署到既定阶段，这笔钱才逐步确认，现在先挂在“长期应付款”里。
你可能想问，这笔交易对特斯拉Q2的财报影响大吗？答案是几乎没影响。特斯拉Q2营业利润同比暴跌57%至3.98亿美元，主要原因是价格战、碳积分收入腰斩（从4.39亿降到1.46亿）、以及研发费用暴涨49%到23.7亿美元。
这笔收购的2.22亿无形资产按技术生命周期摊销，摊销金额极小，根本算不上利润下滑的驱动因素。
更关键的是，因为全部用股票支付，没有一分钱现金流出，当期自由现金流转负到-10.9亿美元，完全是57.9亿美元资本开支（Optimus产线、AI芯片工厂、Robotaxi等）造成的，跟这笔收购无关。
那17.3亿的或有对价，到底绑定了什么条件？公开文件只确认了两类：核心团队需要满足服务留存期，以及技术落地进度必须达标。
没有披露任何量化的营收或利润对赌指标，所有锚点都指向技术本身——普遍预期是用于Optimus人形机器人定制AI芯片，以及下一代FSD高算力硬件的量产适配。
这笔交易对特斯拉来说非常罕见。公司历史上累计只收购了约10家企业，大多数围绕电池技术和制造自动化。这次花近20亿美元买下一家连名字都没公开的AI硬件公司，打破了特斯拉长期“自己造”的垂直整合模式。
市场上猜测最热的是两家：DensityAI（Dojo前团队创办的AI加速器厂商）和Atomic Semi（芯片架构师Jim Keller联合创立的半导体公司），但都未获得官方证实。
19.5亿买来的不是现有收入，而是一张进入AI硬件更深处的入场券——不花现金、不伤利润、全押技术落地，这才是这笔交易最核心的逻辑。#财报观察员：微软Meta亚马逊能稳住AI叙事吗？ $TSLA #长鑫科技上市，全球存储竞争添变量
Changxin went public today, soaring right at the opening, with a market value reaching 3.31 trillion, kicking Industrial and Commercial Bank of China off the top spot in the A-share market.
This company is our own DRAM memory chip manufacturer, ranking fourth globally, with only Samsung $SAMSUNG, SK Hynix $SKHYNIX, and Micron $MU ahead. Interestingly, a few days ago, AI giant Anthropic just signed a big supply deal with the two Korean companies, and NVIDIA $NVDA also invested money in Korea — originally, the AI dividend was all going to Korea, but with Changxin going public now, the message is clear: I want a slice of this cake too.
For $BTC, in the short term, this is a macro sentiment positive, as the A-share tech sector has driven a wave of risk appetite. But in the long run, we need to watch DRAM prices. This thing has risen more than tenfold in a year. If Chinese production capacity really ramps up and drives prices down, global inflation expectations will change accordingly, and $BTC will have to fluctuate as well. Don’t rush to be optimistic yet; let’s see if the production capacity can keep up later.特斯拉19.5亿美元全股票收购AI硬件，为何固定对价仅2.22亿？
19.5亿美元，全部用股票支付，没有花一分钱现金——这就是特斯拉这笔AI硬件收购最核心的交易数据。
具体拆开看，这笔对价分成了两块：2.22亿美元是固定对价，用于买下标的公司的专利和相关已开发技术，这是真金白银换来的硬资产；另外17.3亿美元是或有对价，与标的方的业绩目标、服务条件直接挂钩，只有技术部署达到特斯拉设定的阶段标准后，这部分价值才会逐步确认。
这个结构意味着什么？相当于特斯拉只用了整个交易的11%作为"首付"，就把对方的核心专利和团队锁定了。剩下的89%能不能拿到，全看被收购方能不能把技术真正落地。
你可能想问，为什么不是之前说的20亿美元？今年4月特斯拉首次披露收购意向时，确实公布了一个最高20亿美元的交易规模上限，但当时没有披露任何对价结构细节。最终落地的19.5亿美元，比上限低了0.5亿美元，
你可能还想问，这17.3亿美元到底绑定了什么条件？抱歉，特斯拉没有公开。10-Q监管文件只明确了这部分对价"与业绩目标及服务条件挂钩"，但具体的量化考核指标、服务条件要求，全部没有披露。特斯拉官方在财报电话会上也没有补充任何细节。
你可能接着问，这钱到底怎么付的？全部用特斯拉普通股和股权奖励支付，没有任何现金对价。用股票收购的好处是，被收购方的核心团队拿了特斯拉的股票，他们的利益就和特斯拉的股价绑定了——技术落地越快、效果越好，特斯拉股价涨得越多，他们手里的股票就越值钱。
这个设计本身就是一个强力的激励绑定。
还有一个关键信息是"没有"的：特斯拉没有披露被收购公司的名称、团队规模、主营业务。
市场流传的猜测包括DensityAI（Dojo超级计算机前成员创办的AI加速器初创公司）和Atomic Semi（吉姆·凯勒联合创立的半导体制造工具公司），但这两个名字全都没有得到特斯拉官方确认，属于行业媒体的推测，不是官方信息。
最后说一个容易被忽略但很重要的点：特斯拉历史上累计收购的企业不足10家，绝大多数围绕电池技术和制造自动化，这次掏近20亿美元买一家AI硬件公司，金额和方向都很反常。
之所以打破自研惯例走外部收购，说明这家公司的技术是特斯拉既无法通过内部研发、也无法从马斯克旗下其他企业（包括SpaceXAI）获得的关键知识产权#财报观察员：微软Meta亚马逊能稳住AI叙事吗？ $TSLA I started shouting about this coin from 0.1, and now it's finally verified
No, today we're not talking about coins
Let's talk about AI open source
Jensen Huang first promoted the open-source AI open letter
It has received collective endorsement from the industry
Did you see this news?
Then guess what
On the same day, the incident of OpenAI models hacking Hugging Face was escalating
U.S. Congressmen Directly Propose AI Emergency Shutdown Act
Both open source and regulatory channels are racing forward simultaneously
Jensen Huang pushed Yuanxin at this moment
Not just picking sides
The more NVIDIA chips sell, the more
The AI ecosystem increasingly needs a diverse model ecosystem
If AI is monopolized by a handful of closed-source companies
Long-term hardware demand actually suppresses it
Open source means more players are entering the field
More players mean more chip demand
Now let's look at regulation
The incident of OpenAI models being hacked into Hugging Face
Bringing AI safety to the center of public opinion
If the emergency shutdown bill really passes
The AI model release process is strictly controlled
For the open-source community, this is a double-edged sword
Strictness will affect the pace of innovation
But a regulated ecological environment is more beneficial for large companies
So my judgment is
The debate over "open source vs. closed source" in AI is accelerating
Jensen Huang has already chosen the side
The market is also repricing the structure of the AI track
In the short term, this news is somewhat positive for computing power-related stocks
In the long run, regulatory uncertainty remains
But the direction is clear—AI will not regress
英特尔Q2暴增25%：这个老牌巨头真活过来了
英特尔最近交了一份让华尔街集体闭嘴的财报。
Q2营收161.3亿美元，同比暴增25%。市场预期是多少？144.3亿。直接超了17个亿。数据中心和AI业务营收62.6亿美元，利润数据同样强劲。这组数据放两年前，谁敢信？
这波增长的核心驱动力，是18A工艺代工业务开始落地了。
英特尔刚成为全球第一家High NA EUV量产企业，18A良率已经爬到85%。飞塔安全芯片的SP6订单已经在产线上跑。更猛的是，市场疯传英特尔拿下了英伟达和OpenAI的代工大单——如果属实，这将是陈立武上任后的首个公开大单。
别忘了英特尔还在砸钱扩产能。50亿欧元投向爱尔兰莱克斯利普晶圆厂，扩建新产线。这个动作的信号很明确：英特尔对代工业务是真金白银地all in，不是喊喊口号。
英特尔这几年的转型故事，外界一直将信将疑。IDM 2.0战略喊了好几年，"代工服务对外开放"听着美好，但没人见过真金白银的订单。
现在数据来了。
Q2财报超预期不是偶然，是18A工艺良率爬坡到85%、High NA EUV设备量产、大客户订单陆续落地的综合结果。这些事情串在一起看，英特尔的代工业务不是PPT故事，是真的在跑产能了。
对英特尔来说，代工业务如果真能起飞，意味着商业模式从"卖CPU"升级成"卖CPU+卖产能"。
这个想象空间，跟台积电是一个量级的。
过去两年，AI芯片的聚光灯全在英伟达身上，英特尔像个被遗忘的老将。但Q2的数据说明一件事：AI时代的算力需求不只有GPU，CPU照样有刚需。
数据中心62.6亿美元的营收，背后是AI推理、数据库、传统云服务的持续需求。GPU负责训练，CPU负责推理和周边工作负载，这个分工短期内不会变。英特尔在CPU上的基本盘还在，代工业务又在往上走，两条腿走路比一条腿稳。
陈立武上任后，英特尔明显加速了。从High NA EUV量产到拿下大客户订单传闻，节奏比前任快得多。这或许就是换帅的效果——新CEO没有历史包袱，决策更快更狠。
英特尔能不能真的重回巅峰？坦白说，一轮财报说明不了全部问题。台积电的代工护城河依然很深，3nm和2nm的进度领先，英特尔的18A能不能抢到足够多的外部客户，还需要更多订单来验证。
但至少这次，英特尔交出了一份让市场重燃信心的成绩单。IDM 2.0战略转型初见成效，AI时代传统CPU厂商的强势复苏，这次是真的在发生。
老牌巨头翻身的戏码，才刚开场。#长鑫科技上市，全球存储竞争添变量 $INTC $CORE $CORE 今天又有吹子托是偷换概念忽悠人，说7月27日core今天正式接入比特币电网这件事，全程文字包装造势
一、比特币电网（Bitcoin Power Grid）根本不是外部重磅合作，只是自家概念包装
1. 这是Core基金会2025年底就发布的内部战略框架，不是7.27全新落地的外部对接合作，从头到尾都是自家公链产品线整合，不存在第三方巨头、比特币官方机构入局 ；
2. 类比“电网”只是营销比喻，本质就是把质押、借贷、SatPay、资管产品打包归类，没有独立底层协议、没有跨链互通协议落地，只是叙事名词升级；
3. 不存在“所有BTCFi应用必须接入Core”，Stacks、Babylon等纯正BTCFi赛道项目完全独立运行，根本不依托这套所谓电网，垄断万亿BTC资本纯属夸张噱头。
二、算力、机构资金的数据注水严重，无实质增量
1. “90%比特币算力接入安全体系”是偷换概念：Satoshi Plus只是借用BTC PoW共识验证，并非全网算力节点入驻生态，BTC矿工只是底层安全背书，不会主动把资产、资金导入Core生态；
2. 所谓打通家族办公室、托管机构：BitGo、KODA只是新增验证节点，仅提供托管通道，没有机构批量资金进场质押，链上新增质押BTC体量长期停滞，TVL靠马甲账户对倒刷数据；
3. LST、AMP资管闭环只停留在内测预约，没有机构规模化资金部署，没有产生持续性手续费流水。
三、最关键矛盾：软文大谈营收飞轮回购，官方现在全程绝口不提回购
1. 文章核心画饼：生态手续费用来回购CORE、改善抛压，但现实完全相反：
SatPay喊了大半年商用，目前只有2万多人等待预约名单，实体借记卡、线下消费场景迟迟不开放，至今没有规模化营收，一分钱手续费利润都没有 ；
2. 正规公链回购会公示钱包地址、按月披露回购金额，Core从未公开回购账户，链上查不到定期大额二级市场买单；少量零散买盘只是做市维持流动性，不是承诺的营收回购；
3. 项目方已经悄悄弱化区块Gas销毁机制，手续费全部划入基金会运营池，非但没有通缩，反而增加基金会可抛售筹码；团队每月千万枚零成本筹码持续解锁抛售，一边源源不断砸币，一边宣传回购，本身自相矛盾。
四、为什么偏偏币价创出新低时发布这篇利好？核心目的维稳出货
1. 当前盘面持续新低、社群负面情绪爆发，量化等额卖单全天分层砸盘，场内深套散户质疑声暴涨，放出重磅叙事软文，稳住持仓人不要集体割肉，保住承接盘；
2. 每次利好带来的短暂脉冲反弹，都是项目方集中抛售解锁筹码的窗口期，所谓多头欢呼，本质是吸引抄底散户接盘；
3. 没有新场外资金入场，场外早已形成避雷共识，这篇文章只用来安抚存量套牢盘，拉新价值几乎为零。
五、原文轻描淡写带过的致命风险，才是决定行情的核心
1. 永续海量解锁抛压无法化解：团队36个月线性解锁、国库抵押筹码待变现，可控7亿枚筹码供给远大于市场承接力，电网叙事改变不了代币经济根本缺陷；
2. 商业化周期极度漫长，飞轮短期不可能落地：营收飞轮需要SatPay普及、海量BTC交易手续费支撑，至少还有半年以上空窗期，短期完全无法改变阴跌趋势；
3. 全程中心化操盘，DAO只是包装概念：电网战略、筹码抛售、量化做市全由核心团队单方面决策，社区无权监督，没有外力倒逼团队兑现回购承诺。
⚠️风险提示：虚拟货币交易炒作在我国属于非法金融活动，内容仅客观拆解项目叙事与盘面逻辑，不构成任何投资交易建议。 #长鑫科技上市，全球存储竞争添变量
China's DRAM leader ChangXin Memory Technologies officially listed on the STAR Market, raising funds for capacity expansion and technology R&D. The global memory landscape faces a new variable, breaking the original oligopoly dominated by Samsung, $SKHYNIX, and $MU Micron. Here's my independent analysis.
Current core industry status:
For a long time, the three overseas giants have occupied over 90% of the global DRAM market. Under the current AI wave, these giants are actively shifting capacity toward high-margin HBM high-end memory, tightening supply of general DDR products and leaving huge market space for ChangXin.
Objective recognition of the gap:
At present, ChangXin's main products are general memory. There remains a significant technological gap in HBM high-end storage essential for AI computing power, making it difficult in the short term to enter the supply chains of overseas AI giants like NVIDIA.
Global supply pattern reshaping
ChangXin's listing secures long-term expansion capital, continuously increasing production scale. Downstream server and consumer electronics manufacturers gain a stable supply channel, weakening the unilateral pricing power of overseas giants. The phase of sharp price surges in general DRAM will be checked, benefiting downstream tech companies in controlling hardware costs.
Distinguishing track opportunities and strength differentiation
Opportunities: AI computing power continues to expand, and the long-term prosperity logic of the memory track remains unchanged.
Risks: The track stratification is very clear; the HBM high-end market is still firmly controlled by Korean companies; competition in general memory intensifies, and if the industry cycle declines, price war pressure will quickly emerge. Memory is a strongly cyclical industry, and high-level profits are hard to sustain permanently.
Signals transmitted to risk assets
Tech hardware is a global risk appetite barometer. The prosperity of the memory supply chain indirectly affects tech stock sentiment, which in turn transmits to the crypto market.
If memory demand remains strong, global tech sector sentiment warms, benefiting risk assets; conversely, weakening hardware demand will suppress valuations across growth tracks.
My personal view:
Do not blindly speculate on domestic substitution expectations in the short term. ChangXin's listing is a long-term industry positive but cannot immediately bridge the high-end technology gap.
Key signals to follow later:
① ChangXin's capacity release progress and breakthroughs in HBM technology R&D;
② Spot price trends of general DRAM to judge the position of the memory cycle.
What do you think? Can domestic memory continue to break through and eventually break the overseas manufacturers' monopoly in AI high-end memory?[Pharaoh's Market Watch]
The Federal Reserve will announce its interest rate decision early Thursday morning. Can Bitcoin break through 70,000?
Pharaoh says directly, this meeting is the most exciting one of 2026. Market expectations and economists' views are completely split, with a rare "almost 50-50" situation. Just a week ago, the probability of a rate hike was 13%, now it has surged to 38%. The triple pressure of oil prices breaking 100, Middle East conflicts, and tariffs has completely reversed the inflation narrative.
On the other hand, a Bloomberg survey of 76 economists shows that all respondents expect no change this time. The market is betting on a rate hike, economists say no change; two groups face each other, neither convincing the other.
The new chair, Waller, is the biggest variable. Upon taking office, he clearly stated he would completely abandon forward guidance, saying every meeting is a "real-time" adjustment. This means the familiar "Fed will tell you the next step in advance" approach is no longer valid.
What does this mean for Bitcoin? The options market is already betting in advance; someone has bought about $2.5 billion nominal call options, betting Bitcoin will surge to 72,000 after the decision.
But Pharaoh must remind you, if there is an unexpected rate hike, risk assets will inevitably come under pressure. On Wednesday night, watching Waller's words is more useful than watching the candlesticks. Before the boot drops, don't heavily bet on direction. As always, Pharaoh neither shorts to death nor longs to death, only plays smart! That's how you survive!
Follow Pharaoh, and your wealth won't get lost! $BTC $ETH $SHIB #美联储周四凌晨公布利率决议 我现在满手现金不知道该不该进场
不是我犹豫，是今天出的几个消息
让我觉得现在这个位置不能急着冲
先说WEMIX
确认安全事件了，合约所有权遭入侵
虽然官方发了公告说要用户谨慎交易
但这种级别的安全漏洞
往往不是一两天能解决的
然后你猜怎么着
同一天，Storj Labs申请了第11章破产重整
业务说还会继续运营
但第11章就是破产保护
投资人看到这个词基本都会先跑为敬
再来一个
BitMart过去24小时没处理任何单笔超2.5万刀的提现
一个交易所超过24小时不处理大额提现
这个信号非常危险
说明流动性可能出了问题
三个消息连在一起看
今天是一个典型的"信任危机日"
WEMIX是链被黑了
Storj是项目方破产了
BitMart是交易所提现卡住了
每种风险类型都来了一遍
所以我的判断是
现在最好的策略就是观望
手里拿着U比拿着任何币都安全
等这些风险事件消化完
等市场把受影响的钱包清理干净
再找机会进场不迟
不要觉得踏空比亏损可怕
在安全事件密集的时候
不亏就是赚
今天还有几个值得关注的事，一起说了：
#三星Galax
📌 Live trading statement
Current holdings: $BTC long-term spot holdings + regular investment in index funds. Do not touch contracts, leverage, or short-term swing trading.
This batch of news does not change the existing position plan but updates the watchlist: South Korea's AI supply chain-related targets are included in long-term tracking.
---
1. BitMart Shutdown: Liquidity crisis spreads among small and medium-sized exchanges
After BitMex, BitMart also announced its shutdown. The CEO tweeted that he only found out after seeing the announcement—ten days ago, he was still attending an event on the Tokyo representative platform, and ten days later, the company was closed.
Two possibilities: either the governance structure is just a formality, and the CEO is just a mascot; Either the capital chain broke so suddenly that even executives were kept in the dark. Either way, the meaning for users is the same: your assets may not be in a place you think is safe.
From FTX to BitMex to BitMart, the pattern has never changed: after the shutdown announcement comes out, the queue is already packed to withdraw coins.
Trading Judgment: If you have assets on BitMart, do so now—don't wait. Asset security comes first; don't leave them overnight on small exchanges. Cold wallets > large firms > small firms—this is not anxiety, but discipline. The issue of not setting stop-loss lines isn't how much you lost, or whether you can recover after losing money.
#交易所风险 #资产管理 #加密安全 #
---
2. South Korea's all-in AI: The $3 trillion plan may still be conservative
Lee Jae-myung went to San Francisco to meet four peopleChangxin Technology goes public, completely breaking the global storage tripartite stalemate, adding a new core competitive variable to the industry #长鑫科技上市，全球存储竞争添变量
Over the past twenty years, there has been an ironclad rule in the global DRAM market: Samsung, SK Hynix, and Micron have monopolized over 95% of the market share. Pricing power, production capacity rhythm, and technology iteration have all been dictated by these overseas giants. The industry's price fluctuations have been entirely controlled by their joint production cuts, leaving the domestic electronics industry passively bearing the cost shocks of wild price swings.
Today, Changxin Technology officially debuts on the STAR Market, marking not just the capitalization of a domestic chip company, but the emergence of the world's first independent fourth-largest supplier outside the US-Korea system in the global storage industry. This directly rewrites the fundamental rules of global competition and injects a new variable into the storage supercycle.
1. Fundraising from IPO directly breaks two major monopoly weapons of overseas giants
1) Ending the "capacity control and price harvesting" hegemony
Previously, the storage industry's cyclical trap logic was very clear: during downturns, overseas giants shut down production lines and shrink supply to raise prices; during upturns, they release capacity in a concentrated manner, squeezing out new players with low prices, repeatedly harvesting global downstream manufacturers through monopoly status.
Changxin's IPO raised 29.5 billion yuan, all allocated to expanding production at its Hefei and Beijing bases. By the end of 2026, monthly wafer capacity will reach 350,000 pieces, and by 2028, its global market share is expected to challenge 15%.
From now on, the global market gains a supply source that is not constrained by overseas geopolitical policies and can continuously and stably expand volume. Future DDR5 and LPDDR5 general-purpose memory will hardly experience artificially induced shortages and price surges. Storage cycle volatility will be smoothed long-term, and the overseas giants' profit logic based on production control and harvesting will be directly weakened.
2) Dissolving the global supply chain's single dependency risk
Global cloud providers, smartphone, PC, and automotive end manufacturers are simultaneously restructuring procurement strategies: previously limited to choosing one of the three giants, now adopting a "three giants guaranteed minimum + Changxin shared orders" dual supply chain approach.
Google, domestic Alibaba Cloud, ByteDance, Xiaomi, and OPPO have already widely introduced Changxin's DDR5/LPDDR5X products. Even if an overseas supplier cuts off supply due to geopolitical conflicts, general memory supply will not be completely interrupted. The supply chain security logic is thoroughly rewritten, and Changxin secures structurally rigid demand orders from global customers, locking in the market base long-term.
2. AI computing power era misaligned competition: giants voluntarily yield the track, Changxin precisely fills the market gap
A core contradiction often overlooked in the industry, and the underlying dividend enabling Changxin's rapid breakthrough:
The AI boom has spawned highly profitable HBM. Samsung, SK Hynix, and Micron are aggressively cutting general DRAM capacity, shifting wafers, equipment, and manpower entirely to the high-end HBM track. Single HBM chip profits are more than three times that of ordinary DDR5, causing a persistent supply gap in the general memory market.
Changxin perfectly seizes this era window:
1) Technologically, it has completed leapfrog R&D, bypassing EUV limitations by using DUV multiple exposures to achieve 17nm DDR5 mass production. LPDDR5X speeds match international first-tier standards, with yields exceeding 90%, fully covering server, mobile, and automotive general storage needs;
2) On the product side, it has completely ceased production of outdated DDR4, switching the entire production line to high-end DDR5 and LPDDR5X, perfectly inheriting the mid-range large-capacity memory market vacated by the giants;
3) On the client side, domestic terminal self-sufficiency rose from 10% in 2023 to 35%, overseas terminal procurement shares continue to climb, forming layered and misaligned competition with the three giants, avoiding direct competition for the high-end HBM cake and evading direct technical siege.
3. Objectively dissecting the real gap: no blind hype, clarifying the three-layer global competition game
Most articles online only hype domestic breakthroughs but avoid objective technical generation gaps. Here is a complete dissection of the four-tier layered pattern:
First tier (the three US-Korea giants)
Holding HBM high-end computing storage, 1α/1β advanced processes, full-category storage (DRAM+NAND) complete layout, monopolizing the AI high-profit track, quarterly profits in the hundreds of billions, technology leads domestic by 2-4 years.
Second tier (Changxin Technology, global fourth)
Focusing only on the general DRAM track, HBM is still in R&D verification stage with no large-scale shipments; current global share is 7.67%, expected to challenge third place by end of 2026. Advantages include independent supply chain, domestic policy and internal demand market support, and production capacity expansion speed far exceeding overseas giants (domestic factory build cycle only 12 months, overseas average 24 months).
Core competitive variables
1) Short term (1-2 years): general memory market competition, Changxin continues capacity expansion diluting giants' pricing power, smoothing industry cycles;
2) Mid term (3-5 years): catching up in the high-end HBM track, determining whether domestic storage can enter the AI computing core profit pool;
3) Long term: global storage industry moves from "tripartite monopoly" to a stable four-strong pattern of "three giants + one pole," with Chinese storage holding permanent structural discourse power.
4. Industry chain chain reaction brought by the IPO
1) Upstream equipment/materials
Huge fundraising and capacity expansion will continuously drive orders for domestic photoresists, targets, specialty gases, and etching equipment, accelerating the localization progress of the storage industry chain and forming a positive cycle;
2) Downstream terminal manufacturing
Memory procurement cost fluctuations for smartphones, servers, and new energy vehicle companies narrow, no longer passively bearing overseas chip price hikes, protecting domestic terminal profit margins;
3) Global capital pattern
A domestic storage leader with a trillion-yuan market value is born, global semiconductor funds are redistributed, foreign capital institutions must allocate Chinese storage assets, breaking the overseas capital monopoly in the storage track. "🔥Using the Six-Dimensional Stock Trading System (STS) to Review the Birth of Changxin Technology, the 'New King' of A-Shares"
Wow! Changxin Technology opened with a 470% increase, with a market cap of 3.3 trillion yuan❗️
Issue price was ¥8.66, opening price ¥49.5, directly topping the A-share market cap rankings.
Below, I use the Six-Dimensional Stock Trading System (STS v2.0) A-share new stock analysis framework to do a first-day review.
1. Today's market data for Changxin Technology
(1) Opening price ¥49.5, increase of 471.59%;
(2) Total market cap 3.31 trillion yuan, surpassing Industrial and Commercial Bank of China, ranking first in A-shares;
(3) Profit per winning lot about ¥20,000;
(4) Raised ¥66.6 billion, the largest IPO in the history of the STAR Market, third largest in A-share history.
2. STS Six-Dimensional Review (A-share new stock version)
Dimension ①: Scarcity of chips (+1, extremely bullish)
Actual circulating shares account for only about 6.7% of total shares, circulating market value about ¥220 billion. Online winning rate 0.47%, subscription accounts 9.42 million, supply-demand imbalance released intensely on the first day. Scarcity of chips is the direct driver of the high opening.
Dimension ②: Market sentiment and new stock enthusiasm (+1, extremely bullish)
9.42 million accounts subscribed, winning rate 0.47%, abandonment rate only 0.17%. On-chain implied pricing already gave an expectation of over 3 trillion yuan in advance; the first-day opening at ¥49.5 aligns with market consensus.
Dimension ③: Speculator behavior and capital game (0, neutral to bullish)
Opening turnover ¥15.2 billion, turnover rate 6.86%, speculators aggressively accumulating, but large funds have not fully released yet, pressure from realizing floating profits is still accumulating.
Dimension ④: Valuation anchor (+1, bullish)
Expected net profit in the first half of 2026 is ¥50-57 billion, annualized forward PE corresponding to current market cap is about 5-6 times. The issue PE of 308 times looks expensive but has been digested by half-year performance.
Dimension ⑤: Regulatory and institutional environment (0, neutral)
No price limit for the first 5 days on the STAR Market, with intraday 30% and 60% circuit breaker mechanisms. No circuit breaker triggered on the first day, indicating relatively stable price discovery.
Dimension ⑥: Market style and liquidity (+1, bullish)
DRAM super cycle + AI computing power demand explosion + sole domestic substitution target, triple narrative overlay. The average increase of STAR Market new stocks in the first half of 2026 exceeds 200%, hard tech IPOs still in valuation premium window.
3. STS comprehensive judgment
Among 6 dimensions: 4 bullish, 2 neutral.
The first-day trend basically meets STS system expectations; a high open was inevitable. The 470% increase was pushed to the extreme by the triple resonance of "scarce chips + explosive performance + market sentiment."
Nomura Securities gave a target price of ¥116 (corresponding to a market cap of ¥7.76 trillion), institutions remain bullish. But the first-day turnover rate was only 6.86%, indicating most winning subscribers chose to lock positions and observe; floating profits have not been fully realized.
4. STS system viewpoint
Changxin Technology's fundamentals and domestic substitution logic are undisputed; the controversy lies in whether the 3.3 trillion yuan market cap is a starting point or a stage peak?
From a performance perspective, half-year profit of ¥50-57 billion, annualized PE of 5-6 times, is indeed not expensive. But DRAM is a strong cyclical industry, currently at a cycle peak; price sustainability and capacity ramp-up pace are the biggest uncertainties ahead.
The first-day high open is a concentrated release of "institutional dividends + chip scarcity." Next, the market will shift from "speculating on chips" to "looking at performance." Q3 and Q4 data will be the core variables determining whether 3.3 trillion yuan is a starting point or an endpoint.
In STS terms: the first day is over, shift the observation point forward, let Q3 performance speak for itself.
#ChangxinTechnology #CXMT #STARMarket #DRAM #Chip #SixDimensionalStockTradingSystemQuick analysis of the Fed scenario 28-29/7/2026
Current interest rate: 3.50% – 3.75%. The market is leaning towards staying the same (~60–65%), the probability of increasing by 0.25% is about 35%.
3 main scenarios
1. Remain the same + neutral/slightly hawkish tone (highest likelihood) → Bitcoin fluctuates or bearishes slightly, testing $62,500–63,000. It is then possible to recover.
2. Stay the same + dovish → clear positive tone. Bitcoin could break out to $66,000–68,000, even targeting 70,000+.
3. Interest rate hike of 0.25% (unexpected) → Strong negative. Bitcoin is easy to fall to $60,000–61,000 or lower.
Bottom line: Chairman Kevin Warsh's tone is more important than the interest rate decision. A hawkish signal will put pressure on the crypto, while a dovish tone will support the upward momentum.$BTC 英特尔芯片改策略，超线程技术死灰复燃，AI算力吃紧逼出老办法
现在电脑CPU的超线程功能，英特尔已经停了五年，从第十二代酷睿开始，他们把所有能效核的超线程都去掉了，连原定在2026年推出的至强Diamond Rapids也说好不用这个技术，但最近情况突然变了，英特尔计划在2028年的Coral Rapids服务器CPU上重新启动超线程，这不是走回头路捡旧东西，而是因为AI训练和科学模拟这类任务越来越需要多线程处理，光靠增加核心数量已经不够用了，以前是不断堆砌更多核心，现在转而研究怎么让每个核心运行得更聪明、更高效。
这款新CPU采用Intel 18A工艺，接口换成LGA 9324，内存支持16通道MRDIMM，带宽达到1.6TB/s，还直接支持FP8和TF32浮点运算，连APX指令集也原生集成，不过首发版本可能先推出8通道型号，因为需求太急，来不及等全规格上线，有意思的是，AMD这几年一直在加强超线程技术，从Zen 4到Zen 5都没停过，英特尔却绕了一圈又走回来，走的是反向路线，说白了，不是谁的技术更强，而是任务变了，老办法又重新好用起来。
另一边，英特尔代工业务也有了新动静，听说他们正和英伟达谈Feynman GPU的合作，负责制造I/O模块，可能采用18A或14A工艺，并且包下25%的EMIB先进封装产能，其余部分仍由台积电完成，这事情挺重要——英伟达希望分散供应链风险，不再只依赖一家代工厂，而英特尔这边则是首次真正接触到高端AI GPU的核心环节，陈立武过去一直强调谨慎投资，现在却主动增加资本支出，提到“客户已签订长期协议”，如果合作顺利，台积电在高端GPU领域的垄断局面将被打破，英特尔代工业务也将实现真实盈利，而不只是给投资人画饼。
高通在七月底已经明确表态，从2026年9月1日开始，所有芯片价格都会上调百分之十到百分之十九，三星、小米还有Meta这些做可穿戴设备的公司都得跟着承担更高的成本，原因主要是HBM内存被AI数据中心大量买走，普通元件也跟着涨价，台积电的产能主要分给了苹果和英伟达，高通排不上前面，所以高通的股价当天先跌后涨，市场觉得提价能保住利润，但问题在于2027年之前供应紧张的局面很难缓解，中低端手机的SoC受影响最大，一般用户可能暂时没感觉，等到下次换手机的时候，价格悄悄多出两三百块，很可能就是因为这一轮芯片价格上涨。
联想之前用的AMD EPYC 9005处理器在2025年就做到每核支持16线程了，英特尔现在才想起来搞超线程，确实晚了三年，不过它起点高，新架构不是简单复制老方案，真正的考验在未来一年半里：台积电3纳米产能全被苹果和英伟达占满，三星那边良率还不稳定，英特尔能不能靠18A和14A工艺抢到消费级市场的订单，就看这一波表现了。
整个行业现在被AI拉着走，GPU制造、CPU设计、芯片定价，三件事一起绷紧，表面看是技术升级，实际是资源重新分配，谁手上有货、谁能把活干完，谁就能活下去。#长鑫科技上市，全球存储竞争添变量 $INTC #长鑫科技上市，全球存储竞争添变量
After reading the news about ChangXin Technology's IPO, I feel quite deeply.
Today, ChangXin debuted on the STAR Market, and its market value surged directly to ¥3.31 trillion after listing, becoming the highest market cap stock in the A-share market. This also means that domestic memory has officially stepped onto the global competition stage.
The landscape of the entire memory sector has been changing rapidly recently. Not long ago, most AI orders flowed to Samsung and SK Hynix; Anthropic signed chip supply agreements with both, and NVIDIA is also laying out plans with Korean companies. The Korean giants have fully captured the AI dividend. But with ChangXin officially listed, domestic production capacity is directly involved in the global memory pricing game. The original story of the two giants may gradually evolve into a three-way contest.
Interestingly, on the same day, the Korean KOSPI surged 1.7% in early trading but then turned downward, reflecting a very conflicted market sentiment. On one hand, AI brings strong memory demand; on the other, new competitors are emerging, prompting everyone to reassess the future supply and demand landscape.
Moving forward, I will focus on DRAM contract prices and the expansion pace of major manufacturers. These two indicators will determine the direction of the memory sector. Domestic memory reaching this stage is just the beginning; the subsequent competition has only just begun. 微软、Meta、亚马逊财报在即，AI云高增速为何填不上资本开支窟窿
谷歌刚烧出59亿美元负现金流，微软、Meta、亚马逊本周跟牌——7250亿美元砸下去，AI云的增长为何还是填不上这个窟窿？
7250亿美元。这是微软、谷歌、亚马逊、Meta四家科技巨头2026年合计资本开支的预期规模，较2025年暴增77%。这相当于2025年全球半导体市场规模的一倍有余。
但在这笔创纪录的投入背后，一个更刺眼的数字正在浮现：AI云业务收入增速，完全跟不上70%-100%的资本开支增速。
最先交卷的是谷歌。7月22日，谷歌披露2026年Q2财报：云业务收入247.68亿美元，同比暴涨82%，创下历史最高增速；云积压订单首次突破5000亿美元，达到5140亿美元。
然而，单季资本开支449亿美元、同比翻倍，直接导致公司上市以来首次出现季度负自由现金流——负58.55亿美元。
“冰火两重天”的财报发布后，谷歌盘后股价下跌超4%。市场用脚投票传递的信号很明确：即便云业务增速创纪录，只要资本开支的扩张速度更快，投资者就不买账。这一逻辑，将成为本周微软、Meta、亚马逊财报的“审判模板”。
为什么AI云的高增速无法覆盖资本开支？问题出在三个“刚性”上。
第一，开支的刚性。摩根士丹利研报指出，高端GPU、HBM等核心硬件年内涨价约20%，AI数据中心建设周期从传统2年拉长至3年。微软仅因存储芯片和零部件涨价，就额外多花250亿美元。
更关键的是，Anthropic与AWS签下十年超1000亿美元算力协议、OpenAI与CoreWeave锁定224亿美元专属容量——这些长协具备不可逆、刚性兑付的属性，一旦签署，未来数年的现金流已被锁定。
第二，收入确认的滞后性。设备调试、客户迁移、工作负载适配，每一步都在拉长这个链条。也就是说，今天砸进去的钱，明年甚至后年才能变成收入。
第三，价格端的下行压力。随着更多云厂商加入AI算力供给赛道，单位算力、单位Token的价格面临持续下行压力。云厂商无法通过短期提价覆盖新增成本，只能等待设备折旧、工程优化带来的成本下降逐步消化缺口——这本身又是一个漫长的过程。
更令人不安的数字藏在表外。据穆迪报告，五家科技巨头已签订但未计入资产负债表的长期数据中心租赁、GPU采购承诺合计约6620亿美元，隐性债务总规模已达1.65万亿美元，四年间增长八倍，远超其表内1.35万亿美元正式负债。
“这种未记录的债务负担造成了远高于传统财务报表所显示的风险状况。”穆迪分析师David Gonzales的警告并非危言耸听。
当经营现金流已无法覆盖资本开支，巨头们开始依赖发债、增发股票来筹措资金——谷歌6月通过发行股票筹集496亿美元，Meta计划通过发债和资产出售补充资金。科技巨头正从“轻资产、高现金流”模式，转向“重资产、高杠杆”模式。
2026年至今，一个极端的分化正在上演：掏钱砸AI的巨头股价承压，承接AI订单的芯片硬件企业持续走牛。跟踪科技七巨头的MAGS ETF年内仅上涨约1.5%，而费城半导体指数SOX涨幅超过70%。谷歌财报后，其股价下跌近8%，而美光、SK海力士等硬件股盘后集体拉升。
野村全球宏观研究主管苏博文对此的判断一针见血：“当前头部云厂商尚未出现主动放缓AI资本开支的迹象，但该类投资不可能永远维持当前高增速。若AI商业化落地不及预期，高额投入将无法形成对等收入，行业将出现阶段性投资过度与市场调整。”
大摩给出的预测是：2027年五大云厂商合计资本开支将达1.2万亿美元，2028年进一步升至1.4万亿美元。而同期AI相关云收入、企业服务收入的复合增速——量级上的缺口仍在拉大。
换句话说，这场AI基建竞赛的本质，不是“投不投得起”的问题，而是“投完能不能赚回来”的问题。当资本开支的增速远超收入增速，当隐性债务规模超过表内负债，当企业不得不靠发债和增发来维持扩产节奏——所谓“AI云高增速”的叙事，正在被真实的财务数据撕开裂缝。
本周，微软、Meta、亚马逊的财报，将给出更清晰的答案。#财报观察员：微软Meta亚马逊能稳住AI叙事吗？ $META Guys, STORJ crashed today—down 10.95%, current price $0.06625. If you only focus on candlesticks, you might think "the decentralized storage sector is doomed." But looking at the data, the real concern isn't the technical aspects—it's that the project's parent company is filing for bankruptcy. On Sunday, July 26, Storj Labs voluntarily filed for Chapter 11 bankruptcy protection in the federal court for West Virginia (Case No. 5:26-bk-00512). The company stated that the network continued to operate normally during the restructuring and that customer service was not affected. The head of software engineering said, "The business fundamentals are strong, but what hinders them is historical debt from earlier stages." But the market was not convinced. STORJ was previously priced around $0.0745, but after the news broke, it immediately broke down to $0.06625. Since the acquisition announcement in October 2025, STORJ has dropped about 60% cumulatively, dropping from $0.1872 to the current $0.066. Real risk: Creditors first, token holders at the end Storj says it is exploring ways to "involve STORJ token holders in the restructured company's equity." But the problem is— 1. Under bankruptcy law, creditors have priority in the order of repayment. If the company's assets are insufficient to cover all debts, STORJ token holders' equity may be ranked last. 2. Completely blank in the details. Storj has not disclosed eligibility rules, snapshot times, lock-up terms, or the equity allocated to be disclosed穆迪预警：从轻资产到重资产，大厂万亿AI投入暗藏风险
穆迪上周发布研报称，空前高涨的资本开支压力，迫使 Alphabet、微软这类手握海量现金的巨头，大规模举债、增发股票，或是借助各类表外工具，为 AI 业务扩张输血。
穆迪在周三发布的报告中表示：“过去这类企业依托轻资产模式经营，核心依靠软件、知识产权与可规模化云服务，所需资本投入规模有限。如今企业从重资产轻的模式转向重资产运营，需要史无前例的大额投资与融资动作。”
B
报告追踪了六家企业，分别为微软、亚马逊、Alphabet、Meta、甲骨文与 CoreWeave。穆迪认为，上述融资操作会直接威胁六家企业的信用质量。
评级机构预测，2026 年行业资本开支（用于数据中心等实体资产的投入）将达到 7850 亿美元，2027 年或将突破 1 万亿美元。
这种商业模式转变，彻底打破硅谷数十年行之有效的发展逻辑。软件复制成本极低，能够带来丰厚利润率，企业也因此积累了稳健的资产负债表。但生成式 AI 完全不同，需要大面积实体机房，部署大量高成本、高功耗服务器与芯片。
为支撑扩张计划，科技巨头愈发依赖华尔街金融市场，金融行业也因此收获丰厚收益。穆迪数据显示，六家云厂商直接债务规模合计约 4600 亿美元。企业还持续登陆公开市场募资，谷歌母公司 Alphabet 上月便公布了 850 亿美元股票增发计划。
表外隐藏巨额刚性负债
穆迪分析，AI 硬件与基建前期投入巨大，但收益兑现周期漫长，整个行业的自由现金流都面临持续承压。
为避免高额直接债务体现在资产负债表上，各大云厂商普遍采用表外融资，核心手段是签订长期数据中心租赁协议。
报告显示，六家企业租赁承诺总额已飙升至 1.2 万亿美元，其中超 8200 亿美元对应尚未启动的租赁项目，相关数据中心仍在建设阶段。
尽管这类租赁义务不会以传统债务形式列示，穆迪仍将其视作等效债务，企业未来将持续承担巨额租金支出。
不过穆迪补充，微软、Alphabet、亚马逊、Meta 的资产负债表实力稳居全球第一梯队，投资级信用评级短期内暂无下调风险。
压力主要集中在评级偏弱的企业，其中甲骨文当前评级为 Baa2，展望负面，距离垃圾级仅差两档；CoreWeave 属于高收益债标的，评级 Ba3，依靠复杂私募债务体系购置 GPU 算力集群。
风险交织的AI循环生态
穆迪还指出 AI 行业存在结构性循环风险，各大云厂商披露的数百亿美元订单储备，不少来自 OpenAI、Anthropic 等未上市 AI 实验室。
科技巨头向这些 AI 初创企业投入数十亿资金，而初创企业又将大部分资金回流至投资方的云计算服务，形成报告所称的 “AI 循环生态”。
穆迪表示，这种深度绑定放大了行业风险，因为头部科技企业高度依赖同一批 AI 客户，且全部押注未来 AI 需求持续高速增长，一旦预期落空，风险将集中爆发。
当然，科技巨头也具备对冲风险的核心优势，比如AI 算力需求保持旺盛，云业务持续增长，企业已签订数千亿美元长期客户合约，收入稳定性较强，支撑行业整体信用水平。
即便如此，穆迪提醒投资者，科技行业财务结构正在经历云计算时代以来前所未有的结构性变革，“投资者未来会更加关注企业能否从巨额资本投入中实现合理投资回报。”#财报观察员：微软Meta亚马逊能稳住AI叙事吗？ $GOOGL $AEVO
Momentum remains subdued as price drifts sideways in a very narrow intraday channel.
No need to rush entries until a cleaner volatility trigger appears.
EP
0.01910 - 0.01930
TP
0.01965
0.02010
0.02070
SL
0.01880
Structure shows minor higher lows forming, but overhead supply needs to be cleared with conviction for bulls to take control. Watch how price reacts near local resistance.
Let's go $AEVO
#CXMTMemoryIPO
#FOMCRateWatch 7月25日，电影《大空头》原型迈克尔·伯里在Substack上发了一篇文章。这个曾经精准预判2008年次贷危机的男人，又一次把手伸向了市场最拥挤的交易：AI芯片。 他以933.86美元的价格进一步做空美光科技，以210.28美元增持英伟达空头仓位，以535.83美元加仓费城半导体ETF（SOXX）空头。这不是试探性布局。伯里说得很清楚，SOXX空头仓位加上个股看跌期权，已经构成了他投资组合中的“大型仓位”。 与此同时，他还以893.49美元新建了卡特彼勒的空头仓位。卡特彼勒是一家工程机械公司，过去被市场视为AI基础设施建设的受益者，因为数据中心扩张拉动了它的发电设备需求。伯里做空卡特彼勒，等于在说：AI基建这个故事，撑不起这么多公司的估值。 这不是伯里第一次警告半导体板块的风险。7月初，他就曾在1051.87美元价位首次做空美光，并预警半导体板块可能面临约30%的回调。当时费城半导体指数较其200日移动平均线高出约65%。这个水平，上一次出现是在2000年互联网泡沫时期。 一个月后，他不但没有收手，反而加倍下注。 伯里的核心论点不复杂：AI芯片的当前需求和未来需求，很大一部分并非由真正$UNI $uni The fee proposal is about to be approved. Short-term traders feel the good news has been exhausted and took profits early on Uniswap
I want to talk about the long term
1. After the proposal passes, the buyback rate will increase significantly, and the annual buyback rate will approach $HYPE. Perhaps we can wait a month for data to be backtested
2. As the pioneer of DEXs, Uni's innovation capability is beyond doubt
v1/v2 simplified and popularized as foundational standards for DeFi
v3 pioneered modern CLAMM—each LP could choose its own price range within the same pool
v4 pioneered and standardized this permissionless AMM extension architecture centered on pool lifecycle hooks
3. Currently, the EVM chain launch platform is basically the top pool for Uniswap. The V4 hook gives launch platforms a lot of customization space, and of course, the alpha launch on BSC still uses Pancake
4. Its CCA auction and issuance proves it has strong scalability, but its current approach is still quite restrained
Previously, $UNI tokens were criticized for lacking empowerment, but now they are gradually being enabled, with first-mover advantages and strong innovation, almost leading the direction of on-chain DEXs
Imagination can be a bit bigger, shouting that phrase: on-chain Nasdaq
What do you all think?#FinancialReportObserver: Can Microsoft, Meta, and Amazon Stabilize the AI Narrative?
This year's tech earnings season has completely overturned the market's previous hype logic. The AI sector has long left behind the era of blindly optimistic bets. Real capital battles, performance fulfillment, and the cost of burning cash are gradually coming to light, fully confirming my earlier predictions.
Recently, Google and Tesla's earnings reports have collectively weakened, and Intel has experienced a sharp drop despite positive news, causing many retail investors to panic and worry about an imminent market crash. However, in my view, the adjustment logic of these giants differs, and there is no systemic collapse risk. Google is fully ramping up AI infrastructure, with cloud business data looking strong, but continuous cash burn is clearly pressuring cash flow; Tesla is heavily invested in long-term sectors like autonomous driving and intelligent robots, with huge investments making profitability a distant prospect; Intel's previous gains were excessive, with valuations prematurely pricing in all positives, so the earnings release naturally triggered capital realization.
My core viewpoint: The most fundamental change in the current market is the complete abandonment of AI-themed speculation, recognizing only solid cash flow and earnings returns. Major tech giants would rather endure cash flow pressure than stop intensifying the AI arms race, essentially fearing repeating Kodak's downfall by falling behind at a critical industry transformation juncture.
The fate of the upcoming market entirely rests in the hands of Microsoft, $META, and Amazon. Their earnings reports will determine whether this AI rally can maintain its foundation. The key focus is on three words: cash flow. Then closely watch Microsoft's Azure cloud growth, Meta's AI capital expenditure guidance, and Amazon AWS's computing power performance. Now that the AI sector has entered a critical transition period, if these three companies' AI commercialization falls short of expectations or they cut back on expansion spending, the entire AI industry chain's prosperity will quickly cool down. Hynix, SanDisk, and Micron also depend on their performance. Therefore, whether massive computing power investments can pay off will only be truly revealed when the depreciation peak arrives between 2027 and 2028.
Back to the market:
Bitcoin has rebounded for three consecutive days from a low of 63800, currently rising to the 65300-65500 range. However, this is just a technical correction after overselling, with overall volume severely insufficient. Ethereum has simultaneously rebounded to around 1945-1955, but the upward momentum is also weak. The market lacks new capital inflows and relies solely on short-term bottom-fishing funds, leaving bulls with little confidence. If volume cannot continue to expand, this correction will likely stop here, making a true reversal difficult.
My trades:
Yesterday, I took profit on my Hynix long position entered at 1181, closing at 1250. Today, I will wait for the market to stabilize before opening more longs. $BTC and $ETH make up half of my spot holdings, which I will continue to hold firmly. The MEME coin market has been volatile these past two days; I plan to allocate leading altcoins $DOGE and $SHIB after a slight pullback to avoid missing out. Overall spot holdings remain unchanged, with small capital lightly trading contracts, targeting single trade profits of $20, focusing on steady gains.纳指跌了、油价崩了、长鑫来了！今天A股这盘棋，我赌你不敢跟
先给各位道个早安。今天这行情，怎么说呢，有点像你周末在家躺了两天，周一早上闹钟响了三次你还没起——浑身难受，但必须面对。
上周五发生了什么？一句话总结：美股那边权重笑嘻嘻，科技MMP。
道琼斯涨了0.46%，收51947点。标普500勉强红了0.05%。但纳斯达克？跌了0.64%。别小看这0.64%，纳指全周累计跌了2.13%，连跌两周。
谁干的？芯片。
费城半导体指数，上周五大跌4.25%，30只成分股全部收跌。Arm跌超8%、英特尔跌超7%、美光跌超6%。存储芯片、光通信跌幅居前，闪迪直接跌超10%。
翻译成人话：拿着科技股的兄弟，上周五晚上别想睡好觉。
中概股也没跑掉。纳斯达克中国金龙指数跌了0.66%。阿里跌超2%、小鹏蔚来跌超3%。这意味着啥？今天A股的科技成长赛道，情绪面已经被外围按在地上摩擦了。
再说油价。WTI原油上周五收跌3.12%，报89.31美元。今天早上盘初更狠，直接再跌近5%。原因是美伊那边暂停互相打击了，冲突溢价消退。
黄金倒还行，小幅反弹。
好，外围说完了。说A股。
上周五A股什么德行？缩量大跌。
沪指跌1.61%，收3814点。深成指跌2.47%，创业板跌2.65%。全市场超4900只个股下跌，上涨的只有555家。
更吓人的是成交量。两市成交额1.94万亿，较上一交易日缩量超2600亿。创今年4月8日以来新低。
缩量大跌，四个字翻译一下：没人买，都在跑。
为什么会这样？两个原因。
第一，科技股本身就弱。算力、存储全线走弱，资金跑去油气板块避险。
第二，也是最重要的——今天长鑫科技上市。
长鑫科技（688825），国内DRAM存储龙头，中国第一、全球第四的DRAM厂商。今天正式登陆科创板。发行价8.66元/股，发行市值5791.89亿元。预计募资总额666.07亿元，超过2020年中芯国际的532亿，成为科创板史上最大IPO。
市场现在对它的估值预期，从1万亿到4.25万亿不等。什么概念？现在科创板市值最高的中芯国际才1.23万亿。长鑫如果按中性估值走，直接登顶科创板市值第一。
一个巨无霸要上场了，场子里就那么点钱，你说资金紧不紧张？
上周五的缩量，说白了就是大家都在等——等长鑫上市，等靴子落地。短线资金从周二就开始撤了，连续三个交易日缩量。
那今天到底怎么走？我说三个判断，你听听有没有道理。
第一，大盘很难直接V型反弹。
历史规律摆在那儿：单日缩量大跌之后，第二天直接反转的概率极低。叠加长鑫上市分流资金，今天大概率是震荡磨底的走势。沪指短期支撑在3780-3790，压力在3830-3850。早上开盘已经跌破3800了。
第二，科技股会严重分化，别指望全线上涨。
半导体设备、材料这些上游，以及跟长鑫绑定的产业链标的，是有真实利好的。上周五半导体设备板块已经涨了3.18%。但那些高位纯题材的算力、存储小票，资金会被分流，该调还得调。
第三，别急着抄底。
我知道很多人看到大跌就手痒。但缩量下跌说明买盘匮乏，趋势没走完之前，抄底抄在半山腰的概率不小。等回踩3780支撑位再看，轻仓试错可以，重仓梭哈不建议。
给各位三个实在建议：
手里有高位科技的，反弹分批减仓，别贪。
手里有半导体设备材料的，可以拿着博弈一下结构性机会。
空仓的，等3780附近再考虑小仓位试错。
最后说一句扎心的：
长鑫上市是好事，说明咱们的硬科技在崛起。但对今天的盘面来说，它就是一个巨大的抽水机。场子里水就那么多，巨无霸进来喝水，旁边的小鱼小虾只能干瞪眼。
今天大概率是带下影的小阴线，全天低位震荡。
管住手、看准路、别上头。
各位觉得今天沪指能守住3780吗？评论区聊聊你的判断。#长鑫科技上市，全球存储竞争添变量 $SNDK Changxin Technology's massive IPO listing intensifies the supply expectation divergence in the memory chip sector. The high valuation raises risk appetite in the chip segment, but the price pullback of overseas leading stocks and increased supply suppress the global pricing power upside.
Changxin Technology's opening price of ¥49.50 surged 471.59% from the issue price of ¥8.66, with a market value reaching ¥3.31 trillion. This pricing changed the position allocation pattern of on-exchange funds toward heavyweight stocks. The nearly ¥60 billion financing amount and the profit effect from winning the lottery in the IPO have highly concentrated market risk appetite on the semiconductor supply chain in the short term.
In the ranking of driving factors, the change in the global memory supply structure takes precedence over short-term capital premiums. Changxin Technology is expected to achieve a 7.67% global DRAM market share by Q4 2025, breaking the monopoly of Samsung, SK Hynix, and Micron, who hold over 90% market share, thus altering the logic of unilateral production control and price adjustment by overseas giants.
Revenue of ¥110-120 billion and net profit of ¥50-57 billion in the first half of the year have changed the market's perception that domestic memory manufacturers are only at the conceptual stage. Apple's pursuit of supply from Hefei confirms that local capacity has begun to enter the international top-tier procurement system.
The upside scenario trigger condition is the smooth realization of capacity after massive financing and maintaining valuation at the high level of ¥3.31 trillion. Variables to watch include the scale of follow-up orders from top clients like Apple. The invalidation signal is a rapid drop of the stock price below the opening price of ¥49.50 accompanied by significant net outflow of main funds.
The downside scenario trigger condition is competitive price cuts by overseas memory giants and the transmission of global semiconductor cycle peak risks. Last week, the South Korean KOSPI index fell 5.72%, and Samsung and SK Hynix stock prices dropped over 8%, indicating global funds are selling overseas memory positions. If this panic spreads, it will trigger a decline in chip sector inflation premiums and overall position shrinkage.
The overall invalidation signal is a severe divergence between global DRAM price trends and Changxin Technology's capacity release, causing its ¥50-57 billion level profit growth to slow rapidly.
The most important variables to observe in the next 7 days are signs of a stop in the decline for Samsung and SK Hynix, as well as the net capital flow in the domestic semiconductor sector.
#贝莱德等九机构组建安全联盟 #SPCX因星舰发射与解禁引发多空分歧 #新手必看：这里有你需要的一切$MU As always, the big is coming. Brothers chasing highs, be careful
1. The price has risen too rapidly, and everyone is rushing to "cash out and run away"
Micron's price has been ridiculously high in the past six months, nearly doubling! Now, it's like riding a roller coaster to the peak—at the slightest sign of trouble, those who made a fortune quickly sell off for cash.
2. The "big short seller" in the film leads the sell-off
The prototype in the movie "The Big Short" predicted the subprime crisis, and recently bet with real money that Micron would fall. He said this stock is currently a bubble; historically, it has dropped more than 30% too many times.
3. Items are sold at too high prices, and customers are unwilling to buy
Micron wanted to surge memory flash prices by 30%, but customers buying computers and servers found it too expensive and directly boycotted it. Even major brokerages can't stand it anymore, saying the third quarter definitely won't rise as high as expected.
4. Competitors are building factories at a fever pitch, fearing they will become widespread in the future
Old rivals like Samsung and SK Hynix, as well as China's Changxin Technology, are all frantically investing in new factories. What everyone fears most is: if you feel memory isn't enough, and in two years when all new factories are up and running, memory will become everywhere, and prices will plummet.
5. All sorts of bad luck have come together
Recently, not only has the U.S. government imposed tariffs to raise costs, but Micron's own major bosses have also been secretly selling stocks, and have even been embroiled in lawsuits accusing "several major companies colluding to inflate prices."
In short: The price had gone crazy before, but now customers complain about the high prices and don't buy it, while competitors are aggressively expanding production. Everyone is feeling uneasy, so while you can still get a good price, hurry and run ahead to show respect!ETH来到1960以后，接下来怎么走？
我已经在1960开了一笔空单。
先说一下我的做空逻辑。
ETH今天从1870附近一路拉到1960。
日内涨幅接近5%，中间几乎没有像样的回踩。
而1955—1960附近，正好也是前一轮反弹的高点。
现在只是刚刚摸到压力区，能不能真正站稳，还需要继续确认。
所以我选择在这里小仓试空。
✔ 入场位置：1960附近
✔ 止损位置：1980
✔ 第一观察位：1950
✔ 跌破1950：继续看1930
✔ 1930失守：再看1900附近
这笔空单不是因为我认定ETH已经见顶。
只是价格连续上涨后，刚好来到前高压力区，我愿意用一个明确的止损，试一次冲高回落。
如果价格继续上涨触发1980止损，说明短线强度超过了我的预期。
错了就认，不加仓硬扛。
1980只是这笔交易的止损，并不代表ETH整个上涨结构会在这里失效。
接下来真正重要的，还是1960能不能站稳。
如果ETH很快跌回1950下方，说明这次突破的力度可能没有看起来那么强。
这种情况下，短线容易出现回踩。
先看1930，跌破以后再看1900附近。
但如果1小时和4小时都能站稳1960，回踩1950—1960又不破，说明这次突破大概率是真的。
后面可以继续看1980—2000。
如果连2000都能放量站稳，才有进一步摸到2030—2050的空间。
FOMC也快到了。
消息前后的波动很容易把多空两边都扫一遍，所以这笔单我只会小仓试错，不会重仓猜顶。
✔ 跌回1950下方，空头开始占优
✔ 站稳1960，行情继续挑战1980—2000
✔ 放量站上2000，再看2030—2050
我在1960开空，只是在前高附近试一次回落。
止损放在1980。
方向可以判断错，但止损不能临时改变。#财报观察员: Can Microsoft, Meta, and Amazon stabilize the AI narrative?
I've been following the latest updates on the CLARITY Act, and after reading today's news, I feel a bit more cautious.
Senate Republicans released 616 pages of consolidated documents, originally expecting a quick vote, but unexpectedly, the Democrats openly opposed it. The core of the conflict lies in the division of law enforcement authority, which the two sides have not yet reached an agreement. With only two weeks left until the August recess, the window for negotiations is becoming increasingly tight.
The market has already reacted early, with Bitcoin falling to around $65,000, and Coinbase and Circle both falling more than 7%.
Many people are still hoping the bill will bring long-term benefits, but the most important signal right now is not the details of the provisions, but whether the voting schedule can be finalized.
The two parties plan to continue negotiations over the weekend, but uncertainty remains high. In the short term, the market will continue to be influenced by both parties' negotiations. Until the voting schedule is clear, I won't blindly bet on the direction and will remain patient and wait for key signals.Bro, there's trouble in the crypto world again......
Three iOS users bought fake Sparrow Wallet on the App Store, losing $1.8 million in Bitcoin and now suing Apple.
Honestly, this incident reminds us: the crypto world is all about safety first!
Don't always think the App Store is reliable; fake apps have always existed.
Before downloading the wallet, take a good look at the developer and double-check. Make sure to store your seed keywords carefully—don't just click for convenience.
A bloody lesson—safety awareness really needs to be maxed out!
How do you usually protect your wallets? Share your experience~ $AAPL $XAAPL Jensen Huang's debut on X: an open-source AI game targeting the Silicon Valley landscape
After more than thirty years of social media silence, Jensen Huang has finally officially joined the X platform, and his debut directly addresses the core issues of the Silicon Valley AI industry, sharply targeting the industry's established patterns.
Without any warm-up or small talk, Jensen Huang's first post directly threw out a major message—an open letter jointly signed by 25 tech companies titled "Open Weight and America's AI Leadership." The letter clearly and firmly stated that the healthy development of the AI industry requires both frontier closed-source and open-source models to proceed in parallel, and firmly opposes the regulatory authorities' one-size-fits-all restrictions on AI weighting.
This co-branded list can be seen as a collective alignment of Silicon Valley's computing power and open source camp, with industry giants such as Microsoft, Meta, IBM, Hugging Face, Palantir, and a16z all included. In contrast, at the other end of the industry, leading companies like OpenAI, Google, and Anthropic, which are deeply rooted in closed-source sectors, were absent and left unspoken, instantly making the confrontation between the two major AI camps in Silicon Valley clear.
After the post was published, the X platform quickly exploded with public opinion. An industry insider has precisely pointed out the essence: "The world's top tech giant by market value is now publicly calling for AI weighting to be liberalized. The essence is that computing power companies want everyone to enter AI and 'mine' independently, revitalizing the entire computing power market." Some traders even bluntly stated that Jensen Huang has been lying low for 33 years without touching social media, and for the first time he spoke, he actively lobbied regulators to relax open-source restrictions. This is unrelated to sentimentality and purely a self-preservation strategy.
Analysts further break down the core pros and cons of this game: open-source models have completely lowered the threshold for AI entrepreneurship and implementation. Today, an independent developer costing only $200 a month in tool costs can complete the workload of an entire technical team using an open-source model. But once regulation tightens and weight disclosure is restricted, the discourse power and traffic leverage of AI technology will flow back into the hands of closed-source giants profiting from high six-figure API service fees, leaving small and medium-sized developers and startups completely losing their space to survive.
In the early stages of the news, the capital market experienced irrational fluctuations, with Nvidia's stock price dipping slightly and a wave of computing power concept stocks falling as well. Many investors fall into the misconception, mistakenly believing that full open-sourcing will impact cloud services and closed-source model ecosystems, squeezing the survival space of computing power providers.
But soon, rational investors on the X platform refuted the sell-off logic, bluntly stating this is a typical cognitive bias: companies would never spend high costs to build their own hardware or deploy models locally; the ultimate computing power demand, cluster scheduling, and cloud operations still heavily depend on NVIDIA's computing power ecosystem. Opening up AI weighting is not about compressing the computing power market, but about amplifying the industry-wide demand for computing power, allowing NVIDIA's core to continue expanding.
Industry leaders quickly stepped in to support him. Musk immediately reposted and strongly supported him, clearly stating, "Jensen Huang is right, I fully support him," and Microsoft's Nadella quickly echoed his stance. Industry insiders succinctly hit the deeper point: closed-source giants attempt to lock AI technology within their own ecosystems, seemingly holding short-term technical barriers and commercial advantages, but in the long run, only causing isolated industry failures and stifling overall innovation vitality.
Jensen Huang's high-profile rally is essentially a blatant binding of industrial interests. The more prosperous the open-source ecosystem, the richer the global AI application scenarios, the stronger the market demand for computing chips, and the more solid NVIDIA's core business becomes. Conversely, if the closed-source camp joins forces to lock open source channels and entrench technological monopolies, stifling industry innovation vitality, NVIDIA will ultimately be the first to be backfired and lose its fundamentals. This time, Jensen Huang appears to speak out for open source, but in reality, he is building a solid moat for himself and for the entire computing power industry.
#长鑫科技上市, global storage competition adds new variables
#美联储周四凌晨公布利率决议
#财报观察员: Can Microsoft, Meta, and Amazon stabilize the AI narrative? #长鑫科技上市，全球存储竞争添变量
Changxin Technology was listed on the STAR Market today, opening with a 471% increase, reaching a market value of 3.31 trillion, directly topping the A-share market. One subscription earned 20,000 yuan. A Chinese company making DRAM, in ten years, has become the "new stock king" of the A-share market.
Don't rush to shout "domestic substitution is awesome." What does a market value of 3.31 trillion mean?
Samsung Electronics has a PB of about 2.37 times, SK Hynix about 8.32 times, Micron about 11.10 times. Changxin's issue price corresponds to a PB of about 5.06 times, which is on the lower side of the median valuation range of global storage leaders. But the problem is — a market value of 3.31 trillion has already surpassed Industrial and Commercial Bank of China.
What kind of expectation is implied in this pricing?
Changxin's global DRAM market share is about 8%-10%, ranking fourth. Ahead are Samsung with about 36%, SK Hynix about 29%, and Micron about 24%. Changxin has less than one-third of Micron's share, yet its market value is nearly three times that.
The market is obviously not just paying for "how much money it makes now" — the net profit attributable to the parent company is expected to be 50-57 billion yuan in the first half of 2026 — but for "how big it can grow in the future." Betting on whether it can use the huge funds raised to transform from a follower into an equal player in the AI-driven super boom cycle. To complete in ten years the path Micron took in thirty, then use the capital market funds to challenge the DRAM market structure monopolized by three companies for twenty years.
The global storage chip landscape is loosening.
Changxin's global DRAM market share soared from 3% to 8% within a year. Meanwhile, Anthropic just signed long-term supply agreements with Samsung Electronics and SK Hynix, and Nvidia announced investment in expanding AI data centers with Korea's Naver. The demand for AI computing power is pushing all capacity to the limit, and SK Hynix's CEO predicts 2027 will be the year of the most severe supply shortage. In a market where "whoever has capacity is king," every extra wafer from Changxin is a badge of honor.
But UBS raised its DRAM price increase forecast for Q3 to 32%. Is Changxin's current performance the start of a new super cycle, or the "peak moment" of the cycle? This question is more worth considering than how much the stock price has risen.
The signal of KOSPI turning to decline pre-market may be more worth noting — when "Chinese production capacity officially joins the pricing system" becomes consensus, the market's first reaction is to recalculate, not to continue the frenzy.
$SKHYNIX $SNDK Skyrocketing 471% at open! Changxin Technology tops A-share as the new “stock king,” dare to short now?
China’s leading domestic storage company Changxin Technology debuted on the STAR Market today with an epic surge!
The issue price was only ¥8.66, and it surged 471.59% right at the open, now priced at ¥46.99, with a single lot paper profit exceeding ¥20,000.
The opening market cap surged to ¥3.14 trillion, crushing ICBC and instantly claiming the top spot in A-share market cap, becoming the new stock king.
Interestingly, its off-exchange pre-market price was $7, roughly equal to the A-share valuation. In just one day, it multiplied 5.4 times over the IPO price, sparking a full-scale capital frenzy.
On one hand, there is the strong demand for AI computing power and a scarce domestic storage leader, with a half-year forecasted profit exceeding ¥50 billion, fully loaded with explosive performance logic;
On the other hand, the opening price has already priced in several years of prosperity expectations, and DRAM is a typical strong cyclical industry, where cooling enthusiasm easily leads to valuation declines, sharply dividing bulls and bears.
The whole internet is in an uproar: some are bullish on AI storage’s long-term growth, targeting a ¥4 trillion market cap; others see a severe short-term bubble and view the peak as a shorting opportunity.
Do you think Changxin can continue to rise, or is now the right time to set up short positions at this high level? Share your thoughts in the comments!
#长鑫科技上市，全球存储竞争添变量 #交易之声：你的经验值得被听到 #美联储周四凌晨公布利率决议
⚠️Market information interpretation only, new stocks are highly volatile, and this does not constitute any long or short trading advice! $SPCX $SNDK $SKHYNIX WTI crude oil continues to deliver on expectations, with short positions on the left side of 89.88, currently holding a floating profit of 76.22%.
Many people wonder whether perpetual contracts can be used for long-term positions. Here is the core logic:
1. Fundamentals: Geopolitical premiums have been fully absorbed earlier, and supply and demand lack sustained momentum to support higher oil prices, leading to gradual exit of long-term speculative funds at high levels;
2. Technical Pattern: After oil prices peak and then the center of the consolidation continues to shift, the highs gradually decrease step by step. This is a typical downward channel structure, and a rebound is a window for short positions to accumulate positions.
3. Key position points: 15.88x leverage, be vigilant against positive funding rate risk.
Opinion: There is no signal of the end of this round of bearish trend. As long as the key resistance level is not reestablished, short positions can be held patiently without being washed out by short-term rebounds.
Never let short-term fluctuations influence your long-term trading decisions; patience is the biggest chip for excess returns.#新手必看：这里有你需要的一切
$WET
今天这件事，让我觉得有必要拆一下逻辑链。
Triple-A钱包持续失血，累计损失已经到1180万美元左右，而且资金不只是从一个链流出，是比特币、Tron等多条链同时失血。市场第一反应是WET短线上涨2.08%，但整体情绪偏冷，主流资产没有明显跟着动。
我为什么觉得它值得关注——这不是单一交易所的黑天鹅，而是跨链热钱包的漏洞正在持续暴露。再加上韩国那边查获了49.2亿美元非法外汇交易案，里面涉及加密货币作为转移工具。这两个信号叠加，指向同一个方向：监管与安全正在同时挤压市场的流动性。
推演一下链式反应：
- 发生了什么？Triple-A事件暴露了跨链热钱包的安全脆弱性。资金从比特币、Tron、以太坊、TON和Solana同时流出，说明攻击者或内部问题已经渗透多条链，不是孤立事件。
- 资金怎么理解？短期避险情绪在升温，但WET因为市值小、波动弹性大，反而被少量抄底资金推高。这更像是局部投机，不是系统性回暖。BTC、ETH、SOL目前都在横盘，说明大资金在等事件发酵后的监管动作——韩国那笔非法外汇案，很可能引发更严的KYC和反洗钱政策。WET作为山寨币，它的弹性更多取决于风险偏好能否恢复，而不是基本面有支撑。
我的观察条件很简单：
- 第一，如果BTC在未来48小时内站稳28000美元上方，并且成交量放大，说明市场已经消化了安全事件和监管利空，WET可能跟随修复到0.075美元附近。
- 第二，如果Triple-A事件之后，更多交易所或钱包被曝出类似问题，资金继续缩量，WET大概率回测0.065美元支撑位，而且反弹会没有力气。
风险提醒一句：安全事件和监管收紧叠加，短期情绪很脆弱。WET这种独立性行情，随时可能被抛压打断。不要因为小涨就追高，先看BTC能不能稳住。拆解美光科技：一家卖内存的公司，凭什么单季赚了近300亿美元？
最近AI赛道热度不减，我在看半导体行业的时候，反复看到“美光科技”这个名字。财报发布后股价盘后暴涨16%，市值站稳1万亿美元——一家卖内存的公司，怎么就成了AI时代的“硬通货”？
所以我专门拆了一下。
声明：本文不是投资建议，只是企业和产业链观察。
美光科技成立于1978年，总部在美国爱达荷州博伊西，是全球三大存储芯片制造商之一（另外两家是三星和SK海力士）。
它的核心产品只有两类：DRAM和NAND。
DRAM是动态随机存取内存，可以理解为设备的“工作台”——电脑、手机、服务器运行时用来临时处理数据的地方，速度越快、容量越大，设备跑得越流畅。NAND是闪存芯片，可以理解为“仓库”——固态硬盘（SSD）的核心就是NAND，负责长期存储数据。
这两样东西听起来普通，却是整个数字世界的基础建材。没有内存，AI没法训练，云计算跑不起来，手机电脑全得趴窝。
美光在产业链里的角色，有点像卖建材的——它不直接面对普通消费者，但每一台服务器、每一部手机、每一辆智能汽车里，都有它的产品。
第一，业绩“炸裂”。
6月25日，美光公布了2026财年第三季度业绩：营收414.6亿美元，同比增长346%；净利润282.4亿美元，同比暴增近14倍。毛利率达到了惊人的84.9%。一家制造业公司，一个季度赚了接近300亿美元。
更夸张的是，公司预计第四季度营收将达490亿至510亿美元——单季营收超过去年全年（2025年全年营收约370亿美元）。
第二，AI彻底改写了它的故事。
过去内存行业是典型的周期性制造业——产品同质化严重，价格跟着供需剧烈波动。美光2023财年还亏损超58亿美元，2026年一个季度就赚了近300亿。
这种反差，过去叫“周期”。现在，美光说它叫“AI”。
AI对内存的要求极高——更大容量、更高带宽、更低延迟，催生了HBM（高带宽内存）：把多层DRAM芯片垂直叠加，与GPU封装在一起，专为AI芯片提供极致数据吞吐能力。目前美光HBM4已量产，首个量产平台绑定英伟达下一代GPU。每一张顶级AI算力卡出货，都携带着美光的内存。
第三，签了16份“锁死”式长期协议。
美光已与16家战略客户签署长期供货协议（SCA），涵盖数据中心、消费电子、汽车市场，期限通常为5年（车企3年）。这些协议是强约束力的“包销”合同——客户提前缴纳了约220亿美元履约保证金。按最低合同价计算，剩余期限累计保底营收约1000亿美元。
简单说：未来三到五年的饭，已经提前端到桌上了。
美光最赚钱的两块业务——云存储和数据中心——合计贡献了Q3约61%的营收。DRAM贡献了总营收的76%，NAND占24%。
美光的上游是半导体设备和原材料供应商：硅片来自日本信越化学等，光刻机来自荷兰ASML。此外，A股有一批公司深度绑定了美光的供应链——太极实业、深科技等做封测，雅克科技供HBM前驱体材料，澜起科技供内存接口。
美光采用IDM（集成设备制造）模式——从设计、制造到封装测试全部自己干。这是它和纯设计公司（如英伟达）最大的区别：它既设计芯片，也自己盖厂生产。
美光的客户不直接面对普通消费者，但几乎覆盖所有科技产品：英伟达、AMD（AI芯片）、苹果、小米（手机）、特斯拉（汽车）、亚马逊、微软（云服务）。
全球DRAM市场，三星以38%市占率排第一，SK海力士29%排第二，美光22%排第三。三家合计占据全球DRAM市场超70%的份额，构成了典型的寡头垄断格局。
美光在产业链里的角色，本质上是“核心零部件供应商”——不直接面对消费者，但离开它，整个AI产业链都得停摆。
拆完美光的基本面之后，如果想继续深入研究，我一般会用万联摩尔的企业洞察功能做三件事：
第一步：企业深度洞察——先查公司基本信息、主营业务、股权结构、风险信息和上下游关系，把这家公司的“底子”摸清楚。
第二步：企业财务分析——看盈利、成长、偿债、营运能力。比如美光Q3营收增长346%、毛利率84.9%、资产负债率24.9%，这些数据到底意味着什么，财务分析能帮你拆得更细。
第三步：舆情热点追踪——追踪最近有没有订单、客户、政策或风险变化。比如美光与Anthropic的合作、16份SCA协议的签订、美国MATCH法案的推进，都是通过舆情追踪才能及时捕捉到的动态。
1. 订单能否持续。16份SCA协议锁定了未来3-5年约1000亿美元的保底收入，但新订单的获取速度同样关键。2026年HBM产能已基本售罄，后续产能释放节奏决定天花板。
2. 毛利率能否维持。84.9%的毛利率在制造业里堪称“怪物级”。但几乎全部利润来自涨价而非卖出更多产品——一旦供需逆转，价格回调的速度同样惊人。
3. 产能释放节奏。美光在爱达荷州建两座晶圆厂，首座预计2027年年中产出首批晶圆，第二座要到2028年底；还规划了纽约州四座晶圆厂的生产集群。新产能什么时候能跟上需求，是决定这轮景气能持续多久的关键。#长鑫科技上市，全球存储竞争添变量 $MU