Post

SK海力士到底有多赚钱。二季度营收79.3万亿韩元（约545亿美元），同比暴增257%。营业利润60.5万亿韩元（约420亿美元），同比暴增557%。营业利润率76.3%。每卖100块钱的东西，净赚76块。上半年累计营收首次突破100万亿韩元。任何一个数据，都够让99%的上市公司跪下叫爸爸。然后呢？财报公布当天，韩股盘中一度暴跌超19%，创历史最大单日跌幅。韩国KOSPI指数连续两天熔断。一份创纪录的财报，引爆了整个国家的股灾。市场在怕什么？怕的是那个刻在存储芯片行业骨子里的死亡循环：涨价→扩产→过剩→暴跌→破产。这个循环，过去三十年杀死了无数半导体公司。 2008年，DRAM价格崩盘，德国奇梦达直接破产。2015年，存储芯片寒冬，美光亏了20多亿美元。每一次“超级周期”的狂欢，最后都是一地鸡毛。现在SK海力士在干什么？上半年购置有形资产的现金支出超过18万亿韩元，同比增长超70%。全年资本开支预计达到40万亿至50万亿韩元区间的高端水平。龙仁晶圆厂一期总投资31万亿韩元。清州M15X工厂、M17工厂、P&T7封装厂——全线开工。整个SK海力士，正在用史上最猛的速度扩产。市场看着这一切，脑子里只有一个问题： “这么多产能砸下去，两年后谁来买单？” 但这一次，SK海力士学乖了。它没有像2008年那样闭着眼睛扩产等死。它干了一件存储芯片行业从未有人干过的事——把需求锁死在合同里。 2026年7月，英伟达与SK海力士签署了一项价值可能高达5000亿美元的长期供货协议。协议内容包括共同开发下一代AI内存，确保HBM的稳定供应。这还不是孤例。市场采购模式正在发生结构性变化——大型云服务商从传统的一年期合同，全面转向多年期供货协议。客户用预付款和保证金来履约。产能被提前锁定到2028年，甚至2029年。供给端呢？EUV光刻机交货周期超过两年半——想扩产？等三年。需求端锁死，供给端卡死。这就是SK海力士打破“死亡循环”的办法。比特币过去三次减半，每次都有人说“供给减少，必涨”。结果呢？2022年减半后，比特币从69000跌到16000。为什么？因为只有供给侧约束，没有需求侧锁定。矿工可以减半，但大户可以砸盘。ETF可以买入，但机构可以赎回。加密市场最大的问题从来不是“不够稀缺”——而是“没人承诺一直买”。 SK海力士用7200亿美金（约合1000万亿韩元总投资）告诉世界一句话：穿越周期靠的不是信仰，是把需求锁死在合同里。比特币有“减半”这个供给侧硬约束。但直到今天，加密市场仍然没有“长期协议”这样的需求端锁定机制。 ETF是第一次institutional级别的长期资金入场，但它依然可以被随时赎回。真正的“周期穿越”，需要买卖双方把未来几年的量价都写在纸面上，违约就赔钱。这对我们炒币的意味着什么？第一，别再把“减半必涨”当真理。供给减半不等于价格翻倍——没有需求锁定的减半，只是把同样的故事再讲一遍。第二，ETF是第一步，但不是最后一步。当加密市场出现“机构级长期托管+不可赎回的锁仓机制”那天，才是真正的周期穿越。第三，SK海力士的剧本告诉我们：真正聪明的钱，在周期顶部干的是“锁需求”，不是“赌涨价” 。存储芯片行业用三十年血泪换来的教训—— 周期里死得最惨的，永远是那些只信仰“供给稀缺”、不锁定“需求确定”的人。比特币减半的故事讲了十几年了。下一个十年，该讲点新故事了。 $SKHYNIX $SKHY $xSKHY #海力士扩产提速，资本开支能否兑现回报

NVIDIA spent $21 billion to become a SpaceX shareholder, losing $4 billion in a month—should we be happy or worried? On August 14, NVIDIA filed its 13F with the SEC for the first time disclosing its holdings—as of June 30, it held about 122.8 million Class A shares of SpaceX, with a quarter-end market value of approximately $21 billion. According to FactSet data, NVIDIA is the sixth largest shareholder of SpaceX. Elon Musk himself firmly holds first place with $850 billion, and Alphabet ranks second with $78 billion. But note a key detail: this $21 billion was not recently bought by NVIDIA on the secondary market. Insiders revealed that NVIDIA’s SpaceX shares originated from its roughly $10 billion investment in xAI in January this year. In February, SpaceX acquired xAI at a $1.25 trillion valuation, and the investment was converted into SpaceX shares accordingly. In other words, NVIDIA didn’t bottom-fish; the shares initially invested in xAI were "forcibly converted" into SpaceX stock. Now, let’s talk about what SpaceX has been through recently. On June 12, SpaceX was listed at $135. On the third day, it surged to a historic high of $225. Then what? It went straight down. On August 5, SpaceX plummeted 13.61%, closing at $108.27. On August 3, it hit a new low of $104.83. From $225 to $104, it was a direct 50% cut. Until the past few days—Musk announced "AI revenue will surpass all other businesses in September," and "AI will account for 99% of the company’s value within five years," plus the release of Grok 4.6, the stock rebounded to around $146. But even so, it’s still down 37% from the $225 peak. NVIDIA’s $21 billion holding was calculated at $170.86 per share at the end of June. Now SpaceX closed at $140, meaning nearly $4 billion evaporated on paper. So here’s the question: if you are an SPCX holder and NVIDIA is the sixth largest shareholder—should you be happy or worried? My conclusion: short term is pressure, long term could be a moat. First, short-term pressure—why it might still fall. 1. NVIDIA itself has motivation to reduce holdings. The position shrank from $21 billion to $17.2 billion, with a nearly $4 billion unrealized loss. Any institution facing such a large floating loss would reassess its position. Don’t think NVIDIA is a "long-term value investor"—it’s a public company with its own financial reporting pressures. 2. SpaceX’s technical chart hasn’t stabilized yet. From $225 down to $104, rebounded to $146, but still 37% below the high. Whether $140 can hold is unknown. Whether the $104 low on August 3 will be retested is a real question. 3. Concerns about capital expenditure haven’t disappeared. SpaceX’s Q2 AI-related capital expenditure was $15.83 billion, total capex $18.37 billion—more than twice its revenue. The burn rate hasn’t slowed, so the market won’t stop questioning. Now, long-term value—why it might be an opportunity. 1. NVIDIA is the world’s most knowledgeable company about computing power value. It put real money (though not directly bought) into a $21 billion position, becoming the sixth largest shareholder. If it didn’t believe in SpaceX’s AI prospects, it would have tried to reduce holdings during the conversion. 2. The exclusive chip supply agreement locks out competitors. Musk clearly stated on the earnings call that SpaceX’s AI data centers will fully adopt NVIDIA chips, calling its GPUs the "best architecture." Musk also expects to receive a "large allocation" of NVIDIA Vera Rubin GPUs next year. AMD can’t even get in the door. 3. If Musk’s vision is realized, the current market cap is just the starting point. SpaceX plans to reach 10 gigawatts of AI computing power by the end of next year. At $30 to $50 per watt, that corresponds to annual revenue of $300 billion to $500 billion. The current market cap is only about $1.85 trillion—if this goal is half achieved, the valuation would double. Goldman Sachs just raised its 12-month target price from $205 to $220. Citi sees a long-term target above $900. NVIDIA’s $21 billion stake in SpaceX is not to be a "bag holder"—it’s to be an "accelerator." But accelerators have both gas and brakes. Which side it leans on depends on whether Musk can turn the 10 gigawatt vision into real cash flow. Operational reference (just the essentials): Short-term traders: watch if $140 can hold. If it breaks, the early August low of $104 will likely be retested. Don’t rush to bottom-fish; let the market stabilize first. Long-term investors: focus on two indicators—AI business revenue growth and computing power build-out progress. These are the core catalysts for valuation re-rating. As long as these two exceed expectations, short-term fluctuations are just noise. Position management: don’t go all in. SpaceX’s volatility is there—from $225 to $104 in less than two months. Build positions in batches and keep cash ready for extreme opportunities. $xNVDA $xSPCX #NVIDIA持有SpaceX约210亿美元，AI协同受关注