广告
加载中

机场大屏上的AI云暗战:新云 正迎来它的时代

杨丽 2026-09-02 09:47
杨丽 2026/09/02 09:47

邦小白快读

EN
全文速览

本文核心介绍了AI新云赛道兴起的产业现状,核心干货信息如下:

1. 行业营销端:2026年1-4月国内云服务行业户外投放规模达3.52亿元,较上年同期近乎翻倍,84.7%投放集中在一线枢纽机场。机场是云厂商抢占B端决策人心智的核心场景,受众都是企业数字化、AI转型的核心决策层,广告主打心智占位而非直接转化。

2. 行业格局:当前AI云赛道分为三类玩家,分别是做守垒的通用超云、抢夺入口的MaaS与Agent派、差异化定位的AI新云,三层结构并行发展。AI新云是专为生成式AI与Agent工作负载设计的原生云,架构从底层开始适配AI需求,和通用云长期并存,当前已经成长为AI时代基础设施的主战场之一,市场增速远高于传统云。

本文从机场AI云广告大战中,总结出B端品牌营销的逻辑和AI云行业的最新趋势,干货如下:

1. B端品牌营销逻辑:一线枢纽机场精准覆盖企业CEO、CTO、CIO等AI转型核心决策人群,具备环境封闭、受众停留时间长、决策影响链短的优势,非常适合高端B端品牌做心智占位,不仅能影响客户决策,还可以提振自身销售团队的士气,2026年云服务户外投放同比近乎翻倍,超八成落在机场,验证了该场景的价值。

2. 行业消费趋势:当前企业AI已经从测试阶段全面进入生产阶段,对专业化AI算力的需求快速爆发,催生了AI新云独立赛道,新云凭借AI原生架构、中立定位获得市场认可,行业格局从单一通用云转向通用云与新云长期并存的分层结构,新云市场规模即将迈入千亿级,年增速远超传统云,增长空间广阔。

本文披露了AI云赛道的最新变化,以及面向云服务从业者的机会、风险与可借鉴经验,干货如下:

1. 市场与需求变化:AI新云已经从边缘角色成长为行业增长主力,机构预测中国AI基础设施云2028年规模将突破4100亿元,年复合增长率24.6%;美国新云市场2030年将达到1800亿美元,年均增速69%,增长空间极大。当前企业AI已经从测试进入生产阶段,高频推理、Agent执行、行业模型微调等需求爆发,通用云无法适配这类需求,存在明确市场缺口。

2. 机会与风险提示:新云不需要替代通用云,走差异化互补路线即可,抓住企业通用云做底座+新云做AI执行层的分层采购趋势即可。风险方面,需要警惕通用云自研AI芯片对GPU租赁板块的挤压,可通过强化AI原生架构、中立异构适配能力建立自身壁垒,商业模式上可探索按任务产出计价的新模式。

本文对制造工厂推进数字化转型、布局AI应用有诸多启示,核心干货如下:

1. AI转型的产品与需求层面:当前AI已经全面进入生产应用阶段,工厂落地AI优化生产、研发、管理等环节,都需要稳定高效的AI算力支持,传统通用云架构以通用负载为核心,AI工作负载调度效率低、GPU利用率不足,无法很好满足工厂AI生产级需求,专门优化的AI新云可以提供更高效率、更透明计价的AI算力服务。

2. 商业机会与转型启示:国内AI新云原生支持国产芯片混布、数据合规等刚需,适配国内工厂数字化转型的政策要求,出海工厂也可以借助新云灵活的全球调度能力,适配不同区域的合规要求。转型不需要推翻原有架构,采用通用云做底座+新云做AI执行层的组合模式,就能兼顾成本与效率,适配工厂不同类型的业务负载。

本文梳理了AI基础设施领域的最新行业趋势、客户痛点与可行解决方案,对AI、云相关服务商干货如下:

1. 行业发展趋势:当前AI云已经形成底层算力、中间MaaS、上层Agent入口三层并行结构,AI新云已经从通用云的补充发展为独立的产业层,成为行业增长的核心动力之一,市场规模正从百亿级迈向千亿级,年复合增速远超传统云,未来AI原生调度、按效计价会成为行业标配。

2. 客户痛点与解决方案:当前企业客户的核心痛点是,原有通用云架构无法适配生产级AI负载,存在调度效率低、GPU利用率低、计价模式不匹配,同时需要中立可退出的算力选项、国产异构适配、合规支持,通用云无法完全满足。服务商可参考的路线是,做AI原生架构,从底层适配AI负载,支持异构芯片统一调度,采用按任务产出计价,定位中立AI基础设施层,和通用云互补共存。

本文对云平台、智算平台类平台商的战略规划、运营管理有诸多参考,核心干货如下:

1. 当前市场对AI云平台的核心需求:企业AI进入生产阶段后,需要更可预测的AI容量、更高的GPU利用率、更透明的计价模式,同时需要中立可退出的算力选项,还要适配国产异构芯片、满足不同区域的合规要求,这些都是原有通用云平台无法完全满足的需求。

2. 平台可借鉴的做法与风险规避:平台可聚焦AI负载重构技术架构,做AI原生调度体系,拆分训练和推理生产线,采用按有效产出、按Token交付的计价模式,定位差异化和通用云形成互补,不需要追求全品类覆盖。风险方面,需要警惕通用云自研AI芯片带来的市场挤压,要强化自身在AI原生调度、中立性、异构适配方面的壁垒,抓住区域化算力布局的风口,积极拓展海内外新兴市场,抓住结构性增长机会。

本文梳理了AI云产业的最新动向,披露了核心数据与行业新特征,对产业研究者核心干货如下:

1. 产业新动向:AI云已经形成三层并行的产业结构,AI新云已经从通用云体系中独立出来,成为单独的产业层级,彻底改变了过去企业“选一朵云”的逻辑,形成了“通用云做底座+新云做AI执行层+NeoPaaS做治理层”的组合采购新范式,新云已经从新兴选项变成企业云服务的必选项之一。

2. 新商业模式与核心结论:AI新云创新了计价模式,区别于传统通用云按资源时长计费,改为按有效训练产出、按Token交付计价,完成了AI工作负载的账务重构,更适配AI生产需求。核心数据显示,中国AI基础设施云2028年将破4100亿元,年复合增速24.6%;美国新云2030年将达1800亿美元,年均增速69%,新云不替代通用云,只是重构AI算力的价值链,会和通用云长期并存,是AI时代基础设施的核心主战场之一。

返回默认

声明:快读内容全程由AI生成,请注意甄别信息。如您发现问题,请发送邮件至 run@ebrun.com 。

我是 品牌商 卖家 工厂 服务商 平台商 研究者 帮我再读一遍。

Quick Summary

This article outlines the current industry status of the emerging AI-native cloud sector, with key takeaways as follows:

1. Marketing landscape: Between January and April 2026, total outdoor ad spending on China’s cloud services sector reached 352 million yuan, nearly doubling year over year. 84.7% of this spending was concentrated at major first-tier hub airports. Airports have become the core channel for cloud vendors to build brand awareness among B-side decision-makers, who are exactly the core stakeholders driving enterprise digital and AI transformation. These ads prioritize long-term mindshare over immediate conversion.

2. Industry landscape: The current AI cloud market is divided into three categories of players: defensive general-purpose hyperscalers, MaaS and Agent players targeting user entry points, and differentiated AI-native cloud providers. These three layers are developing in parallel. AI-native cloud is purpose-built from the ground up for generative AI and Agent workloads, with an architecture fully adapted to AI requirements from the bottom up. It will coexist with general-purpose cloud long-term, and has already grown into one of the main battlegrounds for AI-era infrastructure, with a market growth rate far outpacing that of traditional cloud.

This article summarizes B2B brand marketing logic and the latest trends in the AI cloud industry based on the booming airport advertising war for AI cloud services, with key insights below:

1. B2B brand marketing logic: First-tier hub airports precisely reach core AI transformation decision-makers such as enterprise CEOs, CTOs and CIOs. The location offers unique advantages including a captive environment, longer audience dwell time, and a shorter decision influence chain, making it ideal for high-end B2B brands to build mindshare. It not only shapes customer decisions, but also boosts internal sales team morale. The nearly doubling of 2026 cloud service outdoor ad spending, with over 80% allocated to airports, validates the unique value of this channel.

2. Industry trends: Enterprise AI adoption has fully shifted from pilot testing to full production deployment, driving rapid demand growth for specialized AI computing power. This has spawned an independent track for AI-native cloud, which has gained market recognition thanks to its AI-native architecture and neutral positioning. The industry landscape has shifted from a single general-purpose cloud market to a layered structure where general-purpose cloud and AI-native cloud will coexist long-term. The AI-native cloud market is about to cross the 100-billion-yuan threshold, with an annual growth rate far exceeding that of traditional cloud, leaving huge room for future expansion.

This article discloses the latest developments in the AI cloud track, as well as opportunities, risks and actionable insights for cloud service practitioners, with key takeaways below:

1. Market and demand shifts: AI-native cloud has grown from a niche player to the main growth engine of the industry. Institutional forecasts project that China’s AI infrastructure cloud market will exceed 410 billion yuan by 2028, with a compound annual growth rate (CAGR) of 24.6%. The U.S. AI-native cloud market will reach $180 billion by 2030, with an annual growth rate of 69%, indicating enormous growth potential. As enterprise AI moves from testing to production, demand for high-frequency inference, Agent execution, and industry model fine-tuning has exploded. General-purpose cloud cannot adequately accommodate these workloads, creating a clear market gap.

2. Opportunities and risk warnings: AI-native cloud does not need to replace general-purpose cloud; a differentiated complementary strategy is sufficient. Players can capture growth by leveraging the emerging layered procurement trend where enterprises use general-purpose cloud as a base layer and AI-native cloud for AI execution layers. On the risk side, players need to watch out for margin pressure on GPU rental businesses from general-purpose cloud providers’ self-developed AI chips. To build competitive moats, players should strengthen their AI-native architecture and neutral heterogeneous adaptation capabilities, and explore new business models such as pricing based on task output.

This article offers key insights for manufacturing facilities advancing digital transformation and rolling out AI applications, with core takeaways as follows:

1. AI transformation product and demand insights: AI has now fully entered the production application stage. When factories deploy AI to optimize production, R&D and management, they require stable and efficient AI computing power support. Traditional general-purpose cloud architectures are designed for general workloads, leading to low AI workload scheduling efficiency and underutilized GPU capacity, which cannot meet factories’ production-grade AI requirements. Purpose-built AI-native cloud can deliver more efficient AI computing services with more transparent pricing.

2. Business opportunities and transformation insights: China’s AI-native cloud natively supports core requirements such as mixed deployment of domestic chips and data compliance, which aligns with policy requirements for digital transformation for Chinese factories. For factories expanding overseas, AI-native cloud also offers flexible global scheduling capabilities to meet compliance requirements across different regions. Factories do not need to replace their existing infrastructure to transform; a hybrid model of general-purpose cloud as the base layer plus AI-native cloud as the AI execution layer balances cost and efficiency, and accommodates all types of factory workloads.

This article sorts out the latest industry trends, customer pain points and actionable solutions in the AI infrastructure space, with key insights for AI and cloud-related service providers below:

1. Industry development trends: The AI cloud sector has now formed a three-layer parallel structure: underlying computing power, middle-layer MaaS, and upper-layer Agent entry points. AI-native cloud has evolved from a complementary offering of general-purpose cloud to an independent industry layer, becoming one of the core growth drivers of the entire sector. Its market size is expanding from tens of billions to hundreds of billions of yuan, with a compound annual growth rate far outpacing traditional cloud. AI-native scheduling and outcome-based pricing will become the industry standard in the future.

2. Customer pain points and solutions: The core pain point for enterprise customers is that legacy general-purpose cloud architectures cannot adapt to production-grade AI workloads, with issues including low scheduling efficiency, low GPU utilization, and mismatched pricing models. Customers also require neutral, exit-friendly computing options, domestic heterogeneous adaptation, and compliance support that general-purpose cloud cannot fully provide. For service providers, the recommended path is to build an AI-native architecture that adapts to AI workloads from the bottom up, supports unified scheduling of heterogeneous chips, adopts pricing based on task output, and positions itself as a neutral AI infrastructure layer that complements rather than competes with general-purpose cloud.

This article offers valuable references for strategic planning and operations management for cloud platforms and intelligent computing platform operators, with core takeaways as follows:

1. Core market demand for AI cloud platforms: After enterprise AI entered the production phase, customers require more predictable AI capacity, higher GPU utilization, more transparent pricing, neutral and exit-friendly computing options, adaptation for domestic heterogeneous chips, and compliance support across different regions. All these are needs that existing general-purpose cloud platforms cannot fully meet.

2. Actionable practices and risk mitigation: Platforms can focus on rebuilding their technical architecture for AI workloads, build an AI-native scheduling system, separate training and inference workflows, and adopt pricing models based on effective output or token delivery. They should pursue a differentiated positioning that complements general-purpose cloud, rather than trying to cover all product categories. On the risk side, platforms need to prepare for market pressure from general-purpose cloud providers’ self-developed AI chips, and strengthen their moats in AI-native scheduling, neutrality and heterogeneous adaptation. They should also capitalize on the trend of regional computing power layout, actively expand emerging markets at home and abroad, and capture structural growth opportunities.

This paper sorts out the latest developments in the AI cloud industry, discloses core data and new industry characteristics, with key insights for industry researchers below:

1. New industry developments: The AI cloud industry has formed a three-layer parallel industry structure. AI-native cloud has split off from the general-purpose cloud ecosystem to become an independent industry layer, completely changing the traditional "single cloud" procurement logic for enterprises. It has given rise to a new combined procurement paradigm: "general-purpose cloud as base layer + AI-native cloud as AI execution layer + NeoPaaS as governance layer. AI-native cloud has shifted from an emerging option to one of the required components for enterprise cloud services.

2. New business models and core conclusions: AI-native cloud has innovated its pricing model: unlike traditional general-purpose cloud, which charges based on resource usage time, it prices based on effective training output and token delivery, overhauling the billing structure for AI workloads to better align with AI production requirements. Core data shows that China’s AI infrastructure cloud market will exceed 410 billion yuan by 2028, with a CAGR of 24.6%; the U.S. AI-native cloud market will reach $180 billion by 2030, with an annual growth rate of 69%. AI-native cloud does not aim to replace general-purpose cloud, but only restructures the value chain of AI computing power. It will coexist with general-purpose cloud long-term, and is one of the core main battlegrounds for infrastructure in the AI era.

Disclaimer: The "Quick Summary" content is entirely generated by AI. Please exercise discretion when interpreting the information. For issues or corrections, please email run@ebrun.com .

I am a Brand Seller Factory Service Provider Marketplace Seller Researcher Read it again.

那些从第一天起就为AI重写的云,正在收获架构红利带来的增长加速度。

抵达首都机场T3后,在出发大厅,迎面而来的是一排连贯的LED灯箱:阿里云、火山引擎、百度智能云、九章云极。这不是今年才发生的事。

从2025年起,国内一线枢纽机场的AI云广告密度就一路跳涨。中天盈信(CODC)户外广告监测数据显示,2026年1-4月云服务行业户外投放规模达3.52亿元,较上年同期近乎翻倍,其中机场媒体占比84.7%。

把镜头拉到太平洋另一侧,旧金山SFO、拉斯维加斯哈里里德机场的到达大厅里,谷歌GCP、AWS与CoreWeave的牌子并列挂着,AI云同样是核心卖点。

枢纽机场从来不是普通消费品牌的秀场,而是云厂商的“高端局”牌桌。它的受众精准(商旅差旅人群里塞满了CEO、CTO、CIO、央企数字化部总经理),环境封闭(登机前45分钟你抬头只能看见它),且停留时间长、决策影响链短。一位国内头部云厂CMO在2025年中对外聊过同一个判断:决定企业上不上云、All in不All in AI的人,就在这批人里;机场广告打的不是转化,是心智占位,顺便给自家销售团队提气。

当我们把中外机场里那些AI云广告牌连起来读,读到的不是几句slogan,而是一张AI基础设施产业的分布图:通用超云、大模型MaaS与Agent入口派,以及一类近两年越来越显眼的角色——AI新云(Neocloud)。

为什么是“新云”走到了这块牌子下

把2025-2026年机场里出现的AI云广告摊开,能归出三种典型叙事。

第一种,通用超云的“守擂与统揽”。阿里云在首都T3写“AI放心交给阿里云”“2025中国全栈AI云服务(公共云)市场份额超四成”,把云和AI做了强关联,还强化了自身在云时代的品牌占位。华为云、天翼云、移动云把广告铺向虹桥火车站包柱与二线城市机场,讲国产算力与整体替代。AWS在拉斯维加斯不喊“第一”,只挂客户案例——“Netfor用GenAI把等待时间降99%”“PhaseZero全押AWS跑Agentic AI”,这是份额老大典型的守擂姿态:我不自证,我让客户替我说话。

第二种,MaaS与Agent派的“入口抢夺”。火山引擎在国内机场打“你的AI云”“占中国公有云大模型调用量49.5%”,走的是IDC的MaaS Token调用量口径。2026年夏,豆包、千问办公、百度搭子这类Agent工具开始挤掉部分云底座广告,占住登机口屏幕,云厂商的竞争从“谁有算力”外溢到“谁离员工工作流更近”。

第三种,AI新云的“并列声明”。旧金山SFO里,CoreWeave和谷歌GCP的牌子挨着,CoreWeave写“The world's best AI?Yeah,it runs on CoreWeave”。北京首都T3和上海虹桥的同一排灯箱里,九章云极与阿里云、火山引擎并列,不走“第一”自证,讲“中立智算云”“AI工厂”“专业Token”。这类厂商要释放的信号很克制:我和它们是同一个牌桌上的玩家。

三种叙事同场,反而把行业真实面貌说清楚了——AI云不是一条赛道同速跑,而是多层结构并行:底层算力、中间MaaS、上层Agent入口,各自抢各自的心智。机场大屏只是把这套分层压缩成肉眼可辨的并排关系。

令人印象深刻的是,新云(Neocloud)这个概念,2024年前后在海外随CoreWeave、Lambda、Nebius、Together AI的融资与上市节奏被反复讨论。Forrester在2025-2026年连续把Neocloud定义为“AI-native cloud的基础设施层”,专为生成式AI与Agentic AI工作负载设计,以GPU、加速卡为核心,优化计算、存储、网络、调度与AI Ops,目标是比通用云提供更可预测的AI容量、更高的GPU利用率、更清晰的AI工作负载经济账。

这个定义里藏着新云走到机场牌桌下的根本原因。

通用超云的底盘是过去十五年互联网、企业IT上云沉淀的虚机、存储、带宽、数据库、中间件,AI是叠加层。它不是不能做AI,而是架构主键是兼容一切工作负载,AI训练推理队列要和虚机热迁移、块存储快照、公网SLB共用一套调度器。

新云的技术栈从第一天起就是AI工作负载主键:底层做异构芯片统一抽象(英伟达+昇腾+海光等),中间层做训推一体队列与KV Cache感知调度,上层把模型训练、微调、评估、推理、Agent执行封装成可计量服务。CoreWeave能比多数通用云更早铺H100、GB200 NVL72,不是因为它钱更多,是因为它不需要向虚机用户解释“为什么这块卡不能切出来给你”。

商业模型上,新云也在形成自己的计价逻辑。通用云习惯按“虚机规格×小时”或“存储×GB×月”计费,AI客户却想知道跑完这个epoch花多少、每百万Token推理成本多少、缓存命中率提升能否直接折算到账单。CoreWeave、Together AI走“GPU集群租赁+按Token API”双线,其中Together AI约30-40%收入来自per-token API,60-70%来自GPU租赁。

2025年起,新云一词也伴随九章云极、以及智算云专项报告的发布进入中国市场主流视野。架构层面,九章云极的Alaya NeW OS做异构统一调度、把训练工厂和Token工厂拆成两条生产线,也是同一套逻辑——为AI重写字,不为通用负载妥协。商业层面,九章云极提倡训练工厂按有效训练产出、Token工厂按专业Token交付,从按资源时长切换到按任务产出计价。这种改变不是文字游戏,是AI工作负载从实验走向生产必然发生的账务重构。

Forrester与Synergy看到的曲线:新云不是补充

2025年是关键的一年,AI新云从通用云里脱颖而出,成为独立可计价的产业层。

需求端,DeepSeek-R1出圈、Agent在2025下半年井喷、企业从测模型进入跑生产推理。供给端,通用云自研芯片包括Trainium、Maia、TPU,和新云抢卡形成双向挤压,新云厂商凭借更早拿旗舰卡+更纯的AI调度获得窗口期。两端合力,新云完成了从“GPU二房东”到“AI基础设施平台”的语义升级。

研究机构Forrester在《2025年十大云计算趋势》及后续中国AI基础设施云专项里给出过一组判断:通用云时代逐步落幕,AI原生云(含Neocloud与NeoPaaS两层)成为下一阶段主线;2025年中国AI基础设施云TAM约2120亿元,2028年破4100亿元,年复合增长率24.6%;中国企业AI采用率94%,超78%的AI工作负载跑在公有云、智算云上。

报告把CoreWeave、Lambda、九章智算云Alaya NeW Cloud列入同一观察集合,认为新云正从边缘角色跃升为增长主导力量之一。

Synergy Research Group对美国Neocloud市场的测算更加鲜明:2025年约230亿美元,2030年近1800亿美元,年均增速69%。全球Neocloud收入头部是CoreWeave、Crusoe、Lambda、Nebius,二三梯队的Together AI、Fluidstack、NScale、Vultr还在爬坡。需要注意,Synergy的neocloud口径含GPU云租赁,不和通用云IaaS收入直接加减,但方向一致。

这里要澄清一个常见误读:新云不是通用云的替代品。

截至2026年8月,CoreWeave市值约465亿美元,积压订单总额(Backlog)已飙升至1040亿美元以上,客户含OpenAI、Meta、微软。尽管它贵,但客户买的是6个月交付GB200集群和可退出AWS绑定的组合。因为新云的价值公式是敏捷性×可退出性×AI原生效率,不是单价。国内场景里,九章云极这类厂商的锚点类似:不绑单芯片、不绑单模型、做异构与Token运营的中立层,让客户在阿里云、火山引擎、AWS之外有第三个可审计选项。

过去两年圈内喜欢用“新云会不会被通用云碾死”来框架化讨论,这个框架本身过时了。

一方面,企业AI负载正在分层。低频训练、长周期实验,通用云的全栈生态仍最优;高频推理、Agent执行、行业专用模型微调、国产异构合规场景,新云的AI原生调度与中立性更贴。Forrester说的Neocloud+NeoPaaS双层结构,本质就是承认未来企业云战略不是“选一朵”,而是“通用云做底座+新云做AI执行层+NeoPaaS做治理层”的组合。这种组合一旦成为主流采购范式,新云的位置是结构性存在,不是周期泡沫。

但同时,从Trainium、TPU到各家厂商自研加速卡,AI专用芯片正在向上挤压GPU租赁市场,覆盖范围从训练延伸到推理和Token运营。专用芯片与通用GPU在业务边界上有摩擦,但两者烧的是不同预算,一个偏基础设施采购,一个偏弹性算力消耗。所以我们的判断是:新云不替代通用云,新云是把AI算力那一段价值链重新做了一遍。

从地缘维度看,中国市场同样存在三类刚需:国产芯片混布、数据出境合规、行业专云。通用云的历史调度器难以适配这些要求,而新云原生支持这些能力,反而形成了壁垒。海外同样在验证这个逻辑:九章云极在印尼落智算中心、向越南、沙特、欧洲等地积极探索项目,CoreWeave、Lambda盯欧洲EMEA电力缺口。这些动作指向同一判断——AI算力未来是“区域主权+全球调度”双轨,新云比通用云灵活。

所以,AI时代的专用云,迎来了它的时代。就像电刚出来时工厂自己烧蒸汽,电网成熟后蒸汽退场、电变成公共服务。AI算力正在走类似的路径:通用云是电网+部分自营发电机,新云是为AI负载专门优化的电站。两者长期并存,前者覆盖广泛业务,后者在AI负载上更纯、更高效。

AI新云的下一程:定义未来的时刻已经到来

当所有厂商都在争相定义什么是“第一”,真正的赢家将是那些重新定义赛道规则的人。AI新云正在完成这一使命:它将AI算力的交付从资源租赁转向任务产出,从黑盒调度转向透明度量,从通用妥协转向原生优化。这套逻辑不是对传统云的修补,而是对AI基础设施的重新发明。

机场广告牌上的并排展示,映射的是产业格局的真实演变。新云已经从“新兴选项”成长为“必选项”之一。随着企业AI负载全面进入生产阶段,高频推理、Agent执行、行业专用模型微调等场景持续爆发,新云的AI原生调度能力、异构兼容广度和按效计价模式,将不再是差异化优势,而是行业标配。

Forrester与Synergy的数据已经指明方向:新云的市场规模正在从百亿级迈向千亿级,年复合增长率远超传统云。这不是一个补充性的配角市场,而是AI时代基础设施的主战场之一。那些从第一天起就为AI重写的云,正在收获架构红利带来的增长加速度。

注:文/杨丽,文章来源:钛媒体(公众号ID:taimeiti),本文为作者独立观点,不代表亿邦动力立场。

文章来源:钛媒体

广告
微信
朋友圈

FAQ回顾

AI新云是什么?和通用云有什么区别?

AI新云(Neocloud)是AI原生云的基础设施层,专为生成式AI、Agentic AI工作负载设计,以GPU、加速卡为核心,优化计算、存储、网络、调度与AI Ops,可提供更可预测的AI容量、更高GPU利用率、更清晰的AI工作负载成本核算。它无需兼容通用工作负载,与通用云长期并存,通用云做底座,新云做AI执行层。

AI新云行业的市场规模和增速是多少?

Forrester数据显示,2025年中国AI基础设施云市场规模约2120亿元,2028年将突破4100亿元,年复合增长率24.6%;Synergy数据显示,2025年美国Neocloud市场约230亿美元,2030年接近1800亿美元,年均增速69%,整体增速远超传统云行业。

AI新云适合哪些企业业务场景使用?

高频推理、Agent执行、行业专用模型微调、国产异构合规场景更适合选择AI新云,其AI原生调度能力、中立属性、按任务产出计价的模式更贴合这类需求;低频训练、长周期实验场景仍优先选择通用云的全栈生态。

云厂商为什么热衷在机场投放AI云相关广告?

机场受众精准覆盖企业CEO、CTO、CIO等AI云采购决策人群,环境封闭、用户停留时间长,投放广告可快速抢占客户心智,影响采购决策,同时也能增强内部销售团队信心。2026年1-4月云服务行业户外投放3.52亿元,84.7%集中在机场媒体。

这么好看,分享一下?

朋友圈 分享

APP内打开

赞 +1
+1
微信好友 朋友圈 新浪微博 QQ空间
关闭
收藏成功
发送
/140 0