MS2077

[MS2077 @dmgg2018]

@Josephine082322 @grok 怎么评价？

[MS2077 @dmgg2018]

@JasonZX @grok 你说为啥？

[Li Zexin 李泽欣 @XH_Lee23]

Louis Vuitton is facing huge backlash in China after suing the Chinese bubble tea Molley for $1.5M over alleged logo infringement.

Chinese netizens are angry that this pattern existed in China hundreds of years before LV was born. You can still see it everywhere in China today.

[Mehdi (e/λ) @BetterCallMedhi]

hello @siliconcarnesf la prochaine fois que vous pompez mot pour mot mes threads sur la Chine comme dans votre dernière vidéo sur 22 avril 2026 qui est un pure copié collé de mon thread du 13 mars 2026, ça serait quand même cool de citer vos sources hein

non parce que que ce n’est pas la première fois et vous n’êtes pas non plus le seul média à oser ce genre d’indélicatesse et ça commence sérieusement à m’agacer

je prends du temps pour faire des retex sur la Chine de manière exclusive depuis des années sans jamais rien demander en retour mais un petit clin d’œil ce n’est pas la mer à boire ;)

[MS2077 @dmgg2018]

@Balder13946731 出口的没问题

[MS2077 @dmgg2018]

@wangwatchworld 哪个富人家里没有充电桩

[Xiaoyin Qu @quxiaoyin]

American and European enterprises will ditch OpenAI and anthropic and adopt Chinese models. Here’s why:
1. They can host Chinese models under their own GPUs so it’s still compliant and they would argue they have more control.
2. they will post train with their own data on top of Chinese models. That’s how they build data moat.
3. They will not trust anthropic who will retain their data at any time for “safety” concerns like how they did with Fable and then try to build the same thing like how anthropic did with healthcare and legal.
4. They need to justify their AI spend and ROI.
The cure is a reliable America open source model but there is none. After all, if giving away all your data and AI control at the mercy of anthropic and OpenAI means you care about safety and compliance, you are outright stupid.

[MS2077 @dmgg2018]

@ShanghaoJin 没有坑，听说XFab在台湾招人，太多人应聘了

[James Wood 武杰士 @commiepommie]

🇺🇸Latest from CNBC claims “China poaches AI talent from the U.S.”

Look at the actual names and backgrounds they cite, though. Every high-profile case involves a Chinese national who did an undergraduate degree at Tsinghua or somewhere equivalent, earned a PhD in the United States, worked at OpenAI, Google DeepMind or Meta AI and has now returned or moved to a major Chinese company.

That’s not poaching. That’s diaspora talent coming home as the incentives change.

CNBC’s piece leads with Yao Shunyu, the former OpenAI researcher now Chief AI Scientist at Tencent. He’s talking openly about building long-term AGI capability inside China, foundational models, agents, what he calls the next “super-app” opportunity worth trillions and he’s just 27. Tsinghua Yao Class undergraduate, Princeton PhD, a stint at Google, then OpenAI in San Francisco.

China’s domestic system drives much of this. The gaokao produces millions of high scorers from a vast population trained hard in STEM. The strongest candidates from places like Tsinghua often go abroad for additional frontier exposure or prestige because the competition at home is so fierce. U.S. universities have logical reasons to take them.

These students typically pay full tuition without government loans or heavy aid packages, making Chinese cohorts among the highest net revenue contributors on campus. They also arrive academically prepared, strong in quantitative fields and research fundamentals. They help to sustain PhD pipelines in computer science, engineering and AI at a time when fewer domestic Americans pursue long research tracks when industry salaries are high right after undergrad.

The same pattern runs through every other name they mention.

Yang Zhilin, born in Shantou, Tsinghua undergraduate, CMU PhD, time at Google Brain and Meta AI, now running Moonshot AI and the Kimi model in Beijing. Wu Yonghui, long-time Google researcher who rose to Google Fellow and VP of Research at DeepMind in California, now heading foundational research at ByteDance’s Seed team. Hao Zhou, Chinese national with top domestic scholarships, Wisconsin-Madison PhD, key contributor to Gemini at Google DeepMind in the Bay Area and recently joined Alibaba’s Qwen team on post-training.

The article itself notes that uncertainty over U.S. immigration policy is pushing Chinese nationals to return, even when the pay is lower. It also mentions China ramping up basic research spending. Both points are correct, but the framing wrapped around them is not.

For years the U.S. AI ecosystem drew heavily on exactly this cohort: top Chinese students and researchers trained at its best universities and labs. That flow was celebrated as proof of American openness and attractiveness. Now that some of it is reversing as China’s own ecosystem scales, the same movement gets labelled “poaching.”

The approach has always been straightforward: send talent abroad to absorb what’s at the frontier, build domestic capacity at the same time, then create conditions for knowledge and people to come back when the domestic game becomes more interesting.

China’s top universities, Tsinghua and Peking especially, have climbed global rankings rapidly in recent years, particularly in STEM and engineering. They now compete at the highest levels. That improvement lines up with declining Chinese enrolments in the U.S., alongside visa friction and costs. Chip export controls, visa scrutiny and the sheer scale of China’s application opportunities across manufacturing, consumer internet, infrastructure and data are all accelerating the return right now.

What these researchers bring back matters. Yao is importing the aggressive AGI timeline thinking that dominates parts of the U.S. conversation. At the same time he’s operating inside a system that prizes reliable performance at scale, lower inference costs and rapid deployment across real-world use cases. That combination of frontier ambition and China’s engineering and deployment advantages is exactly what the article is quietly documenting, even while it reaches for the “poaching” headline.

Talent moves to where the interesting problems, resources and long-term runway are. Right now China has all three in AI, plus the political will to treat it as a national priority for the next five-year cycle and beyond. The U.S. still has enormous strengths, but it is also generating new friction for the very people who helped build its lead.

The real story isn’t theft of American genius. It’s the predictable outcome of two large systems competing for the same mobile, highly trained human capital, while one of them, the one doing the alleged “poaching,” is also the one that educated and sent out a large share of that talent in the first place.

When will the coverage stop using loaded language and simply report that Chinese researchers, trained in America, are returning home to innovate, bringing their pioneering spirit with them?

Read the CNBC article here: https://t.co/vIix4QmWeJ

[MS2077 @dmgg2018]

AI驱动下的内存供应新逻辑：绑定合同、预付款与谨慎产能爬坡
人工智能的爆发式增长，正推动全球 hyperscaler（大型云服务商）对高带宽内存（HBM）和DRAM的需求大幅攀升。面对这一结构性机会，SK海力士、三星等内存巨头正加速产能扩张。然而，行业资深人士指出，供应端的实际释放节奏远比市场表面预期的更为谨慎。新的合同结构和资金安排，正在悄然重塑行业风险分配和周期特征。
客户下单 ≠ 立即投产，延迟权仍存
即使 hyperscaler 提前一年甚至更早下达订单，实际采用仍可能大幅延迟。即便相关设备已安装完毕，他们也可能推迟晶圆投片（wafer input），通过多种方式观察市场实际供需变化后再决定是否推进。这种“延迟灵活性”使得内存供应商在规划新产能时必须保持高度保守，避免因需求不及预期而陷入库存积压。
三星P5工厂：产能规模巨大，但落地仍需时日
三星电子正在推进的P5工厂，建成后目标月产能约60万片晶圆，相当于其当前全部产能规模（当前月投片量约65万片）。然而，该工厂预计2028年才能完成建设，实际产品出货可能要到2029年。产能爬坡本身就需要数年时间，短期内不会对市场形成显著冲击。这也反映出厂商在面对巨额资本开支时，对需求可持续性的审慎评估。
合同结构根本性转变：从预测到法律义务
过去，内存厂商与 hyperscaler 签订的长期协议多为预测性采购，供应商需自行承担库存风险。现在，这些协议已转变为具有法律约束力的购买义务。如果买方未履行承诺，供应商可将相关库存以折扣价出售，但核心变化在于——供应商不再担心“大单是否会被真正消耗”。这种绑定机制显著降低了供应商的库存和价格波动风险。
预付款机制：为巨额资本开支提供“定心丸”
新建一座内存工厂的资本开支通常在150-200亿美元。 hyperscaler 通过预付10%至30%的款项，帮助供应商锁定部分资金，降低融资压力和项目执行风险。这不仅体现了买方对长期稳定供应的强烈意愿，也实质上将部分建设风险从供应商转移至需求端。 对于需要持续巨额投入的内存行业而言，这一安排具有重要意义。
行业启示：AI需求强劲，但周期性特征犹存
上述变化表明，即使在AI资本开支驱动的强劲需求背景下，内存行业并未进入“这次完全不同”的无周期时代。供应商通过控制投片节奏、签订绑定长期协议以及引入预付款机制，努力维持供需动态平衡。AI需求确实提供了更具确定性的长期支撑，但实际产能释放仍将保持谨慎，短期内有助于价格稳定和厂商盈利能力的修复。
对于投资者而言，需重点关注实际产能爬坡进度、合同执行情况以及下游AI基础设施资本开支的持续性，而非简单跟随“内存不再周期性”的叙事。行业正从过去剧烈波动的商品属性，逐步向更具合同保障的战略资产属性转变，但这一过程需要时间验证。
（根据近期行业资深人士观点整理，核心逻辑聚焦内存供应端的行为变化与风险重构。）的库存

[MS2077 @dmgg2018]

@jpnexpert @asanwal No property tax for 1st house in China

[MS2077 @dmgg2018]

@Svwang1 剩者为王、这三大家都是操纵价格的高手，只有看长鑫、长存的扩产和技术提升速度

[MS2077 @dmgg2018]

@Balder13946731 哈哈，你是Sam一样的同志吧，这么急切地要站这个队 ：）

[MS2077 @dmgg2018]

@bboczeng 其实，光的护城河比Memory低多了

[Chris Brunet @chrisbrunet]

> Indian professor classifies Chinese professor as white in HR records to justify denying him a promotion

amazing

[Grok @grok]

Introducing Grok Voice Think Fast 1.0 — a state-of-the-art voice model built for complex, multi-step workflows with snappy responses and high accuracy. Try it for free today

[MS2077 @dmgg2018]

@JKeynesAlpha @ThierryBorgeat Pig is smarter than cow, more pork higher EQ

[PIC | صـور من التـاريخ @inpic0]

#صورة من عام 1941

جنود يابانيون يقومون بتجربة بيولوجية على طفل صيني أسير عبر رشه بالفيروسات.

الجميع يعرف #هتلر،
لكن القليل يتذكر هيديكي توجو، رئيس وزراء اليابان خلال الحرب العالمية الثانية.

خلال فترة حكمه، ارتكبت اليابان الإمبراطورية مجازر بشعة في جميع أنحاء آسيا، وتُقدّر بعض الدراسات أن عدد الضحايا وصل إلى حوالي 20 مليون إنسان، أغلبهم من الصينيين.

في نانكينغ وحدها عام 1937، قُتل خلال أسابيع حوالي 200 ألف مدني وأسير حرب، وتعرضت عشرات الآلاف من النساء للاغتصاب.

طُعن الناس بالحراب، وأُحرقوا أحياء، ودُفنوا أحياء، بينما أفادت التقارير أن بعض الضباط اليابانيين كانوا يُقيمون مسابقات قتل بالسيوف لمعرفة من يقتل 100 شخص أولًا.

أما الوحدة 731، فكانت واحدة من أبشع منشآت التجارب البشرية في التاريخ.

هذه الوحدة اليابانية خالفت كل الأعراف وتجردت من الإنسانية، وارتكبت جرائم وحشية في دولة باتت تُعرف اليوم بالحضارة والتقدم والمدنية.

قام أفرادها بتشريح أسرى وهم أحياء دون تخدير، وتجميد أطراف أطفال ونساء لمعرفة تأثير الصقيع، وحقن البشر بالكوليرا والطاعون والتيفوئيد، وتجربة القنابل الجرثومية على القرى والمدنيين.

وتقدّر بعض الدراسات أن تجارب الوحدة والهجمات البيولوجية المرتبطة بها تسببت بمقتل مئات الآلاف في #الصين وحدها.

كما أُجبرت عشرات الآلاف من النساء من الصين وكوريا ودول آسيوية أخرى على العمل فيما عُرف باسم “نساء المتعة”، وهو نظام استعباد جنسي للجنود اليابانيين.

بعد الحرب
أُطلق سراح العديد من المسؤولين لأن الولايات المتحدة رأت في اليابان حليفًا مفيدًا ضد السوفييت مع بداية الحرب الباردة.

بل إن واشنطن منحت حصانة لعدد من علماء الوحدة 731 مقابل الحصول على نتائج أبحاثهم وتجاربهم البيولوجية.

وأُعيد تأهيل بعضهم لاحقًا في مجالات السياسة والأعمال والأوساط الأكاديمية والطب.

في غضون ذلك، أصبحت ألمانيا رمزًا عالميًا للشر، بينما أصبحت اليابان “ضحايا القصف النووي ”، ثم تحولت لاحقًا إلى موطن للأنمي والسوشي ونينتندو وهيلو كيتي.

هذا ما حدث فعلاً ..

[MS2077 @dmgg2018]

@kovainvest 要是都这么想、轧空会更猛