#甲骨文 宣布引入 #CBRS Cerebras AI芯片，预计将于2027年上半年正式部署；除Cerebras外，甲骨文还宣布了一系列芯片供应商合作协议，涵盖英伟达、谷歌及英特尔支持的SambaNova Systems。在与英伟达的合作上，甲骨文表示将把部署规模扩展至"多吉瓦级别"，采用英伟达最新一代"Vera Rubin"AI服务器。

据ZDNet报道，三星和SK海力士正准备于2027年第二季度向英伟达Rubin Ultra供应8层HBM4，此举意味着原定采用12层HBM4E的方案或将生变。与此同时，据Wccftech报道，英伟达早在去年9月便已悄然下调Vera Rubin平台的HBM4带宽规格——每颗Rubin GPU的带宽削减约13%，HGX Rubin NVL8系统层面降幅则高达26%。上述调整对HBM供应商及GPU产业链均具直接影响。带宽门槛下调意味着更多HBM4产品有望通过英伟达资质认证，潜在供货来源扩大；而8层配置的引入则将降低每组HBM堆栈的成本与物料清单负担，在HBM供应持续偏紧的背景下，有望缓解产能瓶颈。#内存涨价

Sustainability Becomes a Priority in the Mining Equipment and Machinery Market https://growthmarketreports.com/report/mining-equipment-and-machinery-market-global-industry Environmental performance is becoming an important consideration in the Mining Equipment and Machinery Market. Mining companies face growing pressure to manage energy consumption, reduce emissions, improve resource utilization, and limit the environmental impact of extraction activities. These requirements are encouraging equipment manufacturers to explore machinery that combines operational performance with improved environmental characteristics. Equipment selection increasingly involves evaluating fuel consumption, maintenance needs, emissions, and overall lifecycle costs.

Fluorspar Market Draws Interest from Electronics and Advanced Materials Supply Chains Fluorine Chemistry Supports Specialized Manufacturing The Fluorspar Market is connected to the development of advanced manufacturing through its role as a feedstock for hydrofluoric acid and other fluorine-based chemicals. These chemical intermediates support several applications involving electronic materials, industrial processing, specialty coatings, and high-performance polymers. The mineral itself is not generally used directly in most electronic components. Instead, its importance comes from its contribution to the production of fluorochemicals that serve specialized industrial processes. https://growthmarketreports.com/report/fluorspar-market-global-industry-analysis

Silica Sand Market Responds to Evolving Construction Material Needs Demand for concrete, mortar, glass, and other building materials continues to influence the Silica Sand Market as construction practices and infrastructure requirements evolve. Role of Silica Sand in Construction Silica sand is used in multiple construction-related applications, including selected concrete mixes, mortar products, glass manufacturing, and specialty building materials. Its particle size, cleanliness, and consistency influence its suitability for different formulations. Construction companies and material manufacturers seek dependable supplies that meet technical requirements and support consistent production. The type of sand selected depends on the intended use, local standards, and the performance characteristics required in the finished product. Urbanization and Infrastructure Development Residential construction, commercial development, road projects, industrial facilities, and public infrastructure can all influence demand for sand-based materials. Urban expansion creates requirements for buildings, windows, facades, and supporting infrastructure, connecting silica sand demand to several parts of the construction supply chain. Infrastructure modernization can also support demand for materials used in repair, renovation, and replacement projects. These activities may create opportunities for suppliers serving construction material manufacturers and related industries. Product Specifications and Local Sourcing Construction applications often depend on consistent grading and material cleanliness. Suppliers must ensure that products meet the specifications established for their intended uses. Transportation is a significant consideration because silica sand is generally moved in bulk. Locally available deposits can provide logistical advantages when they meet quality requirements and comply with applicable extraction regulations. Proximity to concrete plants, glass manufacturers, and construction material facilities can influence supplier selection and delivery arrangements. Environmental and Regulatory Pressures Mining operations may face restrictions related to land use, water consumption, dust, biodiversity, and rehabilitation. Permitting delays or changes in extraction rules can affect the availability of material in specific regions. Responsible resource management is therefore an important part of long-term supply planning. Producers may adopt improved dust controls, water recycling, progressive rehabilitation, and more efficient material handling to address environmental obligations. Opportunities for Market Participants Suppliers can strengthen their competitiveness by maintaining dependable delivery schedules, consistent quality, and transparent product specifications. Partnerships with construction material manufacturers may also help align production with changing customer requirements. The Silica Sand Market remains connected to construction activity, but its development also depends on glassmaking, foundries, filtration, and other industrial uses. Producers that diversify their customer base and manage resources responsibly can respond more effectively to shifting demand patterns. https://growthmarketreports.com/report/silica-sand-market-asia-pacific-industry-analysis

#Anthropic 上市前加速囤算力，斥资450亿美元与Nscale签署六年算力大单，采用英伟达最新Vera Rubin芯片，涵盖460兆瓦电力供应，接手微软退出的数据中心 #AIDC 园区，并面向付费用户推出Chrome版Claude。

#英特尔 在Hot Chips 2026上正式展示下一代服务器处理器Diamond Rapids，以最高256个P核心和全新Fan-out Fabric架构为核心卖点，将产品定位于企业级AI Agent基础设施，直面AMD和Arm日益激烈的竞争。Diamond Rapids将采用英特尔18A-P制程打造，支持高达1.6TB/s内存带宽、128条PCIe Gen 6通道及CXL 3.0，较上代Xeon 6核心数提升50%以上。从物理构造来看，整颗芯片由22个独立裸片拼接而成，具体包括16个基于Intel 18A-P制程的核心小芯片（Core Chiplet）、2个基于Intel 3制程的Fabric Hub以及4个基于Intel 3-T制程的基础裸片（Base Tile），各部分通过Foveros Direct 3D混合键合及UCIe-S铜互联接口相连。这一设计思路与AMD EPYC Venice采用多CPU小芯片围绕大型I/O裸片的方案高度相似。英特尔将在 #AIDC 数据中心CPU市场与英伟达 Vera及AMD Venice展开正面竞争。

#英伟达 涨价计划从明年初出货的系统开始生效，影响范围包括搭载旗舰级Vera Rubin和Grace Blackwell芯片的系统。英伟达已经给微软、谷歌和甲骨文等大厂代工服务器的工厂发出涨价通知。#内存涨价

#英伟达 推出容量达102.4T的新一代以太网交换机Spectrum-6，专为Vera Rubin平台与十亿瓦级AI工厂打造。该产品获得微软、Tesla等头部企业采用，通过提升网络效率与能效，解决超大规模GPU集群的通信瓶颈。

Agent时代CPU路线之争：#英伟达 押注“单核更快”，#AMD 押注“并发更多”。英伟达Vera CPU搭载88个自研Olympus ARM架构核心，内存带宽1.2TB/s，片上互联带宽3.4TB/s。英伟达的核心论点是：AI Agent的工作方式是CPU与GPU反复交互——工具调用、代码执行、数据检索、任务编排——每一步都依赖上一步完成，因此CPU的单核速度和延迟直接决定整个Agent的响应效率。Vera的发布为行业引入了一个新的评估框架：“规模化下的最大单线程性能”。这与AMD长期坚持的chiplet多核堆叠路线形成直接对比。在AMD看来，生产级AI系统不是一个Agent在串行推进任务，而更像一个由数据库、API、向量存储、编排引擎、缓存和中间件组成的分布式软件平台。在这种场景下，瓶颈不是单任务的完成速度，而是固定功耗预算内能同时承载多少工作流。AMD给出了一组具体测算：在模拟的100千瓦机架部署场景中，EPYC 9965（Turin）的机架级吞吐量约为英伟达Vera基准的2.4倍；下一代EPYC 6（Venice）预计达到3.3倍。AMD的逻辑是：吞吐量密度更高，意味着同样的能耗能服务更多用户、处理更多请求——这才是 #云端AI 部署的真实成本函数。两套框架，两套KPI，本质上是两家公司在争夺行业叙事的主导权。市场最终会用哪个指标来采购服务器CPU？