AI 模型和平台
CoreWeave 激活丹麦AI能力并扩展全球网络背骨
在 Google 上的首选来源中添加 Unite.AI
CoreWeave 宣布于2026年10月27日扩张其全球AI云平台，使丹麦新的计算能力在线化，并将其全球AI网络骨干扩展到中东和亚太。
称公司在伦敦发布的新闻稿中，将规划中的22.5兆瓦的IT容量，该设施由与NVIDIA量子-X800 InfiniBand连接的NVIDIA Blackwell系统加速。
丹麦在北DEN01部署
DEN01将扩展 CoreWeave的欧洲基础设施足迹部署在包括英国、西班牙、瑞典和挪威在内的五个地理区的九个数据中心。
数据中心已签订热再利用协议，向区供暖网络提供剩余热量。 根据释放情况，这项安排减少了对替代供暖来源的依赖，并有助于丹麦的绿色过渡。
在北区首席执行官Eyjólfur Magnús Kristinsson中，他称表示，“与北方公司合作，CoreWeave获得了用于GPU密集型加工的DEN01的基础设施，同时通过电力购买协议支持可再生能源开发，并通过热量再利用帮助温暖的当地家庭，为绿色过渡做出贡献。”
北区总部位于冰岛雷克雅未克，在北欧地区设有7个数据中心。 该公司在芬兰的库沃拉、丹麦的乌尔戈德、瑞典的索洛费德奥和挪威的豪加兰开发了四个大型站点，另外在瑞典的斯德哥尔摩正在建造一个地铁站点。 该公司将其数据中心描述为利用可再生能源和支持循环经济原则。
全球AI网络背骨
除了新的计算能力之外，CoreWeave正在扩大全球AI网络的骨干，在阿拉伯联合酋长国、印度、印度尼西亚、新加坡、澳大利亚、韩国和日本设立了新的站点。 建设工程于2026年第四季度晚些时候开始，计划于2027年中期延长至全环。 将扩大CoreWeave现有的网络
CoreWeave 称正在建设第一个全球AI主干网，这是为Petabit规模的AI工作量而专门建造的，连接北美、欧洲、中东和亚太的地点。
该公司将骨干的技术设计描述为结合了每个存在点的多条400千兆塔每秒连接，基于硬件的MACsec加密以保护过境数据，以及旨在维持跨区域性能和可靠性的冗余网络路径。 CoreWeave 称这些能力意在让客户移动大型数据集和访问计算基础设施，而不牺牲其工作量所需的速度或安全性。
“我们的客户需要的不仅仅是获得计算能力。 他们需要能够可靠地运作规模的基础设施，需要能够将数据和工作量移动到他们需要的地方。” 他补充说，该公司正在扩大其全球基础设施和将它连接在一起的网络，因此客户可以专注于建设和改进AI，而不是管理其基本的复杂性。
公司规模和背景
CoreWeave最近在印度尼西亚和印度宣布，截至2026年6月30日，该公司在全球运营了51个数据中心，拥有超过1.5千兆瓦的主动功率，截至2026年8月11日，该公司拥有约4.2千兆瓦的合同功率。
CoreWeave 称它提供了创纪录的MLPerf 基准测试在推论和培训方面的结果，并且它是唯一一个连续三次获得半解集束MAX中最高白金排名的AI云。 这些业绩数字是公司自己报告的结果。
CoreWeave成立于2017年，并于2025年3月完成了在Nasdaq的公开上市。 公司将自己定位为一个专门设计的AI云平台，将计算能力、联网和软件服务结合起来，并将丹麦的部署和网络主干线设定为一个完全一体化的平台的一部分，目的是减少可减缓AI开发和部署的基础设施复杂性。
新站点计划于2026年第四季度晚些时候开始上线，全网环路计划于2027年中完成。
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
AI Models & Platforms
CoreWeave Activates Denmark AI Capacity and Extends Global Network Backbone
Add Unite.AI to your preferred sources on Google
CoreWeave announced on October 27, 2026 an expansion of its global AI cloud platform, bringing new computing capacity online in Denmark and extending its Global AI Network backbone into the Middle East and Asia Pacific.
In a news release issued from London, the company said it has brought the majority of its planned 22.5 megawatts of IT capacity online at atNorth’s DEN01 data center in Copenhagen, marking its first deployment in Denmark. The facility is accelerated by NVIDIA Blackwell systems connected with NVIDIA Quantum-X800 InfiniBand.
Denmark Deployment at atNorth DEN01
The DEN01 deployment expands CoreWeave’s European infrastructure footprint to nine data centers across five geographies, including the United Kingdom, Spain, Sweden, and Norway.
The data center has a heat reuse agreement in place to supply surplus heat to the district heating network. According to the release, the arrangement reduces reliance on alternative heating sources and contributes to Denmark’s green transition.
atNorth chief executive Eyjólfur Magnús Kristinsson described the partnership in terms of the site’s design. “By partnering with atNorth, CoreWeave gains access to DEN01’s infrastructure, designed for GPU-intensive processing, while supporting renewable energy development through power purchase agreements and helping warm local homes through heat reuse, contributing to the green transition,” he said.
atNorth is headquartered in Reykjavik, Iceland, and operates seven data centers across the Nordics. The company has four mega sites under development in Kouvola, Finland; Ølgod, Denmark; Sollefteå, Sweden; and Haugaland, Norway, plus an additional metro site under construction in Stockholm, Sweden. The company describes its data centers as running on renewable energy resources and supporting circular economy principles.
Global AI Network Backbone
Alongside the new compute capacity, CoreWeave is extending its global AI network backbone with new points of presence in the United Arab Emirates, India, Indonesia, Singapore, Australia, Korea, and Japan. The buildout begins later in the fourth quarter of 2026 and is planned to extend to a full loop in mid-2027. The expansion will extend CoreWeave’s existing network across North America and Europe into four continents.
CoreWeave said it is building the first global AI backbone network purpose-built for AI workloads at Petabit scale, connecting locations in North America, Europe, the Middle East, and Asia Pacific.
The company described the backbone’s technical design as combining multiple 400 gigabit-per-second connections at each point of presence, hardware-based MACsec encryption to protect data in transit, and redundant network paths designed to maintain performance and reliability across regions. CoreWeave said these capabilities are intended to let customers move large datasets and access computing infrastructure without sacrificing the speed or security their workloads require.
“Our customers need more than access to computing capacity. They need infrastructure that performs reliably at scale and the ability to move data and workloads wherever their business requires,” said Sachin Jain, chief operating officer of CoreWeave. He added that the company is expanding its global infrastructure and the network connecting it together so customers can focus on building and improving AI rather than managing the underlying complexity.
Company Scale and Background
The expansion follows CoreWeave’s recent announcements in Indonesia and India. The company operated 51 data centers worldwide with more than 1.5 gigawatts of active power as of June 30, 2026, and had approximately 4.2 gigawatts of contracted power as of its August 11, 2026 earnings call.
CoreWeave said it has delivered record-breaking MLPerf benchmark results in inference and training, and that it is the only AI cloud to earn the top Platinum ranking in SemiAnalysis ClusterMAX three times in a row. These performance figures are the company’s own reported results.
CoreWeave was established in 2017 and completed its public listing on Nasdaq in March 2025. The company positions itself as a purpose-built AI cloud platform combining computing capacity, networking, and software services, and frames the Denmark deployment and network backbone as part of a fully integrated platform intended to reduce the infrastructure complexity that can slow AI development and deployment.
The new points of presence are scheduled to begin coming online later in the fourth quarter of 2026, with the full network loop planned for completion in mid-2027.
Auto-fetched and translated by AI Atlas. All rights belong to the original author.