情

月之暗面引爆AI交易变局，盘点新格局下的赢家与输家_我的网站

一 |     LOS ANGELES -- Talks have broken off between Hollywood actors and studios, killing any hopes that the strike by performers was coming to an end after nearly three months, as the writers strike recently did. The studios announced that they had suspended contract negotiations late Wednesday night, saying the gap between the two sides was too great to make continuing worth it. On Oct. 2, for the first time since the strike began July 14, the Screen Actors Guild-American Federation of Television and Radio Artists had resumed negotiations with the Alliance of Motion Picture and Television Producers, which represents studios, streaming services and production companies in strike talks. When negotiations resumed with writers last month, their strike ended five days later, but similar progress was not made with the actors union. The studios walked away from talks after seeing the actors' most recent proposal on Wednesday. “It is clear that the gap between the AMPTP and SAG-AFTRA is too great, and conversations are no longer moving us in a productive direction,” the AMPTP said in a statement. The SAG-AFTRA proposal would cost companies an additional $800 million a year and create “an untenable economic burden,” the statement said. Representatives from the actors union did not immediately respond to requests for comment. Actors have been on strike over issues including increases in pay for streaming programming and control of the use of their images generated by artificial intelligence. The AMPTP insisted its offers had been as generous as the deals that brought an end to the writers strike and brought a new contract to the directors guild earlier this year. From the start, the actors talks had nothing like the momentum that spurred marathon night-and-weekend sessions in the writers strike and brought that work stoppage to an end. Actors and studios had taken several days off after resuming, and there were no reports of meaningful progress despite direct involvement from the heads of studios including Disney and Netflix as there had been in the writers strike. Members of the Writers Guild of America voted almost unanimously to ratify their new contract on Monday. Their leaders touted their deal as achieving most of what they had sought when they went on strike nearly five months earlier. They declared their strike over, and sent writers back to work, on Sept. 26. Late night talk shows returned to the air within a week, and other shows including “Saturday Night Live” will soon follow. But with no actors, production on scripted shows and movies will stay on pause indefinitely.。    月之暗面发布的Kimi K3模型引爆“DeepSeek效应”，在全球科技股市场掀起新一轮重新定价浪潮，迫使投资者审视支撑AI交易逻辑的核心假设。         月之暗面上周五发布新一代模型Kimi K3，称其性能可与OpenAI和Anthropic的顶级产品媲美，成本却仅为后者的一小部分。市场开始将此次发布与去年的“DeepSeek时刻”相提并论，IG市场分析师Tony Sycamore表示，这一冲击导致两家未上市美国公司OpenAI和Anthropic的预期估值合计蒸发3140亿美元。

二 |          评论认为，Kimi K3的问世从两个维度动摇了既有市场逻辑：其一，美国限制先进芯片出口将阻碍别国AI进展的判断面临挑战；其二，性能更强、成本更低的模型在加速全球AI应用普及的同时，是否会引发AI基础设施产能过剩，亦成为投资者的新忧虑。         然而，此次冲击并非全面利空。从市场表现来看，不同赛道的玩家正在经历截然不同的命运分化。以下是潜在赢家与输家的梳理。         赢家之一：中国半导体生态系统          目前尚不清楚月之暗面开发K3所使用的具体芯片，不过，新模型的发布已直接提振中国半导体板块。本周一盘中，中芯国际A股一度涨超5%，市场预期AI模型开发商之间的竞争加剧将推动相关基础设施支出持续增加。         Allspring Global Investments投资组合经理Gary Tan表示：“我们预计最大的赢家仍将是AI基础设施层。在政府的大力支持下，中国对开源AI的推动可能加速国内先进AI模型的部署，这将需要更多算力资源，并带动底层硬件需求，尤其是网络设备和存储芯片。”          上周末，阿里巴巴发布了旗舰模型Qwen3.8 Max的预览版，该模型参数量达2.4万亿，性能据称可与全球领先的前沿模型比肩。两款产品的相继发布，叠加中国AI开发商日益倾向于针对国内芯片优化模型的趋势，共同推动科创板代表性指数创业板综周一涨幅一度达3.6%。         赢家之二：存储芯片厂商          Kimi K3拥有2.8万亿参数，系中国迄今最大规模的AI模型，其稀疏比——衡量计算效率的关键指标——已创下历史新高。

三 | 更高的稀疏比意味着每次推理任务中被激活的参数占总量的比例更低，从而降低算力消耗。         然而，这2.8万亿参数仍须全部存入内存。

四 | 即便经过低精度格式压缩，Kimi K3仍占用约1.4太字节的内存空间。这意味着部署该模型须依赖配备大容量内存的AI处理器集群，例如英伟达的Blackwell GB300系统，对SK海力士、三星电子等存储芯片供应商的需求由此得以延续。         Sumitomo Mitsui DS Asset Management高级投资组合经理Stanley Tang表示：“由于供应商数量有限，存储芯片厂商仍将是半导体板块中处境最为有利的一类企业。”他进一步指出，如果像Kimi这样的模型加速普及，整体需求不太可能因此下滑，反而可能推动AI智能体加速渗透，从而进一步拉动存储需求。         赢家之三：AI智能体与软件开发商          Kimi K3的发布还提振了外界对中国软件企业与AI智能体开发商的信心——即便在获取全球最先进芯片受限的背景下，这些企业依然有能力持续提升模型性能与效率。

五 |          “可以确定的是，一款性能出色且成本更低的开源模型将降低应用开发成本，并可能加快在编程、客户服务和工业流程领域的应用普及，”Saxo Markets首席投资策略师Charu Chanana表示。         不过，晨星分析师Malik Ahmed Khan对于软件开发商所面临的竞争风险持更为审慎的态度。         Khan认为，将上周五Alphabet、亚马逊和微软遭抛售归因于市场担忧更多企业将转向受K3启发的开源模型，这种判断“站不住脚”。

六 | 他说，我们认为，美国企业、政府机构以及与美国政府合作的公司，不会为了降低推理成本而采用别国的开放权重模型。

七 |          输家之一：AI模型开发商          如果说K3催生了新的赢家，那么对于AI模型开发商而言，竞争格局已明显趋于恶化。         在K3发布前被普遍视为中国最佳开源模型开发商的智谱，港股股价在过去两个交易日内已累计下挫近40%，折射出市场对于模型层竞争白热化的担忧。

八 |          此后竞争亦不会就此平息。阿里巴巴周一股价随Qwen3.8 Max的发布而上涨，该公司将这款模型定位为能力仅次于Anthropic Fable 5的产品。

九 | MiniMax Group Inc.据报也计划推出M3模型的升级版本。月之暗面本身则宣布将于7月27日公开发布Kimi K3的模型权重，届时企业可自行运行该模型，而无需依赖其云服务。         Sumitomo的Tang表示：“我们仍处于早期阶段，更好的模型还会不断涌现。

十 | 一些模型可能变得毫无价值，我认为市场尚未充分计入这一风险。”          输家之二：高端芯片制造商          此次发布同样令市场重新审视英伟达和AMD等AI热潮核心受益者是否还能持续享有稀缺性溢价。

十一 | 与DeepSeek冲击不同的是，K3的突破点不在于训练效率，而更多体现在以更低成本实现对标顶级模型的推理性能——这一路径对高端算力的依赖程度存在本质差异。         Siebert Financial首席投资官Mark Malek表示：“东京和台北的投资者当天将对AI资本支出最终能否获得回报的疑虑计入价格。无论世界上哪家实验室——不管来自哪个国家——每次推出一款免费的前沿级模型，这种疑虑就会再次浮现。”          从更宏观的视角来看，Kimi K3与Qwen3.8 Max的接连亮相表明，前沿AI模型的竞争正在超越OpenAI、Anthropic等美国头部实验室的范畴持续扩展。据彭博报道，这一趋势可能对美国模型提供商的定价权形成压力，同时继续推动全球范围内支撑更强大AI系统运行所需的基础设施投资。AI赛道的赢家与输家格局，或正在经历深刻重构。

Current article:http://www.shangfochengcongshuaizhuang.buzz/list_ml3e4oo/nui.html

Published on:12:34:18