跳到正文
11月18日周三
  1. elvis58

    elvis 转发 SGLang 峰会回顾,指出想了解 AI 模型与推理走向可查看该回顾。回顾称首届 SGLang Summit 为期两天、2000 多人参与,会上宣布 SGLang v1.0、发布 Miles v0.2,并成立由 NVIDIA、AMD、Intel、Google、Meta、thinkymachines、Cloudflare、modal、radixark 九家创始成员组成的 SGLang Consortium。SGLang 目前被 332 个组织使用,在 Meta 支撑服务数十亿用户的生产推理,Modal、Cloudflare、LinkedIn 等团队分享了生产经验,峰会包含 90 多场技术演讲和六场实操工作坊。

    引用SGLang@sgl_project

    2,000+ people. Two days. The first SGLang Summit. Together, we announced SGLang v1.0, released Miles v0.2, and launched the SGLang Consortium with nine founding members: @nvidia, @AMD, @intel, @Google, @Meta, @thinkymachines, @Cloudflare, @modal, and @radixark. SGLang is now used by 332 organizations. At Meta, it powers production inference for products serving billions of users. Teams from Modal, Cloudflare, LinkedIn, and others shared their production lessons. Across 90+ technical talks and six hands-on workshops, the community explored what comes next for inference, reinforcement learning, and AI agents. To everyone building, deploying, and contributing: thank you for making this possible. Explore the announcements, engineering lessons, and moments from SGLang Summit 2026 in the link below.

11月17日周二
  1. 量子位58

    奥特曼首谈黄仁勋离职:英雄也有缺陷,行业不会失速

    在11月16日Semafor硅谷峰会上,OpenAI CEO奥特曼首次公开谈及黄仁勋辞去英伟达CEO一事,称黄仁勋一直是他心目中的英雄,自己对黄仁勋的处境有强烈个人共情,但表示掌握的信息不足以判断他是否应重返英伟达。奥特曼认为英伟达体量庞大、担保能力仍受信任,AI需求旺盛,行业不会因黄仁勋离开而失速;他还澄清内存条言论被误解,并称OpenAI尚未启动完全自主的递归自我改进,上市不急于求成。

  2. IT之家58

    期权市场预计英伟达财报发布后市值或波动 3350 亿美元

    路透社报道称,英伟达将于北京时间 18 日凌晨公布第三季度财报,期权市场预期财报后公司市值波动幅度可能达到 3350 亿美元。萨斯奎汉纳国际集团数据显示,期权市场预计财报后首个交易日股价可能上涨或下跌约 8.5%,超过过去八个季度财报后平均波幅的两倍。LSEG 数据显示,英伟达市值从 10 月初约 5.8 万亿美元缩水至 13 日约 3.9 万亿美元,周一股价再跌 4%。

11月16日周一
  1. IT之家58

    英伟达机器人学习工程师戴士钦宣布辞职,称黄仁勋离职令其格外失望

    英伟达机器人学习工程师戴士钦(Darrell Dai)发文宣布离开公司,称黄仁勋突然离职让他格外失望。他提到黄仁勋在深度学习热潮前投入 CUDA、提前布局 AI,并在具身智能领域投资机器人技术、分享开源工具和项目;他还称黄仁勋会亲自阅读员工的 Top 5 邮件并直接交流。戴士钦表示,英伟达未来方向与企业文化将受到何种影响目前仍不清楚,黄仁勋的离开让他决定结束自己在英伟达的职业篇章。

11月14日周六
  1. Mira Murati58

    Thinking Machines 在 Connectionism 上详解了 Inkling 2 Small 的端到端 4-bit 训练方法,在 NVIDIA GB300 上从预训练第一步到最终 RL rollout 全程使用原生 4-bit 前向。Mira Murati 表示这是整套技术栈多项改进之一,将预训练时间从数周缩短到 68 小时。引用内容指出,用更少比特训练速度更快,但量化误差可能干扰学习。

    引用Thinking Machines@thinkymachines

    Training with fewer bits is faster, but quantization error can interfere with learning. Today on Connectionism, we detail how we trained Inkling 2 Small with a native 4-bit forward pass from the first step of pretraining through the final RL rollout. https://thinkingmachines.ai/blog/end-to-end-4-bit-training

  2. 宝玉58

    宝玉梳理 Will Depue 11 月 12 日推文与报告原文的差异,指出报告承认指纹只能说明初始权重如何创建、支持但不能证明 Blackwell 参与训练,5 亿美元等数字属作者自称没把握的估算,报告也承认证据不能说明芯片在哪、是否违反出口管制。

    引用will depue@willdepue

    I have a major result to share. I’ve found conclusive forensic evidence that the top 4 Chinese AI labs—DeepSeek, Moonshot, MiniMax and Ant Group—have built their newest models on NVIDIA’s most advanced Blackwell chips, which are barred from export to China. This sounds insane, but I discovered that almost every AI model carries a hidden signature of the GPU that trained it. When a model is first set up, the starting values of its weights are laid out in a pattern that depends on the exact GPU. Each combination of GPU and seed number produces a unique pattern—a sort of fingerprint of the hardware used to create the model. Traces of that pattern survive training into the published model. Again, this evidence isn’t based on any insider information or leaking. It’s based on a forensic analysis of the model weights released by the companies themselves. I searched more than six trillion combinations of chip layout and seed number and matched them to open-weights models. In doing so, I found conclusive evidence that six Chinese AI models—including DeepSeek-V4-Pro, Kimi K3, and MiniMax-M3—were initialized on NVIDIA Blackwell systems. Four of the models match Grace Blackwell, the GB200 and GB300 parts in NVIDIA’s flagship liquid-cooled rack-scale systems. I then verified these results on real hardware. Running unmodified PyTorch with the seeds I’d recovered, a Blackwell B300 reproduced DeepSeek-V4-Flash’s untouched starting weights bit for bit, and a rented GB200 reproduced Kimi K3’s to a correlation of 0.999995. A Hopper H100 reproduced none of DeepSeek-V4-Flash’s, and changing the seed by one broke the match. (DeepSeek’s seed is 953049583 by the way.) The method also was verified to match 24 known open-weights models to the hardware that developers said they used. It is exceedingly unlikely that these results are a coincidence. Across six trillion combinations, the best accidental match should reach only about 8 sigma (i.e., eight standard deviations above what random chance would produce). The real matches range from 49 to 1,385 sigma, and even the weakest single test hits 20. These were data-center-scale training runs. With some analysis of release schedules and model details, I estimate that the largest, Kimi K3 and DeepSeek-V4-Pro, used 2,000 to 3,000 Blackwell GPUs each. Clusters of this size cost on the order of $150 to $200 million each to buy, and across these labs alone would add up to more than half a billion dollars of export-controlled GPUs, not counting experimental compute or otherwise. It’s quite likely that we see multi-billion dollar expenditure on export-controlled GPUs. I am highly confident in this result, and it has been independently verified by senior researchers at OpenAI and Anthropic. You can read the full report at http://unemploymentlabs.com/checkpoint-forensics. There is much more I’ve identified about these models through other forms of checkpoint forensics like this, and will have more to share soon.

  3. Rohan Paul58

    Reuters: Senate Banking Chair Tim Scott asked Jensen Huang to testify on Dec. 3 and threatened a subpoena if he does not appear voluntarily. Commerce Secretary Howard Lutnick was asked to testify separately on Dec. 10 because his department issues and enforces chip export controls. The Justice Department says it has not yet determined whether she is a Chinese spy, so the espionage claim remains unproven.

    引用Rohan Paul@rohanpaul_ai

    Reuters: Senate Banking Chair Tim Scott asked Jensen Huang to testify on Dec. 3 and threatened a subpoena if he does not appear voluntarily. Commerce Secretary Howard Lutnick was asked to testify separately on Dec. 10 because his department issues and enforces chip export controls. The Justice Department says it has not yet determined whether she is a Chinese spy, so the espionage claim remains unproven.

11月13日周五
  1. IT之家58

    抄底资金涌入,英伟达总市值重回 4 万亿美元上方

    英伟达股价当地时间 11 月 12 日盘中反弹,收盘跌幅收窄至约 4.34%,报 166.19 美元/股,总市值重回 4 万亿美元上方。Vanda Research 估算散户在开盘后前半小时买入约 1.11 亿美元英伟达股票,创该机构有记录以来最高水平;富达投资客户截至中午提交近 39.5 万笔买入订单,接近卖出订单的 5 倍。此前英伟达股价一度大跌约 14%、市值蒸发近 6000 亿美元。

11月12日周四
  1. IT之家58

    外媒盘点英伟达潜在接班人:黄仁勋去留悬而未决

    IT之家援引外媒报道称,黄仁勋在英伟达的去留仍不明朗,预测市场 Kalshi 上他本周留任的概率已跌至 30%。报道称英伟达目前没有明确的继任计划,并列出苏姿丰、埃里克·施密特、陈立武、布雷特·泰勒、伊恩·巴克、安德鲁·贝尔、科莱特·克雷斯以及黄仁勋妻子劳丽·黄等潜在人选。

11月11日周三
11月10日周二
11月9日周一
  1. 钛媒体58

    黄仁勋风波背后:英伟达国家安全调查与国会权力更替

    作者认为黄仁勋 11 月 8 日全员信承认私德问题,实为回应 10 月 29 日《纽约时报》关于国家安全疑虑的报道,英伟达董事会已成立独立董事委员会并聘请外部法律顾问审查。文章梳理民主党中期选举拿下参众两院后,参议员沃伦有望执掌银行委员会并获得传唤权,认为黄仁勋从游说白宫的角色转为需应对国会调查的人。

11月7日周六
11月6日周五
  1. IT之家58

    Brightspeed 与英伟达洽谈将 130 多座电信建筑改造为小型数据中心

    互联网电信服务提供商 Brightspeed 正与英伟达洽谈,计划把在美国的约 130 座老旧电信建筑改造为小型数据中心。英伟达预计这些设施合计可提供约 1 吉瓦算力,容纳 2200 多个 Supermicro Vera Rubin NVL72 机柜,每个机柜含 72 颗 Rubin GPU 和 36 颗 Vera CPU。改造仍需部署冷却系统等配套基础设施,并需获得许可和地方政府批准。

11月3日周二
  1. 机器之心58

    从第一性原理看 AI 基础设施:算力的尽头为何是电力

    机器之心从第一性原理拆解 AI 基础设施,把有效智能写成可用电力、硬件能效、算法效率与利用率的乘积,指出瓶颈已从 GPU 迁移到变压器、燃气轮机和并网排队。文章引用 IEA 数据称,2025 年全球数据中心用电增长 17%,其中 AI 数据中心用电暴增 50%,而大型变压器交付周期在美国最长已达四年。

10月31日周六
  1. 钛媒体58

    芯片的大买主们,快没钱了?

    半导体产业纵横分析称,芯片大买家并未集体陷入资金枯竭,但持续扩大采购正考验各自筹资能力。Alphabet三季度经营现金流约423亿美元、资本支出约514亿美元,自由现金流为负约91亿美元;CoreWeave二季度经营现金净流入约6.79亿美元,设备支出达64.22亿美元。

10月30日周五
10月27日周二
10月25日周日
10月24日周六
10月23日周五
  1. MiniMax (official)58

    MiniMax 宣布 MiniMax M3.1 Flash 已获 SGLang 支持,从推理到 RL 均可 Day 0 投入使用。据引用的 SGLang 项目介绍,该集成由 MiniMax、NVIDIA、AMD 与 radixark 联合完成,在 bs1 decode 下达到 >1K tok/s,支持 NVIDIA Blackwell 与 Hopper(GB300/B300/B200/H200)及 AMD Instinct MI350X、MI355X,并可用 Miles 运行 RL。

    引用SGLang@sgl_project

    Space Bunny is here! 🐰 MiniMax M3.1 Flash hits >1K tok/s, bs1 decode with SGLang, production-ready from Day 0. This integration is a joint effort by @MiniMax_AI, @NVIDIAAI, @AMD, and @radixark, and you can also run RL with Miles! -Broad hardware support: NVIDIA Blackwell and Hopper (GB300/B300/B200/H200), plus AMD Instinct MI350X and MI355X. -Agentic serving at scale: ~100K total tok/s per GPU on the AgentX coding-agent trace with 96 concurrent sessions. -Speculative decoding: Supports MiniMax's DSpark draft and a DFlash2 draft trained by @radixark; supports HiCache and multimodal input. -Ultra-low-latency inference: MegaKernels deliver 2.15× faster single-request decode.

10月22日周四
  1. 钛媒体58

    黄仁勋押注Agent PC,英伟达RTX Spark能否成为Agent最佳宿主

    英伟达CEO黄仁勋在柏林GTC大会上称PC需要被重新发明,并展示与微软秘密合作四年的AI PC产品RTX Spark,该设备采用融合CPU与GPU的芯片设计,最高128GB统一内存和最高1 petaflop的FP4 AI算力。文章认为,业界正尝试用本地模型和AI PC解决Agent在隐私、安全和成本上的问题,但也有人提出盒子产品、虚拟主机等分布式模式可能更适合承载Agent。

10月21日周三