CPU重返舞台：2026年数据中心CPU格局

RLandAgentUsage,ContextMemoryStorage,DRAM PricingImpacts,CPUInterconnectEvolution,AMDVenice,Verano,Florence,IntelDiamondRapids,CoralRapids,ArmPhoenix+Venom,Graviton5,Axion 强化学习与代理使⽤、上下⽂记忆存储、DRAM价格影响、CPU互连演进、AMDVenice、Verano、Florence、IntelDiamondRapids、CoralRapids、ArmPhoenix+Venom、Graviton5、Axion 微信公众号 404K微信公众号 404KFEB10,2026 2026年2⽉10⽇∙PAID∙已⽀付 Since 2023, the datacenter story has been simple. GPUs and networking are king. The arrivaland subsequent explosion of AI Training and Inference have shifted compute demands awayfrom the CPU. This meant that Intel, the primary supplier of server CPUs, failed to ride thewave of datacenter buildout and spending. Server CPU revenue remained relatively stagnantas hyperscalers and neoclouds focused on GPUs and datacenter infrastructure. ⾃2023 年以来，数据中⼼的故事很简单。GPU 和⽹络成为王者。AI 训练和推理的到来及随之爆发的需求，使计算需求从 CPU 转移。这意味着作为服务器 CPU 主要供应商的英特尔，未能乘上数据中⼼建设和⽀出的浪潮。随着超⼤云和新云⼚商将注意⼒集中在 GPU 和数据中⼼基础设施上，服务器 CPU 收⼊保持相对停滞。 微信公众号 404K微信公众号 404K微信公众号 404K微信公众号 404KAt the same time, the same hyperscalers have been rolling their own ARM-based datacenterCPUs for their cloud computing services, closing off a significant addressable market for Intel.And within their own x86 turf, Intel’s lackluster execution and uncompetitive performance torival AMD has further eroded market share. Without a competent AI accelerator offering, Intel was left to tread water while the rest of the industry feasted. 与此同时，同样的超⼤云⼚商⼀直在为其云计算服务⾃⾏研发基于 ARM 的数据中⼼CPU，从⽽封锁了英特尔的⼀⼤可寻址市场。在其⾃⾝的 x86 领域内，英特尔平庸的执⾏⼒和在性能上不如竞争对⼿AMD 的表现，进⼀步侵蚀了其市场份额。由于缺乏有竞争⼒的 AI 加速器产品，英特尔只能原地踏步，⽽⾏业的其他部分则⼤快朵颐。 微信公众号 404K微信公众号 404K微信公众号 404K微信公众号 404KOver the last 6 months this has changed massively. We have posted multiple reports toCoreResearchand theTokenomics Modelabout soaring CPU demand. The primary drivers wehave shown and modeled are reinforcement learning and vibe coding’s incredible demand onCPUs. We have also covered major CPU cloud deals by multiple vendors with AI labs. Wealso have modeling of how many CPUs of what types are being deployed. 在过去的六个⽉⾥，情况发⽣了巨⼤变化。我们在核⼼研究和代币经济模型中发布了多份报告，指出 CPU 需求激增。我们展⽰并建模的主要驱动因素是强化学习和 vibe编码对 CPU 的巨⼤需求。我们还报道了多家⼚商与 AI 实验室达成的重要 CPU 云计算交易。同时我们也对正在部署的各种类型 CPU 的数量进⾏了建模。 微信公众号 404K微信公众号 404KIntel Q4’25 DCAI Revenue. Source: Intel However, Intel’s recent rallies and changing demand signals in the latter part of 2025 haveshown that CPUs are now relevant again. In their latest Q4 earnings, Intel saw an unexpecteduptick in datacenter CPU demand in late 2025 and are increasing 2026 capex guidance on foundry tools and prioritizing wafers to server from PC to alleviate supply constraints inserving this new demand. This marks an inflection point in the role of CPUs in the datacenter,with AI model training and inference using CPUs more intensively. 然⽽，英特尔在 2025 年后期的股价回升和需求信号的变化表明，CPU 现在再次变得重要。在其最新的第四季度财报中，英特尔在 2025 年底意外地看到数据中⼼CPU 需求回升，并上调了 2026 年在代⼯设备⽅⾯的资本⽀出指引，同时将晶圆从个⼈电脑优先调配到服务器，以缓解满⾜这类新需求的供应限制。这标志着 CPU 在数据中⼼⾓⾊的⼀个拐点，AI 模型训练和推理对 CPU 的使⽤变得更加密集。 2026 is an exciting year for the datacenter CPU, with many new generations launching thisyear from all vendors amid the boom in demand. As such, this piece serves to paint the CPUlandscape in 2026. We lay the groundwork, covering the history of the datacenter CPU andthe evolving demand drivers, with deep dives on datacenter CPU architecture changes fromIntel and AMD over the years.微信公众号 404K微信公众号 404K微信公众号 404K微信公众号 404K 2026 年对于数据中⼼CPU 来说是激动⼈⼼的⼀年，各⼚商在需求激增的背景下纷纷推出了多代新品。因此，本⽂旨在描绘 2026 年的 CPU 全景。我们奠定基础，回顾数据中⼼CPU 的历史及不断演变的需求驱动因素，并深⼊分析多年来 Intel 和 AMD 在数据中⼼CPU 架构上的变化。 We then focus on the 2026 CPUs, with comprehensive breakdowns on Intel’s ClearwaterForest, Diamond Rapids and AMD’s Venice and their interesting convergence (anddivergence) in design, discussing the performance differences and previewing our CPU costinganalysis. 微信公众号 404K微信公众号 404K微信公众号 404K微信公众号 404K接着我们聚焦 2026 年的 CPU，全⾯解析 Intel 的 Clearwater Forest、Diamond Rapids和 AMD 的 Venice，以及它们在设计上的有趣趋同（与分歧），讨论性能差异并预览我们的 CPU 成本分析。 Next, we detail the ARM competition, including NVIDIA’s Grace and Vera, Amazon’sGraviton line, Microsoft’s Cobalt, Google’s Axion CPU lines, Ampere Computing’s merchantARM silicon bid and their acquisition by Softbank, ARM’s own Phoenix CPU design and look at Huawei’s home grown Kunpeng CPU efforts. 随后，我们详述 ARM 阵营的竞争，包括 NVIDIA 的 Grace 和 Vera、Amazon 的Graviton 系列、Microsoft 的 Cobalt、Google 的 Axion CPU 系列、Ampere Computing的通⽤ARM 芯⽚竞标及其被软银收购、ARM⾃⾝的 Phoenix CPU 设计，并关注华为⾃研的鲲鹏（Kunpeng）CPU 努⼒。 微信公众号 404K微信公众号 404K微信公众号 404K微信公众号 404KFor our subscribers, we provide our datacenter CPU roadmap to 2028 and detail thedatacenter CPUs beyond 2026 from AMD, Intel, ARM and Qualcomm. We then look ahead towhat the future looks like for datacenter CPUs, discuss the effects of the DRAM shortage,what NVIDIA’s Bluefield-4 Context Memory Storage platform means for the future ofgeneral purpose CPUs, and the key trends to look out for in the CPU market and CPU designsgoing forward. 对于我们的订阅者，我们提供了直到 2028 年的数据中⼼CPU 路线图，并详细介绍了来⾃AMD、Intel、