微软 AI 战略解构：从能源到代币的全局布局

"TheBigPause",AITokensFactoryEconomicsStack,OpenAI,NeocloudRenting,GitHCopilot,MAIandMaia "⼤暂停"、AI通证⼯⼚经济学架构、OpenAI、新云租赁、GitHubCopilot、MAI与玛雅 ,,,AND4OTHERSDYLAN PATEL JEREMIE ELIAHOUONTIVEROS MYRONXIE 迪伦·帕特尔、杰雷⽶·埃利亚胡·昂蒂⻙罗斯、迈伦·谢等四位作者微信公众号 404K NOV13,2025 2025年11⽉13⽇∙PAID∙付费内容 Share分享 Microsoft was at the top of AI in 2023 and 2024, but then a year ago they changedcourse drastically. They paused their datacenter construction significantly and slodown their commitments to OpenAI. We called this outa year ago to datacenter mclientsandlater wrote a newsletter piece about it. 微软在2023⾄2024年间曾是⼈⼯智能领域的领跑者，但⼀年前却突然急转⽅向。⼤幅暂停数据中⼼建设，并降低了对OpenAI的投⼊承诺。我们⼀年前就向数据中型客户发出预警，随后还在通讯专栏中专门撰⽂分析此事。 Microsoft'sDatacenterFreeze-1.5GWSelf-BuildSlowdown&LeaCancellationMisconceptions 微软数据中⼼冻结⸺1.5吉瓦⾃建项⽬放缓与租约取消误解 DYLAN PATEL,JEREMIE ELIAHOUONTIVEROS,ANDMAYA BARKIN 迪伦·帕特尔(DYLAN PATEL)、杰雷⽶·埃利亚胡·昂蒂⻙罗斯(JEREMIE ELIAHOUONTIVEROS)和玛雅·巴尔⾦(MAYABARKIN)·微信公众号 404K微信公众号 404K Readfullstory阅读全⽂ 2025 was the story of OpenAI diversifying away from Microsoft, with Oracle,CoreWeave, Nscale, SB Energy, Amazon, and Google all signing large computecontracts with OpenAI directly. But now Microsoft’s investments in AI are back, a the AI giant has never had such high demand for Accelerated Computing. TheRedmond titan has woken up to it going down the wrong path and has dramaticallshifted course. With thenewly announced OpenAI deal,Azure growth is set toAcceleratein the upcoming quarters as forecasted by ourTokenomics model. 2025年的故事线本是OpenAI寻求微软之外的多元化合作，甲⾻⽂、CoreWeave、Nscale、SB Energy、亚马逊和⾕歌陆续与OpenAI直接签署了⼤规模算⼒合约。但今微软重燃AI投资热情，这家AI巨头对加速计算的需求从未如此强烈。雷德蒙德已然意识到此前路线偏差，正全⼒扭转航向。根据我们通证经济学模型的预测，随OpenAI新协议的签订，Azure增长势头将在未来⼏个季度显著提速。 Microsoft plays in every single part of theAI Token Economic Stack, is witnessingaccelerated growth, and we expect the trend to continue in coming quarters and y微软在AI通证经济栈的每个层级都深度布局，当前增长势头持续加快，我们预计这趋势将在未来数季度乃⾄数年内延续。微信公众号 404K微信公众号 404K The firm is actively looking for near-term capacity and pulling the trigger oneverything it can get its hands on. Self-build, leasing, Neocloud, middle-of-nowherlocations – everything is on the table to accelerate near-term capacity growth (exacnumbers available to ourDatacenter Model subscribers). 该公司正积极寻求短期扩容⽅案，并全⼒获取⼀切可⽤资源——⽆论是⾃建数据中租赁设施、采⽤Neocloud⽅案，还是偏远地区选址，所有能加速短期容量增长的选都在考虑之列（具体数据可参阅我们的数据中⼼模型订阅服务）。 On the hardware side, Microsoft even has access toOpenAI’s Custom Chip IP,thmost exciting custom chip ASICs currently in development. Given the trajectory oOpenAI’s ASIC developments look much better than that of Microsoft Maia, it mathat Microsoft ends up using the chip to serve OpenAI models. This situation issimilar to Microsoft’s situation with OpenAI models. While they have access toOpenAI models: they are still trying to train their own foundation model withMicrosoft AI. Like the models, access to the chip roadmap is not persistent giventhose IP rights do not go into perpetuity. Being reliant on the OpenAI’s ASIC doeachieve the goal of being self sufficient in hardware that is a common goal sharedamong hyperscaler ASIC programs. 硬件层⾯，微软甚⾄拥有OpenAI定制芯⽚IP的使⽤权，这是当前研发中最令⼈瞩的定制ASIC芯⽚。鉴于OpenAI的ASIC发展轨迹明显优于微软Maia芯⽚，最终软可能会使⽤这款芯⽚来⽀持OpenAI模型。这⼀局⾯与微软处理OpenAI模型的如出⼀辙：虽然能使⽤OpenAI模型，微软仍在尝试⽤Microsoft AI训练⾃⼰的基模型。同理，由于知识产权并⾮永久授权，微软对OpenAI芯⽚技术路线图的访问⾮长期稳定。依赖OpenAI的ASIC⽆法实现超⼤规模云服务商ASIC项⽬共同追求硬件⾃主⽬标。微信公众号 404K微信公众号 404K OpenAI ChipTeamIsNowSeriousOpenAI芯⽚团队现已全⼒投⼊ DYLAN PATEL迪伦·帕特尔·2024年6⽉3⽇Readfullstory阅读全⽂ Much of this is motivated by a drastic simplification of the OpenAI ownershipstructure as well as Microsoft’s laser focus on serving the infrastructure needed foconverting model capabilities into product use cases (and revenue) via stateless AP这⼀举措很⼤程度上源于OpenAI股权结构的极⼤简化，以及微软将核⼼战略聚焦过⽆状态API，为模型能⼒转化为实际产品⽤例（及收益）提供必需的基础设施⽀ In this report, we will dive into all aspects of Microsoft’s AI business. We begin byreviewing the history of the OpenAI relationship, covering the historic surge inMicrosoft’s datacenter investments in 2023-24, as well as the mooning scale of theOpenAI training clusters – from tens of MWs to Gigawatts. 本报告中，我们将深⼊剖析微软AI业务的各个维度。⾸先回顾微软与OpenAI的合渊源，涵盖2023-24年间微软数据中⼼投资的激增历程，以及其OpenAI训练集群模跃迁——从数⼗兆⽡级跃升⾄吉⽡量级。 We then analyze every part of Microsoft’s positioning in the AI Tokens EconomicStack:微信公众号 404K微信公众号 404K 随后我们将逐层解构微软在AI代币经济栈中的战略布局： Applications应⽤场景 LLMsPaaS平台即服务(PaaS)IaaS基础设施即服务(IaaS)Chips芯⽚ In each section, we’ll dive deep into Microsoft’s product portfolio, competitivepositioning and outlook. It’s not all good news for Microsoft as the software behemfaces a slew of new entrants and challengers to their dominant productivity suite aAI compute platform.微信公众号 404K微信公众号 404K 在每个章节中，我们将深⼊剖析微软的产品组合、竞争优势及前景展望。对这家软件头⽽⾔并⾮全是好消息——⾯对⽣产⼒套件和AI计算平台领域涌现的⼤量新⼊局者挑战者，微软的统治地位正受到冲击。 Microsoft&OpenAIin2023-25:FromAll-InonAItotheBig“Pause”微软与OpenAI的2023-2025：从AllinAI到"⼤暂停"时代 2023-24:selfbuild,leasing,andbuildingtheworldʼslargestdatacentersforOpenAI2023-2024年：⾃建、租赁并举，为OpenAI打造全球最⼤数据中⼼集群微信公众号 404K微信公众号 404K ChatG