Investing The Future
香港
Joined February 2020
- Tweets4.9K
- Following4.2K
- Followers3.1K
- Likes6.5K
0xInsights.eth retweeted
Everyone believes CUDA is NVIDIA's moat, the reason we've maintained our dominance in the AI era. I've never believed that.
NVIDIA has only one true moat: its people. And the most important of them is Jensen Huang.
I've worked on CUDA compilers, libraries, and DSLs for eight years, from Volta to Rubin. I think I've earned the right to share my perspective on what NVIDIA's moat really is. I hope you'll bear with me.
1. What is CUDA?
Many people think CUDA is just a programming language and its compiler, and that NVIDIA's moat comes from the massive amount of software written in CUDA, making it impossible for developers to leave NVIDIA GPUs.
This is completely wrong. It's an incredibly naive understanding of CUDA.
To me, CUDA is everything NVIDIA does.
CUDA isn't just a language, software, or hardware architecture. It's a product defined by a unique customer experience.
When you find GPU programming too difficult, CUDA gives you an easier way. When your GPU code isn't fast enough, NVIDIA engineers respond and help you solve it. When a handful of GPUs can no longer push your technology forward, NVIDIA has already scaled its computing platform from desktops to massive data centers. When you need more data centers than anyone else can supply, NVIDIA has already secured wafer and HBM capacity years in advance to meet your demand.
That experience is CUDA.
It's made possible by the relentless, day-to-day work of everyone at NVIDIA. And the person who brings all these people together, who gives them a shared vision and purpose, is Jensen Huang.
2. What made CUDA possible?
Patience. Perseverance. An unwavering commitment to a vision.
Around 2006, NVIDIA bet that the future would demand far more computing power, and began transforming GPUs from graphics processors into a platform for general-purpose computing. Initially, it was a commercial disaster. The added costs hurt margins, the market wasn't ready, and the strategy faced enormous skepticism and pressure, including questions from the board.
But Jensen persisted.
For roughly a decade, NVIDIA kept investing in CUDA, building its hardware, software, and developer ecosystem long before the commercial payoff became clear.
Eventually, deep learning arrived. And NVIDIA was ready.
Today, this sounds like an inspiring story. But for those who lived through it, it meant a decade of perseverance, doubt, and being misunderstood.
Looking back at the history of technology, how many people could have done the same?
This is what I've always loved most about NVIDIA.
And now, Jensen has been pushed out.
This won't kill NVIDIA. The company can still create tremendous commercial value. But it has killed NVIDIA's soul. And how far can a body go without its soul?
If this is how Jensen Huang is treated today, how will those of us who built CUDA be treated tomorrow?
I love NVIDIA deeply, but I have to say it: NVIDIA is dead.
公司核心动荡的形势下，CFO确实是一个非常好的守成CEO替补，因为财务总监是面上的，方方面面都了解，确实是当下最好的选择。
黄仁勋的老朋友，公开说英伟达离了他照样转。
可没过多久，同一个人在评论区又补了一句完全不同味道的话。
Brad Gerstner，硅谷知名科技投资人，Altimeter Capital的创始人，今天在X上说，黄仁勋是朋友，也是传奇，但英伟达从来不是他一个人的独角戏。
他给了一组数字，英伟达有4.2万名员工，光研发就有3.1万人。他还点了一串高管的名字，让大家自己去问Claude这些人是谁，说正是这些人每天在撑着公司的核心业务。
他说风波归风波，团队照样在交付芯片，下一代Rubin Ultra定在2027年，再下一代Feynman定在2028年，都在按计划走。临时接任CEO的财务长Colette Kress最了解公司的角角落落，是眼下最合适的人选。他还特意说，这不是在把坏事说成好事。
转头他在评论区说，替这位朋友难受，希望他早日洗清，跟家人修复关系，重新回到场上。
一边告诉市场公司不靠一个人，一边盼着这个人回来，硅谷大钱这是在两头稳军心。对 $NVDA 来说，大佬给团队背书能托住情绪，但真正定方向的，还是接下来的业绩指引。
人家公司离了老板照样转，散户离了消息就找不到北。
65
越大的人物，就更需要面对各种无情的斗争，直到变成一台无情的斗争机器。
善于斗争，以退为进，伺机而动。
川普透露黄仁勋刚辞职就打电话，说对不起妻子和家庭。
这位一手打造了英伟达（NVIDIA）的掌舵人，突然退出了自己建立的算力帝国。
英伟达董事会已经任命首席财务官科莱特·克雷斯（Colette Kress）接任临时首席执行官。
辞职背后，是一起涉华风波。
核心人物是一名被称为爱丽丝·黄（Alice Huang）的女子。
她被指控涉嫌违反美国出口管制向中共国企业转移芯片，
还涉嫌为中共国国家安全部工作。
突然从最高管理层退下来，黄仁勋的真实状态如何？
川普在访谈中讲到了两人通电话的细节。
按照川普的转述，黄仁勋在电话里说自己正在经历地狱。
他心里最难受的，就是对妻子和家庭造成的伤害。
谈到黄仁勋的技术能力，川普给的评价很高：
说他是个难得的天才，带领英伟达造出了谁也抄不走的顶级芯片。
川普还表示，凭他的本事，以后迟早能重新站起来。
至于涉案女子到底是不是间谍，川普也说，目前官方还没有定论。
英伟达一把手的位置空了出来，谁来接手？
川普在访谈里给出了自己的设想：
让库克（Tim Cook）来管英伟达。
按川普的想法，刚从苹果（Apple）半退休的库克是个绝佳人选。
当年乔布斯（Steve Jobs）留下了天才的创意和设计，
正是库克接手，把苹果带成了全球顶尖的企业。
如今英伟达的技术底子已经无人能及，正好需要一位擅长供应链和大体量运营的帅才。
由库克来接任，在他眼里再合适不过。
当然，这只是川普个人的提议。
英伟达董事会还在寻找正式的永久首席执行官。
一边是涉嫌间谍活动与违规出口的法律调查，
一边是万亿芯片帝国的权力更迭。
这位曾经站在全球人工智能浪潮最顶峰的人物，
以这种方式退场，确实出乎了所有人的预料。
51
0xInsights.eth retweeted
黄仁勋说一语成谶：要包容失败，但失败的过程要快，发现苗头不对，要及时止亏…黄仁勋用了33年，把英伟达从0带到市值5万亿美元。华盛顿有些人却觉得，凭几场听证会、几句政治口号，自己就有资格决定这场AI革命该由谁来领导。不荒唐？英伟达的护城河是什么？只是昂贵的GPU吗？
男女关系的事情，只要不是当场捉奸，到底爱了还是没爱，爱的刻骨不刻骨，当事人不承认，这都无法做实的。
成人的世界，玩起了小孩子的游戏。
11 月 9 日，我做多了英伟达。
理由很简单：当时能决定黄仁勋去留的人都在保他。董事会说信任，特朗普说他该留任。
两天后，这两个理由都靠不住了。
今天早上我把整件事从头复盘了一遍，结论是：这不是两周的事，引线埋了四年。
2022 年美国出台首轮对华 AI 芯片禁令，黄仁勋的应对主要是商业上的：推出降规版芯片接着卖，嘴上批评管制，但没有亲自下场游说白宫。
2024 年初他认识了黄乐。2025 年起，他开始高调游说华盛顿，多次面见特朗普，结果是 H20 和 H200 先后放行。中国的大模型也从这时候开始快速缩短与美国模型的距离。美国情报部门的备忘录认为，推动英伟达游说正是黄乐的任务之一。
放行惹怒了两党的对华鹰派。与此同时，情报部门在调查 Megaspeed 时发现了两人的关系。2025 年 10 月和 2026 年 5 月，特朗普两次听取汇报，都没有动作，5 月还带着黄仁勋去了北京。10 月 6 日，英伟达收于 239.24 美元，创历史新高。
内部渠道走不通，就只剩媒体。10 月 29 日《纽约时报》首次曝光，11 月 8 日黄仁勋承认出轨。
11 月 9 日我的判断是：间谍这条线已经摊开，董事会和白宫都公开力挺，最坏的消息出完了，这事算翻篇。
结果今天早上第二轮报道出来：《纽约时报》称他与一名直接下属有染，《华尔街日报》称董事会准备罢免他。两条都是匿名信源，公司还没确认。这是我完全没预想到的黑天鹅。第一段关系是国家安全问题，白宫可以保；第二段是公司治理问题，董事会很难再保。
董事会不算黄仁勋本人有 10 人。按我的推测，保黄派只有 3 位元老，倒黄派有 5 人，另外 2 人摇摆，过半只需要 6 票。倒黄派里有 4 位持股很少，罢免带来的股价下跌对他们几乎没有影响，不罢免反而可能被股东追责。所以我认为罢免大概率会通过。没有黄仁勋的英伟达，估值要大打折扣，分析师估计还要再跌 25% 到 30%。
不管今晚结果如何，我都会在开盘后平掉全部 $NVDA 期权。即使罢免没通过，放料的人手里可能还有东西，短期看不清。
往后看，对华鹰派很可能成为华盛顿的主流。一直对华发疯的 Anthropic 会是 AI 板块里最大的赢家。
45
拭目以待，看看世界最大的公司公司治理是否完善，这是商科教授一次难得的现实教材。
这个公司治理理论发扬光大的地方，能否真的践行股东利益最大化原则。
给大家科普一下，为什么这一次爆料老黄跟直属下属有恋情，杀伤力比女间谍还大？
2019年麦当劳曾经发生过一起董事会保 CEO 的案件：当时的CEO Easterbrook与员工发暧昧短信，董事会一开始保他，让他体面地拿了4000万美元的离职红包。
结果几个月后的调查发现，他又与另外三名女员工有不正当关系。最终美国证监会（SEC）直接下场处罚，以证券欺诈罪起诉Easterbrook本人，还把麦当劳公司也给告了，理由是董事会放水行权误导了股东。
很多人的直觉是Jensen是英伟达的灵魂，董事会肯定得保他。其实第一次爆料黄仁勋的时候，董事会确实是想保他的，甚至当时还说了对这个调查充满信心。可是自从当年有了麦当劳这个先例之后，如果黄仁勋现在跟直属下属发生不当关系，董事会还想保他，在法律内部调查程序上就属于 corrupting internal processes。
这种案子是有先例的，在西方法律里面，precedents are very very important.
董事的法定责任是 fiduciary duty，意思是他们效忠的是全体股东跟公司法人，而不是 CEO。
当英伟达现在面临美国司法部刑事调查、FBI 介入、国会两党发传票，以及 CEO 多个未申报的利益冲突时（我们还不知道之后会不会有更多爆料），这时候董事会继续保他就等于违反合规红线。
一旦日后发生更多的股东诉讼，甚至起诉这 10 位独立董事个人违反义务，那这 10 个人都会面临倾家荡产的调查。你们自己看看这 10 个人都是谁：他们都是身家百亿的风投教父，包括前高盛高管和知名商学院院长，没有人现在会为了保另外一个人的帅位，去赌上自己一辈子的声誉和个人身家。
investor.nvidia.com/governan…
44
0xInsights.eth retweeted
$MRVL has lost about 50% of its value in one month.
Today CEO Matt Murphy emailed every employee to explain it, under the subject line "Putting the Week in Perspective."
What happened, in his words:
1. Last week Marvell agreed to combine with $LITE. Murphy says "some of the initial analyst commentary and media coverage reflected a misunderstanding of the strategic rationale."
2. The large shareholders he met in Boston and New York "told a very different story." He says they understood the case for joining silicon and optics, and recognized Marvell's acquisition track record.
3. Their most common question was "why now?" His answer points to Inphi, which put Marvell ahead of the move to scale-out optics. Optics are now moving into scale-up and scale-in, and "customers are making important technology decisions now that will shape the next generation of AI infrastructure."
4. On the stock, he gives the number plainly. "Marvell shares declined approximately 51%, closely in line with the 49% average decline among our connectivity peers." $LITE saw a similar decline.
5. He ties that to a selloff across AI stocks. Part of the cause is "uncertainty about how long the current AI infrastructure investment cycle will continue." He adds that the midterms "brought increased attention" to data center expansion and "the possibility of additional restrictions."
6. His counterpoint is the longer window. "Marvell shares are still up approximately 66% year to date," ahead of both connectivity and broader AI peers, and he says the long-term trends behind AI infrastructure spending "remain strong."
7. He ends by telling staff the deal alone won't pay off. "Simply acquiring a company doesn't create value. It's what we do afterward." That part is still far off, because the deal isn't expected to close until the first half of 2028 and the two companies stay independent until then.
0xInsights.eth retweeted
Fable 5.2 shows substantially improved skills in controlling physical robots, even on tasks it has never explicitly been trained to do.
On RoboDojo, a benchmark with a simulated two-arm robot, it outperformed other models and most AI systems built to control robots.
作为公众人物，斗争是永远的剧本，学会如何聪明的斗争是核心技能之一。
We just issued a press release from our board of directors. Read the release: nvidianews.nvidia.com/news/s…
68
0xInsights.eth retweeted
$LITE 跌了 45%，CEO Hurlston 出来回应：需求订单强得不能再强，和股价完全脱节；被问到买家 $MRVL，他说董事会从没考虑过全股票之外的方案，看好 Marvell 从当前位置往上的空间，还做了尽调才拍板。
卖方 CEO 愿意把对价全押在买家股票上，这个信号比嘴上表态实在。交易目标 2028 年上半年交割，他说剩下的最大障碍是监管审批——后面就盯这一关的进度。
$LITE CEO Michael Hurlston addressed the 45% selloff today.
On demand, he said “There’s a massive disconnect between what we’re seeing in our demand book and what the markets are showing. Our demand couldn’t be stronger.”
Asked whether the market reaction gives him pause on $MRVL as the buyer, he said “We really like Marvell. We think they’ve got a lot of upside from where they are, and that’s part of the reason we took all stock.”
He admitted “the timing was bad” but didn’t back off. “Our board of directors never considered anything but taking all stock,” citing diligence on Marvell’s books and design wins.
On getting the deal closed, targeted for 1H28: “It’s going to be a bumpy one.” He named regulatory approval as the biggest hurdle left.
$LITE $MRVL
0xInsights.eth retweeted
China is deploying AI infrastructure using the same real-estate development model that previously imploded. There are no obstacles from environmental impact assessments or local residential communities; the only catch is that it is bound to end in a mess of "unfinished projects"—the equivalent of the "half-built, abandoned buildings" seen in the property sector. Both the limits of financial credit and the physical constraints of hardware have already begun to manifest.
参考光伏，新能源车，只要国内掌握了核心科技，AI能给你干到白菜价。
工业内卷能力，全世界第一☝️
We Visited China’s Stargate in Inner Mongolia
A wind-powered 800MW campus,
120MW built within a year,
a 5MW SST-native 800V DC pilot and a cooling design without central chillers
newsletter.semianalysis.com/…
1
68
0xInsights.eth retweeted
ARM CEO ：
美国原本希望把AI数据中心留在本土，是为了把算力、就业、税收、能源投资和产业链一起留在美国。
但如果审批、电力和社区阻力越来越大，算力不会消失，只会流向加拿大、东南亚、中东和欧洲。
美国少建 ≠ AI需求消失，
更可能只是全球CAPEX重新分布。
真正流失的，反而可能是美国自己的就业、税收和配套产业。
所以“限制美国数据中心”不等于“AI结束”。
对于这些公司，需求更可能是换地方建设，而不是消失。
他妈的股民成了党争的牺牲品😭
“The Stock Market suffered one of the Worst Days in History based on the results of the Election. Such a sad thing to watch! It took so much work to get us to a level that no one has ever seen before, and then this. Your 401(k)s have, therefore, dropped. Remember, it wasn’t Republicans, it was Dumocrats...”
2
98
