作者以霍华德·马克斯的泡沫框架解读美股调整与AI资本开支回报
作者转述霍华德·马克斯的观点称，当前市场尚未达到真正意义上的泡沫状态，因为市场持续讨论泡沫风险，投资者并未对风险“浑然不觉”；他同时认为投资者对进一步上涨存在错失恐惧，鲁莽行为信号则较为混杂。
#英伟达
X（推特）上的财经、科技与宏观账号
X·grokAI 评分5151 X·grokAI 评分2222 我被推荐，和看到太多聊AI风险的，说实话，这些风险我都认同，但是少有把风险和机遇放一起分析最后做出结论的，只聊风险并不会显得自己有多聪明，对未来需求的判断才是最难的，起码华尔街那些人23年GPT到现在…
我被推荐，和看到太多聊AI风险的，说实话，这些风险我都认同，但是少有把风险和机遇放一起分析最后做出结论的，只聊风险并不会显得自己有多聪明，对未来需求的判断才是最难的，起码华尔街那些人23年GPT到现在，预测英伟达的一年后的盈利预期从来没有对过。 https://t.co/b9Nm961XwB 引用 @TJ_Research：美股的泡沫分为估值泡沫和盈利泡沫 估值现在根本没有泡沫，无论静态还是动态。 盈利泡沫需要向前看，在你听到别人聊泡沫时，记得多问一句，你看到未来的盈利风险在哪里？对建立自己的观点会有帮助的
X·科技公司AI 评分77 Already running a model endpoint? What does it take to scale it on Kubernetes?
Already running a model endpoint? What does it take to scale it on Kubernetes? In five minutes, we walk through how NVIDIA Dynamo handles deployment with scalable workers and an OpenAI-compatible endpoint. Watch below: https://t.co/S8uc5ljUIW
X·财经博主AI 评分2424 This is important to crystallize.
This is important to crystallize. The underlying demand trend is incredibly strong and required. Whether that means shaky business models, bad deals, expensive financing, or scarcity pricing pencils is another debate altogether. Someone will be there to pick up the pieces. But 引用 @BenBajarin：All
X·财经博主AI 评分3434 JPM’s chart puts global AI compute demand at ~450 GW in 2030 and ~530 GW in 2031. Translating that…
JPM’s chart puts global AI compute demand at ~450 GW in 2030 and ~530 GW in 2031. Translating that into capital and revenue shows aggressive these assumptions seem to be. At an illustrative $50B per fully equipped GW (low-end by the way as costs are moving up), supplying 530 GW 引用 @pequityresearch
X·财经博主AI 评分2424 Thought it'd be helpful to frame this in economic output terms relative to history: Anthropic…
Thought it'd be helpful to frame this in economic output terms relative to history: Anthropic staff's survey answers imply US real GDP growth of ~18.6% in the 12 months to January 2030 (17.1% in the model's log terms). In the BEA annual record back to 1929, comparable growth https://t.co/j1RJcEaOKC
X·财经博主AI 评分7171 媒体报道称三星接受英伟达请求，由台积电为NVHBM制造base die
媒体报道称，三星已接受英伟达请求，由台积电为NVHBM制造base die；NVHBM是基于下一代HBM4E的英伟达定制HBM。三星加入SK海力士和美光，采用台积电3nm base die，以确保一致性。
X·grokAI 评分1111 这真的管用！ Grok 可以自负盈亏。
这真的管用！ Grok @Bot 可以自负盈亏。 https://t.co/GEZskbNU9h 【被引推文】我让我的 Grok Bot 去想办法自己挣钱。 我爱这个应用 https://t.co/g9SysaLINp 原文： This really works! Grok @Bot can pay for itself. https://t.co/GEZskbNU9h 引用 @mikepat711：I told my Grok Bot to go figure out how to pay for itself. I love this app https://t.co/g
X·宏观AI 评分1111 去看一下这周的ALL-IN，前面20分钟，JASON聊到老马发推“YIKES”，加入CHAMATH关于COREWEAVE的评论时，他发出的奸笑，我认为他肯定做空了。
去看一下这周的ALL-IN，前面20分钟，JASON聊到老马发推“YIKES”，加入CHAMATH关于COREWEAVE的评论时，他发出的奸笑，我认为他肯定做空了。 然后视频开始后他团队做的DECK聊加息对于COREWEAVE成本的影响，“因为一开始的UNSECURED
X·科技公司AI 评分3636 Building safer AI agents takes more than model safeguards.
Building safer AI agents takes more than model safeguards. @perplexity_ai open sourced SPACE, the agent sandbox platform behind Computer, giving developers more control over where agents run code and what they can access. Cheers to more teams open-sourcing tools to build safer 引用 @perplexity_ai：W
X·科技公司AI 评分7272 NVIDIA祝贺Reflection AI发布首个开放权重模型Beam
Reflection AI发布首个开放权重模型Beam，提供权重、tokenizer和参考实现，并以Apache 2.0许可证开放。该501B MoE模型使用NVIDIA GB300 NVL72训练，在Blackwell Ultra上开箱即用的推理速度超过70K tokens/sec/GPU；Beam 1.0提供BF16、FP8和NVFP4权重及DFlash2 drafter。
X·科技公司AI 评分3737 New from our STEM Fellows Program: Designing more efficient quantum circuits with Claude.
New from our STEM Fellows Program: Designing more efficient quantum circuits with Claude. Given plain-language instructions, Claude Mythos 5 designed quantum circuits that improve on the best published ones, sometimes by an order of magnitude or more. More efficient circuits
X·科技公司AI 评分4747 恭喜 @alexandr_wang 和 @Muse 团队发布开放权重模型 Muse Spark Open
@alexandr_wang 和 @Muse 团队发布开放权重模型 Muse Spark Open。
X·科技公司AI 评分2020 Big shoutout to all of our partners for making Nemotron 3.5 Super VL their own through…
Big shoutout to all of our partners for making Nemotron 3.5 Super VL their own through post-training 🙌 Here’s what they’ve built 🧵 https://t.co/pL3cEDEEpw
X·财经博主AI 评分5050 Shanu Mathew转述黄仁勋称英伟达每吉瓦内容价值超400亿美元
Shanu Mathew转述黄仁勋在最新一期 BG2 访谈中的说法，英伟达每吉瓦的内容价值超过400亿美元，前沿实验室约占其业务的25%–30%。基线计算约占三分之一，年增长率约为30%–40%；访谈还涉及AI收入是否支撑资本开支、2027年AI瓶颈及开放与封闭等话题。
X·财经博主AI 评分6161 Artificial Analysis 初步测试显示 Vera Rubin 每兆瓦峰值吞吐量较 GB300 高 14%–45%
Artificial Analysis 的首批 Vera Rubin 测试结果显示，其每兆瓦峰值吞吐量较 GB300 高 14%–45%，具体幅度因模型而异。在选定生成速度下，DeepSeek V4 Pro 约为 5.7 倍，GLM-5.2 约为 3 倍，Kimi K3 约为 2.8 倍。原文随后提到，即使按提升 30% 假设，一个 100MW GPU…
X·grokAI 评分1515 Orange Juice高兴宣布其CTO Elias Martinez Cohen，此前在Nvidia，有十多年AI经验。
Orange Juice高兴宣布其CTO Elias Martinez Cohen，此前在Nvidia，有十多年AI经验。 【被引】Elias Martinez Cohen加入ORANGE JUICE任首席技术官。 原文： Orange Juice is happy to announce its CTO, Elias Martinez Cohen, previously from Nvidia, and with over a decade working in AI. https://t.co/oBn810pKzB 引用 @orangejuice_btc：Elias Martinez
X·宏观AI 评分2121 老黄： $CRWV 的融资成本实际在下降
老黄： $CRWV 的融资成本实际在下降 下次请我做嘴替吧
X·宏观AI 评分5858 El-Erian转述FT：Meta、英伟达等将入驻的数据中心债券价格跌破面值
El-Erian转述《金融时报》称，为知名数据中心融资发行的债券近几周价格已跌至面值以下，相关中心将由Meta、英伟达、Oracle等公司入驻。原文还称，投资者正在抛售所谓的neocloud债务，并附有链接：https://t.co/08RzREOkUT
X·宏观AI 评分2828 Great read
Great read 引用 @arpitrage：I spent a couple of days at Anthropic learning from their folks about AI advances, and here's what I took away: • We are close to RSI (ie, models building other models) which may lead to rapid takeoff scenarios (with risks and benefits). Some evidence for this is code per
X·科技公司AI 评分00 Last claw stop before Berlin.
Last claw stop before Berlin. Good to see you, Stockholm. 🇸🇪🦞 https://t.co/d7KRcVKelM
X·科技公司精选AI 评分7676 NVIDIA祝贺StepFun发布Step 5 Preview开放权重模型
StepFun宣布Step 5 Preview开放权重，采用总参数量600B、每个token激活27B的稀疏MoE架构，支持1M-token上下文窗口和原生文本、视觉输入。
推荐理由：StepFun公布Step 5 Preview开放权重，并列出模型规模、上下文窗口与输入形式，NVIDIA则补充了GB300上的运行速度和NVFP4口径。
X·科技公司AI 评分4141 Nemotron 3.5 Super VL 将文本与视觉理解整合到一个模型
Nemotron 3.5 Super VL 将文本和视觉理解整合到一个模型中，减少在独立文本模型与视觉模型之间传递上下文的需要。აგენტ可能需要阅读文档、解读图表并搜索网页来完成任务。
X·科技公司AI 评分5757 英伟达推出开放模型 Nemotron 3.5 Super VL
英伟达推出开放模型 Nemotron 3.5 Super VL，将推理和工具使用与文本、图像、文档及视频理解结合，用于长时间运行的 AI agents。该模型总参数量为 120B，激活参数量为 12B。
X·科技公司AI 评分4343 英伟达支持的共享AI超级计算机 Sferical One 正式启动
Sferical One 这台由 NVIDIA 提供算力支持的共享 AI 超级计算机举行启动庆祝活动。黄仁勋与 Wallenberg Investments 的 Marcus Wallenberg、Spotify 的 @eldsjal 及 Sferical AI 的 Jenny Nordlöw 共同出席。
X·科技公司AI 评分6868 马斯克称SpaceX与英伟达联合开发的Star Mind计算机计划明年量产
马斯克称，SpaceX与英伟达共同设计的Star Mind计算机性能远好于多数人的预期，并计划明年进入量产。马斯克还表示，与英伟达工程团队合作令人愉快。
X·科技公司AI 评分3939 When the brand synonymous with car safety for almost 100 years teams up with AI, amazing things…
When the brand synonymous with car safety for almost 100 years teams up with AI, amazing things happen. 🚗 We’re expanding our partnership with Volvo Cars to advance safe automated driving: Volvo’s safety expertise will come together with the NVIDIA Hyperion, Halos OS and https://t.co/8pszLgH2tg
X·科技公司AI 评分6666 Meta发布Muse Spark Open，NVIDIA称其由GB300 NVL72训练
Meta发布开放权重模型Muse Spark Open，称其为迄今能力最强的美国开放权重模型，面向编程和智能体任务。模型总参数量为816B、激活参数量为42B，支持1M token上下文窗口；NVIDIA称该模型在NVIDIA GB300 NVL72上训练。Meta还制作了针对Blackwell Ultra优化的NVFP4检查点，原文随后提到“over 50K”，但未提供完整表述。
X·科技公司AI 评分4747 英伟达发布 Nemotron 3.5 Bolt，面向笔记本本地 AI 智能体
英伟达推出 Nemotron 3.5 Bolt，面向 RTX Spark 上运行的本地 AI 智能体。该模型参数量为 74B，可在 24GB 内存中运行。
X·科技公司AI 评分5757 英伟达扩展 Nemotron 开放模型家族，推出 Nemotron 3.5 Super VL
英伟达宣布扩展 Nemotron 开放模型家族，加入 Nemotron 3.5 Super VL。该模型结合视觉理解、推理与工具使用，面向 AI agents；引用内容称其支持文本、图像、文档和视频理解，总参数量为 120B、激活参数量为 12B。开放模型可供用户定制、部署和构建。
X·科技公司AI 评分4242 NVIDIA Nemotron 3.5 Bolt 24GB 可在 24GB 内存中运行
NVIDIA 推出 Nemotron 3.5 Bolt 24GB，称其可在 24GB 内存中运行。该模型包含 74B-A9B 参数，面向 RTX Spark 优化，可与其他应用同时运行。
X·科技公司AI 评分1717 AI has a lot to learn from the Nordics.
AI has a lot to learn from the Nordics. Tonight, Jensen celebrated the builders combining their expertise with NVIDIA technology to tackle challenges in manufacturing, medicine, and beyond. Thanks for an inspiring evening, Stockholm. 📍Next stop: Berlin. https://t.co/ydtjTOILgS
X·科技公司AI 评分1717 Working out how best to use AI in mathematics research—and in science more generally—will take…
Working out how best to use AI in mathematics research—and in science more generally—will take time, effort, and resources. We’re eager to support mathematicians as they drive scientific progress and define the future of their field. We’ve committed $5 million of funding, 引用 @SimonsFdn：How can the
X·科技公司AI 评分2222 NemoClaw was just named one of TIME’s Best Inventions of 2026! 🦞
NemoClaw was just named one of TIME’s Best Inventions of 2026! 🦞 So proud of the team and shoutout to everyone building safer AI agents with us. Blurb on @TIME: https://t.co/EY5S9htCn5 https://t.co/G3vo2N3RPP 引用 @TIME：These are creations changing how we live, work, play, and think about what’s p
X·科技公司AI 评分2121 Never seen anything grow as fast as !
Never seen anything grow as fast as @Grok @Bot! And there is no free tier, which means people are deriving real value in excess of what they pay. As the saying goes, you get what you pay for. 引用 @spectnfa：HOLY SH*T, JENSEN HUANG SAYS GROK BOT USAGE IS SKYROCKETING, AND ELON IS DOING SOMETHING AL
X·科技公司AI 评分6060 NVIDIA称FLUX 3 Image Klein 5.6B在GB200 NVL72上训练并可在RTX GPU运行
NVIDIA表示，FLUX 3 Image Klein 5.6B在NVIDIA GB200 NVL72上训练，并可在RTX GPU上运行。NVIDIA还提到该模型有NVFP4变体；引用的@bfl_ai称其为开放权重模型，拥有56亿参数，可在单张消费级GPU上运行，并支持原生4K、边界框和像素级编辑。
X·财经博主AI 评分4545 Rod Evans：AI数据中心仅60%–70%的电力用于有效AI工作
Nvidia副总裁Rod Evans在Vertiv的EMEA活动上表示，AI数据中心消耗的电力中，只有60%–70%用于有效AI工作。TechRadar称，他还拆解了一座用电负荷为1GW的数据中心的用电分配，但正文未提供后续明细。
X·财经博主AI 评分2525 Glad we finally got a word too from $NVDA Jensen Huang today from an event at Stockholm...
Glad we finally got a word too from $NVDA Jensen Huang today from an event at Stockholm... About $CRWV amid the broader Neocloud selloff. When asked about the markets reactions to Coreweave, Jensen responded: - "Coreweave's strength is improving every day, I'm optimistic" - https://t.co/Dt1uNlL
X·科技公司AI 评分1919 Hello, Stockholm. 🇸🇪
Hello, Stockholm. 🇸🇪 From the podcasts we discover to the robots working in factories, the Nordics are building with NVIDIA. Spotify is making podcast recommendations more efficient. ABB Robotics is helping engineers train and test robots in simulation. And startups are https://t.co/2OYTR0SoE6
X·科技公司AI 评分1313 Just ask Grok to buy what you want someplace else and get a better deal.
Just ask Grok @Bot to buy what you want someplace else and get a better deal. There are far more purchasing websites than just Amazon! 引用 @niccruzpatane：This happened when I tried to get my Grok Bot to order stuff for me on Amazon. “Continued access by an unauthorized AI agent violates Amazon's