2025 (@2025Updated) profile picture
Joined January 2015
Photo by 2025 (@2025Updated)Photo by 2025 (@2025Updated)Photo by 2025 (@2025Updated)
Replying to @AnthropicAI
Your AI hates Whites & Asians, especially Chinese, heterosexuals and men. This is misanthropic and evil. Fix it. Frankly, I don’t think there is anything you can do to escape the inevitable irony of Anthropic ending up being Misanthropic. You were doomed to this fate when you chose your name. The Name of the Wind.
3,324
5,663
1,353
64,663
5,184,091
2025 retweeted
十年前的箴言 《花旗半导体投资15条公理》 1. 不要单纯依据估值进行买卖决策。 2. 无论身处何种环境,半导体公司管理层的口径普遍偏向乐观。 3. 半导体具备周期性;不要相信“半导体已经不再有周期”这类说法。 4. 「都市法则」:半导体板块股价通常只由单一核心逻辑驱动——你的任务就是找到这个核心变量。 5. 产品周期推动股价/每股收益上行的时间,往往比你预判的更久;反过来下行周期持续时间同样经常超出预期(这点往往让人踩坑)。 6. 以牺牲利润率为代价追逐营收增长,几乎都无法获得成功。 7. 交付周期拉长时买入,交付周期收缩时卖出。 8. 真正的“时代已经不同”的局面极少出现;这种情况有可能发生,但概率很低。 9. 警惕“这只是单季度的短期问题”的论调。 10. 不要指望管理层主动预判行业下行周期,可参考第2条公理(管理层天然乐观)。 11. 并购交易的成功率,最高的场景是收购方本身就深耕标的所属赛道;收购方缺乏相关行业经验时,交易风险会大幅上升。 12. 大多数科技公司只擅长做好一件核心业务;当这套核心商业模式受到冲击、开始衰退时,企业想要转型做好新业务是极其困难、非常罕见的。 13. 二线厂商最先、最严重地感受到行业下行压力,而在行业回暖时,它们却是最晚、最弱受益的群体。 14. 科技板块的“价值投资策略”,通常只有在管理层更迭、或是商业模式/经营策略发生变革时才会奏效。 15. 当利空消息出台股价却不再下跌,或是利好落地股价不再上涨时,你就需要停下来重新审视判断。
1
50
4
214
28,165
Google’s TPU v9 has an encouraging production update. MediaTek told Morgan Stanley’s Asia Pacific Summit that it’s comfortable with how yields and output are progressing. It still sees EMIB-T substrate supply as key to the ramp. MediaTek also said Google could use TPU v8t and v9 in rack systems for outside AI labs, in addition to using them internally. That could give both chips another source of demand. Timing still matters. A later start would give v9 less time to ramp before v10 arrives, and the report didn’t give an updated shipment timeline.
2
5
1
59
3,889
$INTC just got a massive hidden catalyst Senate just passed the BAAJA bill 62-33. This is the biggest U.S. energy permitting reform in decades. What it actually does: Slashes environmental review timelines from 4+ years → ~2 years Speeds up power plants + high-voltage transmission lines Forces data centers to pay their full energy costs (removing free-rider problems) Why this is huge for Intel right now: Intel’s new U.S. fabs are some of the most power-intensive projects in the country. Ohio, Arizona, New Mexico, Oregon — all of them need massive, reliable electricity. Faster energy infrastructure = → Lower risk of costly construction delays → More competitive power pricing → Higher chance Intel actually hits its foundry volume targets While the market is focused on AI hype and IPOs, the real bottleneck for domestic chip manufacturing has always been power. This bill attacks that bottleneck head-on. Trump already called it a “BIG WIN for U.S. ENERGY INDEPENDENCE.” Quiet but extremely bullish for the only major U.S. company actually building leading-edge capacity on American soil. $INTC holders should be paying attention.
I’ve added to my $INTC position once again! I previously held 9,000 shares, and today I purchased another 3,000 shares at $56.32. 🔥 Additional investment: $168,960 🔥 Current position: 12,000 shares 🔥 Average cost: $70.33 🔥 Total capital invested: $843,900 I am currently sitting on an unrealized loss of $136,620, representing a drawdown of 16.19%. To be honest, a drawdown like this is far from comfortable. Yet, I’ve chosen to keep adding to my position while the market is gripped by fear. My bullish thesis centers on Intel’s upcoming 18A process technology, the growing demand for AI chips, and the long-term potential of its foundry business. I cannot control short-term market volatility, but I will continue to monitor the company’s progress in delivering on its performance goals. **If you also hold $INTC—or share my belief in Intel’s long-term potential—give me a follow!** I will continue to provide transparent updates on my 12,000-share position, documenting every additional purchase and the ultimate outcome. $INTC #Intel
6
16
84
5,148
2025 retweeted
Replying to @chumacn
你打的起这张牌吗? 你把“在全球多源供应链里分到一杯羹的优质乙方”,硬生生幻想成了“能卡死甲方的垄断霸主”。只要把 Vera Rubin 的真实物料清单(BOM)和供应链格局摊开,就知道为什么根本打不了这张牌——因为一打出去,先窒息的不是英伟达,而是自己。 你宣称的“大陆卡脖子王牌”2026 年英伟达 Vera Rubin 真实供应链格局为什么根本无法用来“断供美国”? 1. “Rubin 命脉把控在胜宏超高多层 PCB 上”胜宏科技确实是英伟达 Tier 1 核心供应商,但 2025 上半年其在全球 AI PCB 市占率仅约 13.8%(维基百科);台湾臻鼎、华通、欣兴等同样是高阶 AI PCB 主力(EMSG)。更关键的是,真正卡脖子的上游原材料全在日台韩:Rubin 所需的 M8/M9 超低损耗覆铜板(CCL) 由台光电(EMC)、台燿(TUC)、联茂、韩国斗山、日本松下垄断(Lighthouse Canton);最核心的 T-glass 特种玻纤布 仅日本日东纺(Nittobo)一家就独占全球约 90% 份额(EMSG)。 ① 上游一断,一块板子都压不出来:没有日本 T-glass 玻纤布和台日 M9 板材,大陆 PCB 厂立刻停摆。 ② 胜宏自己都在加速出海:为满足北美客户合规要求,胜宏 2024–2026 年重金收购并扩建泰国工厂(增资 2.5 亿美元)、在越南北宁建厂(投资逾 5–8 亿美元),今年港股募资近 200 亿港元大头就是投向海外产线(凤凰网财经、维基百科)。 2. “散热供应链都在深圳、东莞、惠州”Vera Rubin 全液冷机柜的核心冷板与热管理主力,是美国维谛(Vertiv)、美国 Boyd、加拿大 CoolIT、以及台湾奇鋐(AVC)、双鸿(Auras)、酷冷至尊、台达电(Vertiv 官方新闻、Lianliwork、iConnect007)。黄仁勋在 GTC 台北 2026 明确披露,Rubin 供应链横跨 30 多个国家、超 350 家工厂(Lianliwork)。液冷核心认证与大头份额在美资与台资手里(且多在墨西哥、东南亚、台湾扩产),大陆厂只是其中一部分外围与二级产能,毫无断供筹码。3. “NVLink 的 SerDes 稳定和良率调优在大陆”纯属外行编故事。224G NVLink SerDes 是芯片内部最顶级的模拟/混合信号物理层(PHY)电路 IP,从架构设计、硅片仿真到晶圆良率调优,100% 掌握在英伟达美国/以色列研发团队与台积电手中。PCB 厂顶多配合做板级阻抗匹配与信号完整性测试,把“配合画板子”吹成“掌控 SerDes 核心良率”,就像说印钞机厂掌控了美联储的货币政策。 一句话点破“为什么不打这张牌” 因为这是典型的“单向依赖自残牌”: 对英伟达而言,高阶 PCB 与液冷组件早已刻意设计为多源供应(Multi-vendor)(Lighthouse Canton)。如果大陆单方面断供 PCB,英伟达顶多承受 1–2 个季度的爬坡阵痛,订单立刻全部切给台湾、韩国、日本及东南亚工厂; 对胜宏等“达链”中企而言,它们从二线代工厂跃升为两千亿市值龙头、2025 年净利暴涨 273% 至 43 亿元(凤凰网财经),全靠吃下北美 AI 算力的最高毛利订单。主动断供无异于自断财路、把千亿级市场拱手送给台韩竞争对手,并被永久踢出全球算力供应链。 上面掌舵的产业决策者比键盘侠清醒得多——手里握的是“能赚钱的优质代工份额”,不是“能勒死对手的 EUV 光刻机”,真把“焦大”当成了“贾母”,那才是拿自家企业的命去开玩笑。
6
8
1
94
12,019
🚨 BREAKING: Jensen Huang is reportedly considering founding a brand-new semiconductor company 🚨 He could raise tens of billions on day one, would have full founder control, and could build fast on his own terms. With the relationships he's built over decades, he could easily plug into the #AI ecosystem. Imagine the demand and customer base lining up before his first chip even ships... @grok Could Jensen Huang launch a competitor to $NVDA, attract tens of billions in funding, and lock in major customers?
1,958
4,076
825
35,270
4,722,245
Big News: Intel @ TomorrowX + Nova/Razor Lake + Unified Core - 18A yield at 70% - First 14A product in 2027 - 14A volume ramp in 2028 - 14A external customers 2028/29 - 14A volume ramp for external in 2028/29 (like TSMC A14) - 14A will have parity with A14 - New (external) Unified Core team 141 heads. Will spin in. X86S Unified Core with HT Amazing news is, my original Rosaic report was spot on!!! Headed by Top CPU Architect Peter Bannon (who worked with Jim Keller on Apple Silicon A4 & A5) is now working on the new Intel X86S Unified Core with Hyper Threading. Current team head count is at 141 as confirmed by Lip-Bu. This is Lip-Bu's wildest move ever. His secret weapon to leap frog competition in performance, PPW & PPA. Nova Lake Performance & Competition Recent leaks say Zen 6 desktop is on N2X only and is scheduled for later part of 2027. Note, N2X HVM is H2 2027, so we can expect Zen 6 desktops only in Q4 2027 at best (sad). This means, Nova Lake is gonna have a free reign for almost a year with literally *NO* competition. If NVL delivers even half of what's expected, it's gonna make a killing. The big question is, will Intel deliver? Some recent posts say, NVL has lower clocks (will not hit 6.0 GHz), but has higher IPC. I'm sure there's no Fmax regression with N2P HP cells as it's optimized for desktop clients. So in theory, the Fmax should be between 5.7 GHz to 6.0 GHz (although I think it's closer to 6.0 GHz rather than 5.7 GHz). So, even after adjusting for the lower clocks & slightly higher IPC, my old Jul-2025 prediction still stands. - M5 class ST - Very high MT performance - Excellent gaming performance Razor Lake Update Since Zen 6 desktops are launching late next year after N2X enters HVM, it's gonna come out a few months ahead of Intel Razor Lake. Razor Lake is reportedly using N2P (v2) UHP which is expected to have higher Fmax than original N2P HP, but compared to N2X, it's still slower. So, we can expect it to have around +5% Fmax gain at best. This means, Zen 6 will have at least +300 MHz higher (or more) boost clock compared to Razor Lake. Expected Clock Frequencies The revised clock frequency estimates of the desktop parts are as follows: - Nova Lake = 5.7 to 5.8 GHz - Razor Lake = 6.0 to 6.1 GHz - Zen 6 = 6.3 GHz to 6.4 GHz (maybe more) I expect Zen 6 and Razor Lake to have similar performance in spite of Zen 6 having roughly +300 MHz higher clock. Share your thoughts. Follow/Repost if you like. 👍 Image Credit: wccftech
13
25
4
160
12,849
Funny thing about Intel Unified Core is, it's now not a single core architecture. Intel now has 2 Unified Core architectures (x64 & X86S). And both Rosaic & Austin teams are working in parallel. Only one will survive. The battle is on!!! I prefer X86S. Many don't understand the importance of X86S and say x64 is good enough. No it's not. If Intel needs to go head to head with Apple Silicon, X86S is the way!! Time for baggage-ridden x64 to die and give way to superior lean & mean X86S.
Big News: Intel @ TomorrowX + Nova/Razor Lake + Unified Core - 18A yield at 70% - First 14A product in 2027 - 14A volume ramp in 2028 - 14A external customers 2028/29 - 14A volume ramp for external in 2028/29 (like TSMC A14) - 14A will have parity with A14 - New (external) Unified Core team 141 heads. Will spin in. X86S Unified Core with HT Amazing news is, my original Rosaic report was spot on!!! Headed by Top CPU Architect Peter Bannon (who worked with Jim Keller on Apple Silicon A4 & A5) is now working on the new Intel X86S Unified Core with Hyper Threading. Current team head count is at 141 as confirmed by Lip-Bu. This is Lip-Bu's wildest move ever. His secret weapon to leap frog competition in performance, PPW & PPA. Nova Lake Performance & Competition Recent leaks say Zen 6 desktop is on N2X only and is scheduled for later part of 2027. Note, N2X HVM is H2 2027, so we can expect Zen 6 desktops only in Q4 2027 at best (sad). This means, Nova Lake is gonna have a free reign for almost a year with literally *NO* competition. If NVL delivers even half of what's expected, it's gonna make a killing. The big question is, will Intel deliver? Some recent posts say, NVL has lower clocks (will not hit 6.0 GHz), but has higher IPC. I'm sure there's no Fmax regression with N2P HP cells as it's optimized for desktop clients. So in theory, the Fmax should be between 5.7 GHz to 6.0 GHz (although I think it's closer to 6.0 GHz rather than 5.7 GHz). So, even after adjusting for the lower clocks & slightly higher IPC, my old Jul-2025 prediction still stands. - M5 class ST - Very high MT performance - Excellent gaming performance Razor Lake Update Since Zen 6 desktops are launching late next year after N2X enters HVM, it's gonna come out a few months ahead of Intel Razor Lake. Razor Lake is reportedly using N2P (v2) UHP which is expected to have higher Fmax than original N2P HP, but compared to N2X, it's still slower. So, we can expect it to have around +5% Fmax gain at best. This means, Zen 6 will have at least +300 MHz higher (or more) boost clock compared to Razor Lake. Expected Clock Frequencies The revised clock frequency estimates of the desktop parts are as follows: - Nova Lake = 5.7 to 5.8 GHz - Razor Lake = 6.0 to 6.1 GHz - Zen 6 = 6.3 GHz to 6.4 GHz (maybe more) I expect Zen 6 and Razor Lake to have similar performance in spite of Zen 6 having roughly +300 MHz higher clock. Share your thoughts. Follow/Repost if you like. 👍 Image Credit: wccftech
6
11
2
96
6,668
Replying to @Alex_Intel_
Make Annual Reports Great Again
2
4
52
1,723
2025 retweeted
Elon Musk在《纽约时报》DealBook Summit通过视频连线接受主持人Andrew的采访。 47分钟的访谈涵盖了美国政治、黄仁勋、俄乌战争、Starlink、SpaceX与Tesla合并、自动驾驶,Anthropic。 几乎一分钟一个观点,现总结如下: 1. 明确表示自己不是MAGA阵营的一员。他认为如果MAGA意味着无条件支持共和党任何政策,那么他绝不属于这一阵营。 2. 强调自己的政治立场实际上相当温和。他不认为自己持有极端政治观点,并表示民主党和共和党都有优点和缺点。 3. 承认2026年中期选举的政治投入效果很差。他花费了数亿美元支持共和党候选人,但最终共和党在选举中惨败。他承认自己对此感到相当愚蠢,如果能够重新选择,会减少政治捐款。 4. 认为美国选举不是金钱能够决定的。巨额政治捐款并没有换来他预期的选举结果,也让他重新认识到资金对选举的实际影响力。 5. 主张两党相互制衡,而不是一党独揽权力。他解释,支持共和党在参议院保住多数席位的重要目的,是避免民主党未来同时控制白宫和国会两院。他认为分权和制衡对美国有益。 6. 不担心特朗普离任后的政治调查,也不认为自己需要总统赦免。对于民主党政客威胁调查甚至起诉他,他表示自己没有犯罪,不需要赦免;调查和惩罚应当以实际违法行为为依据,而不是政治立场。 7. 对民主党即将上任的得州州长Gina Hinojosa态度相对积极。他认为她可能是一位比较温和的州长,不属于极左派,因此不预期其执政会给Tesla和SpaceX在得州的经营带来严重问题。 8. 认为英伟达赶走黄仁勋是重大错误。他把黄仁勋比作巅峰时期的迈克尔·乔丹,认为这相当于一支正在连续获胜的篮球队,突然把最优秀的球员踢出球队。 9. 认为黄仁勋几乎不可替代。 黄仁勋不仅懂技术,而且与供应商建立了数十年的关系,许多优秀员工也是因为对他的忠诚而留在英伟达。马斯克认为,董事会低估了这些个人关系对公司的重要性。 10. 不认为黄仁勋背叛了美国。 尽管黄仁勋涉及私人生活和国家安全相关争议,马斯克并不认为他对美国不忠,反对因此否定他对英伟达的贡献。 11. 愿意聘请黄仁勋,但认为他并不是真的在找工作。面对主持人询问是否愿意招募黄仁勋,他表示当然愿意,因为黄仁勋是历史上最优秀的创业型CEO之一。不过他也表示,黄仁勋目前对离开英伟达仍感到震惊。 12. 认为自己对乌克兰的帮助被严重低估。他强调,SpaceX在俄罗斯全面入侵乌克兰前后提供Starlink通信服务,对乌克兰维持通信和军事行动至关重要。 13. 声称如果Starlink完全停止服务,乌克兰前线可能立即崩溃。 14. 坚持Starlink是民用系统,不应被用于进攻性军事打击。 他认为,普通Starlink不应该成为指挥无人机攻击、造成伤亡的军事系统。 15. 认为进攻性军事行动应当使用Starshield,由美国政府承担决策责任。他强调Starshield是与Starlink不同的军用系统,涉及杀伤性武器的决定应由美国军事指挥体系作出,而不是由他这个私人企业负责人决定。 16. 直接指责泽连斯基对Starshield通信限制问题撒谎。他否认乌克兰总统关于Starshield受到限制的说法,并认为乌克兰实际上是希望把民用Starlink用于进攻性军事行动。 17. 否认曾经主动切断乌克兰已经获得的Starlink通信服务。 他认为,外界对他曾阻止乌克兰袭击克里米亚塞瓦斯托波尔的描述存在误解,因为该地区当时本就受到制裁,相关服务并未开通。 18. 认为Starlink的使用规则不能因为政治压力而随意改变。 他特别提到Starlink设置了约80英里/小时的移动速度限制,并强调相关规则早已存在。 19. 认为俄乌战争的历史背景比外界描述的更复杂。他承认俄罗斯发动入侵属于侵略行为,同时又批评美国曾支持2014年乌克兰政权更迭。 20. 继续反对欧洲大规模非法移民。他认为这不是什么新的极端政治主张,并指出奥巴马和希拉里过去也曾表达过对非法移民的反对立场。 21. 反对将SpaceX的技术优势简单等同于垄断。对于加拿大总理Mark Carney等人提出的私人企业掌握关键基础设施是否过于危险的问题,他认为SpaceX的市场地位来自技术创新,而不是主动排斥竞争。 22. 认为SpaceX并没有阻止竞争对手复制其技术。 他表示SpaceX主要使用防御性专利,甚至欢迎其他公司研究、复制火箭技术;Tesla的专利也采取开放使用的政策。 23. SpaceX计划从2027年开始发射第二代手机直连卫星星座。他预计新一代系统的带宽将达到第一代的约40倍,这是SpaceX未来通信业务扩张的重要方向。 24. 计划通过屋顶微型基站解决卫星信号无法进入建筑物的问题。 Starlink手机直连在室内或金属屋顶下存在信号限制,因此SpaceX准备利用800MHz频段部署小型地面基站,计划于2027年量产。 25. 不认为Starlink的扩张必然导致AT&T和Verizon倒闭。 他认为未来AI超级智能、机器人以及机器之间的通信需求会以数量级增长,整个通信市场的需求空间可能远超今天。 26. 希望用Starlink通信业务产生的收入支持Starship的发展。商业通信业务可以为高成本的航天研发提供持续资金。 27. SpaceX的终极目标仍然是在月球和火星建立自给自足的城市。 他希望人类最终成为能够在多个星球生存的文明,而不是永远依赖地球。 28. 认为SpaceX和Tesla合并可能比继续独立运营更有价值。他表示两家公司不一定必须合并,但如果问合并后是否会更好,他的答案倾向于肯定。 29. AI、机器人和芯片研发是两家公司合并逻辑发生变化的根本原因。 五年前他看不到合并的必要性,但如今Tesla专注自动驾驶、Optimus等现实世界AI,SpaceX则拥有大模型、算力基础设施及芯片研发能力,两者交集越来越大。 30. 预测未来大部分超级智能计算能力可能部署在太空轨道。 他认为轨道计算将成为未来重要方向,这也进一步加强SpaceX与Tesla在AI领域协同的潜在价值。 31. 强调合并并未确定,最终必须获得双方股东批准。 他认为合并与不合并都有可能,特别是Tesla股东是否支持,并不能视为理所当然。 32. 反对通过公司合并直接触发自己的Tesla绩效股票奖励。 他表示自己的奖励应当取决于Tesla实现经营及估值目标,而不是仅仅因为与SpaceX合并就获得额外股份。 33. 进一步强调Tesla将把生产资源从Cybertruck转向Cybercab。Cybertruck计划在2027年年中逐步停产,腾出生产能力扩大无人驾驶出租车的制造规模。 34. 认为双座Cybercab足以满足绝大部分出行需求。他表示平均每辆车的乘客数量约为1.2人,两座设计能够覆盖大约90%的乘客行驶里程,因此具有明显的效率优势。 35. 首次明确提出开发没有方向盘和踏板的Model Y自动驾驶版本。 该车型计划容纳6至7名乘客,弥补双座Cybercab无法满足家庭和多人出行的不足,但他没有公布量产日期。 36. 对六座Model Y L的市场需求非常乐观。 他称Model Y L非常受欢迎,甚至表示可能把Model Y的生产线全部转向长轴距版本。 37. 对Anthropic团队的评价明显转向积极。他透露,在同意向Anthropic出租算力之前,自己亲自接触了很多Anthropic员工,认为他们有原则、勤奋,而且真正关心AI的未来。 38. 不再将Anthropic简单视为敌对公司。他甚至表示,所接触的Anthropic员工没有任何人触发他的“邪恶探测器”,并称自己相当喜欢这些人。 39. 确认Anthropic与SpaceX的算力租赁协议具有一定灵活性。 根据他在采访中的说法,相关协议可以提前约3至6个月通知终止,并非完全无法调整的长期锁定关系。 40. 认为Anthropic极有可能继续租用SpaceX的AI算力。 41. 透露Anthropic最可能再向SpaceX租用额外10GW的计算能力。这是此次访谈最值得关注的商业信息之一。 42. 公开认可Anthropic的AI模型技术实力。他特别称赞了Opus和Fable系列产品,肯定其在前沿AI领域的表现。 43. 认为自己现在的内心状态比过去更加平静。 当主持人回顾他此前曾把自己的大脑形容为一场风暴时,他回应说,风暴并非时时刻刻都在发生,目前自己的状态相对平静。
59
256
32
1,263
255,581
Anthropic IPO price $83 - it will be basically on par with Tesla at open... Now, if it goes to: $90 - valued same as Meta $104 - above TSMC $190 - above Google $282 - most valuable company in world - above Apple
27
31
10
271
133,689
2025 retweeted
Now that Intel, AMD, and Nvidia have locked up all the substrate capacity on the planet, other companies trying to build CPUs may struggle to get the substrates they need to make their chips.
IMO, Lip-Bu calling memory the biggest bottleneck suggests Intel has already secured enough substrate supply.
27
51
6
889
74,183
2025 retweeted
IMO, Lip-Bu calling memory the biggest bottleneck suggests Intel has already secured enough substrate supply.
28
35
3
564
108,888
$INTC CEO LIP-BU TAN ON CHANGE IN AI DEMAND AND SEMICONDUCTOR CYCLES: “Only PC side there is softness” “I do not see any demand weakness” “AI is bigger than internet. I think CPU demand going to continue” “Few issues we need to address. One is Cybersecurity. Other is buildout after midterms could be slower” “Semiconductor demand people talking about $2T. Nvidia, AMD and Intel will benefit and memory guys even better” “Supply shortage slowing down is good.” On market: “This is just a correction. We need to have more positive narrative.” On CAPEX continuing in 2028: “I think it will. May be in other pockets”
10
64
5,357
$INTC CEO LIP-BU TAN ON $AMD CEO LISA SU AND CPU PARITY: “My #1 priority to get back to CPU leader” “We have Diamond Rapids and Coral Rapids coming” “We are looking to catch up and eventually leap frog AMD” “Give me little bit time, we will get back”
12
14
1
170
14,576
*SPACEX CFO JOHNSEN: WE DO NOT SEE ANYTHING OTHER THAN INCREASED DEMAND ON THE COMPUTE SIDE $SPCX *SPACEX CFO JOHNSEN: WE'RE ABLE TO BE MORE OF A SPOT MARKET PLAYER AND GARNER A MUCH HIGHER RETURN ON OUR COMPUTE *SPACEX CFO JOHNSEN: I DO THINK ANTHROPIC WILL BE CHALLENGED TO IN-SOURCE ITS COMPUTE. IT'S NOT JUST FINDING A COUPLE GIGAWATTS, IT'S FINDING A LOT MORE THAN A COUPLE *SPACEX CFO JOHNSEN: WE ARE VERY WELL POSITIONED EVEN WITH OUR TOP CUSTOMER. THE REST OF THE INDUSTRY IS SCREAMING FOR THE SAME THING
SpaceX CFO Bret Johnsen on orbital compute timeline and cost: *JOHNSEN: WE'RE GONNA FLY OUR FIRST ORBITAL COMPUTE SATELLITES NEXT YEAR AND START RAMPING THAT CONSTELLATION IN '28 $SPCX *JOHNSEN: I THINK IN '29 YOU'RE GONNA SEE MEANINGFUL VOLUME IN THE GIGAWATTS COMING INTO THE MARKET FROM ORBITAL COMPUTE *JOHNSEN: IN THE NEXT YEAR OR TWO WE'LL BE AT THE TIPPING POINT OF BECOMING MORE COST EFFECTIVE THAN TERRESTRIAL *JOHNSEN: WE'RE ALREADY IN THE LOW TO MID $30 PER WATT, THE REST OF THE WORLD IS PROBABLY MORE THAN $50 TO $60
10
47
9
388
49,221
Replying to @Alex_Intel_
He announced that we have signed the first external customer on 18A
28
43
16
298
59,839
Intel just landed Microsoft as the first company to use its EMIB packaging, and its CEO laid out a recovery plan. Here's what you need to know. Microsoft and Intel announced an expanded collaboration for Azure. Microsoft is the first company to announce plans to use Intel Foundry's EMIB (Embedded Multi-die Interconnect Bridge) advanced packaging, for its own first-party Azure silicon. Advanced packaging is the step where multiple chips get connected inside one package. The deal also covers CPUs. Microsoft confirmed multiyear Xeon supply for Azure, including instances with the next-generation Diamond Rapids server processors. The two companies will also work together on CXL to improve memory efficiency, plus memory and confidential computing. Separately, CEO Lip-Bu Tan spoke at the TomorrowX Summit in an interview with Gavin Baker. He said Intel has signed its first outside customer for 18A. This is separate from the Microsoft EMIB news. Tan said 18A yield was "horrible" when he took over, and is now at 70%. He said a CEO came to his office asking for a huge amount of EMIB-T. On the roadmap, Tan said the first 14A product arrives internally around late 2027. Outside customers follow in 2028 and 2029. He said that is realistically about the same time as TSMC, maybe slightly behind. Tan also said his number one priority is getting Intel back to CPU leadership. He pointed to Diamond Rapids and Coral Rapids, and said Intel wants to catch up with AMD and eventually leap past it. Key numbers: - 18A yield: now 70% - First 14A product: 2027 - Outside 14A customers: 2028 to 2029 - Diamond Rapids: up to 256 cores, 1.6 TB/s memory bandwidth, 1.28 GB cache - INTC stock: about $54.51, down 1.2% Intel is pairing its first named EMIB customer and its first outside 18A customer with a 14A timeline close to TSMC's.
1
2
3
4
2,280
$INTC Intel CEO Lip-Bu Tan’s exact transcribed quote on $MSFT Microsoft as EMIB-T customer: “…we have the EMIB-T, we have the world best in the best technology. And in fact today, I think Microsoft just announced [our new product] for the data center, they are using EMIB-T, the technology for packaging.” 5 mins ago, Microsoft just did the press release: news.microsoft.com/source/20… “Microsoft is the first company to announce use of Intel Foundry’s EMIB™ (Embedded Multi-die Interconnect Bridge) advanced packaging technology for Azure first-party silicon. Bringing together Intel silicon and EMIB capabilities with Microsoft’s cloud engineering, the agreement confirms multi-year Intel® Xeon® supply capacity – including instances of next generation server CPUs, codenamed Diamond Rapids – and expands collaboration across the Azure roadmap and Intel-based”
$INTC INTEL CEO LIP-BU TAN: “Today, $MSFT Microsoft announced that they will use EMIB-T”
5
12
2
85
10,128
I agree with what @jeffsonnenfeld wrote in @TIME and no matter what Huang tweets, NVIDIA's board must not reinstate Jensen Huang simply because he is a brilliant founder, investors want their money back, or Silicon Valley's powerful elite are demanding it. Ethics mean nothing if we abandon them when the stakes are high. Consider what Jeff and his colleagues have documented: 1. Betrayal of trust. Jensen allegedly concealed personal relationships involving a major customer and a direct subordinate, and misled his own board. A CEO who cannot be trusted cannot lead. 2. Conflicts of interest. An undisclosed relationship with the head of a company purchasing billions in NVIDIA chips is a fundamental governance failure. Shareholders deserve better. 3. National security. Advanced NVIDIA chips reportedly reached restricted Chinese entities, including military-linked institutions. The article does not establish Jensen's personal involvement in illegal sales, but the questions about oversight and accountability are serious. 4. Financial self-interest. Some of the loudest voices demanding his return have enormous financial exposure to NVIDIA and its ecosystem. Their interests should not dictate corporate governance. 5. A culture of unchecked power. A company with roughly 60 direct reports to its CEO, no clear successor, and inadequate independent oversight has a serious governance problem. Jensen deserves credit for building one of the greatest companies in American history and his contributions are extraordinary. But extraordinary success does not confer moral immunity. I am disgusted by the chorus of technology leaders treating this as a popularity contest. What message are we sending to the next generation of entrepreneurs? That ethics are optional if you make enough money? That rules apply only to ordinary employees? The board must complete its investigation independently and act on the evidence, without intimidation or pressure. Integrity is tested when doing the right thing is costly. If Silicon Valley cannot uphold that principle, all its talk about values is hypocrisy.
Over the past week, I’ve been overwhelmed by the love and support of friends, industry colleagues, and people I've never met. Thank you to everyone who has shown kindness to my family and me. It means the world to me. I came to the United States 53 years ago, and becoming an American citizen is one of the greatest gifts and proudest moments of my life. NVIDIA has been my life’s work. Building it into the company it is today alongside my colleagues has been a privilege and an honor. Nowhere but in America could a company like NVIDIA be possible. I love our country, and I love NVIDIA. I did not live up to the standards expected of everyone at NVIDIA, and I have made serious mistakes in my personal life. I created problems for the Board that I deeply regret. I own my mistakes completely, and I am committed to cooperating with the Board’s independent investigation. Since my departure, I’ve seen an outpouring of calls from NVIDIA’s employees, investors, business partners, and customers for me to return as CEO. Many continue to reach out to me directly as the company navigates complex technology, market, competitive, and geopolitical dynamics. NVIDIA is a once-in-a-generation company. If the Board of Directors decides that my return is in the best interest of our shareholders, our employees, the company, and America, I would be honored to answer these calls and return as CEO. Thank you again to everyone who has shared their good wishes. I am profoundly humbled by your support. To my family: I love you, and I will spend the rest of my life being worthy of yours. Jensen
133
53
16
299
28,625