Logo
Become a Client

Resource Regions: Asia Pacific

At the IDC Quanta launch webinar, Joe Bradley, CTO at IDC, made the case for why trusting an AI-powered answer shouldn’t require faith; it should require an architecture you can actually inspect. You’ve heard “AI-powered” enough times this year that a healthy dose of skepticism is the right response. Fair. So instead of asking you to trust that IDC Quanta gets its answers right, here’s what Bradley says is actually happening under the hood when it does.

Bradley breaks it down into two layers. The first is an MCP server: essentially a pipe that gives Claude direct access to IDC’s data, the trackers, the forecasts, the market figures. It also carries instructions for how that data is structured and how to use it. The second is IDC’s own Claude plugin, which goes further, Bradley explains. It shapes how the AI reasons about that data, tells it what’s relevant for what purpose, and requires it to surface a source before handing over any answer. Put together, when someone asks a question, Claude isn’t searching a phrase in a database, Bradley says. It’s reasoning with IDC’s own methodology built into the process.

That’s the theory. Here’s what it looks like in practice.

The Moment It Earned Trust

In the product demo, Bradley walked through a scenario: an FP&A analyst building a market forecast ahead of a CFO review, projecting 16% growth in a segment his company competes in. He asked IDC Quanta to check that number against an external benchmark, right inside Excel.

In under a minute, Quanta surfaced IDC’s actual forecast for that market: 12.1% growth through 2029, with the category decelerating to single digits in the later years. His model hadn’t caught up to where the market was actually headed. That’s the payoff of the architecture above. The answer arrived with its source attached.

Why Even Build an App

Technical buyers reasonably ask why IDC needs its own app when Claude and ChatGPT already exist. IDC isn’t positioning itself as a competitor to the assistants people already use daily. It’s building something with a narrower job.

The case for IDC Quanta comes down to control over how IDC’s own data gets handled and delivered. It exists because of what only a dedicated app can guarantee: a single place that collects everything relevant across an IDC relationship, a direct line to a live analyst when the automated answer isn’t enough, and data handling built on tenant isolation, enforced access controls, and audit logs that capture every user action. Which raises the next question technical buyers ask first.

Provenance You Can Check Yourself

Bradley’s last test for anyone skeptical of AI-generated answers is provenance. Can you verify where it actually came from? In a second demo, he showed IDC Quanta processing a strategy document sent over email, then breaking its answer down into cited data cuts, each one tied explicitly to the filters and definitions behind it.

Nothing here is asserted without a source attached, and nothing requires trusting the AI’s summary over the underlying data itself. That’s the actual answer to “how do you know it’s not confidently wrong”: you don’t have to take Quanta’s word for it. You can check.

See It Yourself

The architecture, the demo, and the sourcing are easier to evaluate firsthand than to take on faith. Request a demo, or talk to your IDC account team if you already have one.

Ryan Smith - Content Marketing Director - IDC

Ryan Smith is the Director of Content Marketing at IDC, where he leads brand-level content and social media strategy, aligning research insights with compelling storytelling to engage technology decision-makers. With a background in both IT and marketing, Ryan brings a unique blend of technical understanding and creative strategy to his work. He’s also a seasoned storyteller, speaker, and podcast host who believes the right message, told the right way, can drive both trust and transformation.

过去几年，大语言模型和生成式AI率先改变知识处理、内容生产和软件流程。随着多模态模型、世界模型、VLA模型和边缘AI持续发展，AI正在进入车辆、机器人、无人机、工业设备和城市设施，能力范围也从生成内容、调用软件，扩展到感知环境、规划任务和控制设备。

物理AI面对持续变化的真实环境，需要处理设备状态、空间关系、环境变化、物理规律和安全要求。系统既要形成“感知—决策—执行—反馈”的实时运行闭环，也要形成“真实运行—数据回流—模型训练—仿真验证—重新部署”的研发迭代闭环。前一类闭环影响任务能否稳定完成，后一类闭环影响产品能否持续改进并扩大应用范围。

IDC 2026年全球CEO调查显示，35.2%的受访CEO将物理AI列为未来12—24个月重点关注的新技术投资方向。资源行业的比例达到50%，制造和零售均为46%，医疗为38%。当前市场已经进入投入和生产验证并行阶段，企业普遍从任务清晰、数据可获得、收益可衡量的场景切入。

IDC认为，硬件决定系统能否稳定进入现场，软件影响学习速度、验证效率和复制能力，数据闭环负责把应用需求、软件工具和硬件执行连接起来。物理AI未来的差距，会越来越多地体现在运行数据能否沉淀为场景资产、验证用例和模型更新。

数据闭环贯穿三层产业架构

为了更系统地拆解数据闭环如何落地，IDC将中国物理AI产业划分为三层相互咬合的结构。 应用和业务层定义真实任务、运行流程和商业价值；软件基础设施层组织模型与策略、仿真验证和场景数据；硬件基础设施层提供计算、连接、供能、感知和物理执行条件。

三层围绕真实运行形成循环：应用产生任务和现场问题，硬件完成感知与执行，软件把运行数据转化为训练数据、仿真场景和验证用例，更新后的模型再进入设备和业务系统。下面逐层展开，看每一层在闭环中具体承担什么角色、当前进展到哪一步。

应用和业务层：闭环质量影响商业化节奏

任务边界、数据反馈、安全责任和投入产出，是判断场景成熟度的四项重要条件。运行问题能够被记录、复现和验证的场景，更容易进入规模部署。

智能驾驶已经形成较完整的数据采集、场景挖掘、仿真测试和版本回归流程，是当前数据闭环较成熟的物理AI场景。

具身智能终端正在从动作演示进入连续任务执行阶段。任务成功率、人工接管频率、异常恢复能力和跨环境适应能力，将逐步成为主要评价指标。

无人机和空天系统需要把飞行状态、空间环境、通信链路、任务调度、运行监管和异常处置纳入统一的运行与验证体系。

工业现场与工程装备拥有清晰的工艺、安全和成本目标。质量检测、预测维护、物料搬运、巡检、精密操作和无人作业等任务，已经开始产生可量化的效率、质量和安全收益。制造业具备流程明确、数据基础较好和指标体系成熟等条件，将继续成为物理AI的重要验证场。

智慧城市相关应用需要打通感知、分析、调度和执行。AI进入交通、设施、低空和应急等物理系统的实际运行后，才会形成更完整的物理AI能力。

无论哪个场景，应用层产生的海量运行反馈，都需要经由软件层转化为可复用的能力——这正是下一层的核心使命。

软件基础设施层：把运行反馈转化为可复用能力

软件基础设施层包括模型与策略、仿真与验证、场景数据与合成数据。三类能力通过数据闭环持续协同，决定物理AI的进化速度和规模复制能力。

模型与策略方面，世界模型仍处于多条技术路线并行发展阶段，可用于交互环境生成、状态变化预测和行动结果推演；VLA模型连接视觉、语言和动作，为机器人及自主设备生成任务策略。未来系统将更多采用分层架构，由高层模型理解任务与环境，中间层完成预测和规划，底层控制系统负责实时、稳定和安全执行。

仿真与验证方面，数字孪生、空间智能和仿真平台之间的协同正在加强。数字孪生提供设备结构、空间关系、物理参数、工艺规则和实时状态；空间智能帮助模型理解三维环境和对象关系；世界模型扩展场景生成、状态预测和策略探索；仿真平台承担测试、回归和安全验证。虚拟环境由此可以覆盖训练、方案推演、系统测试和运行优化。

场景数据与合成数据方面，物理AI需要的数据已经从单帧图像和单点记录，扩展到包含时间、空间、设备状态、动作过程和执行结果的连续场景数据。真实运行中的故障、接管、任务失败和高风险事件，需要经过筛选、标注和结构化处理，沉淀为可检索、可复现和可重复使用的场景资产。

数据闭环将分散的模型、数据和仿真工具连接成持续研发体系。任务覆盖度、异常覆盖度、问题复现率、回归关闭率和跨版本一致性，将逐渐成为物理AI软件平台的重要评价指标。物理AI软件平台商业模式也会从单次工具采购，延伸到场景资产管理、模型生命周期管理、持续验证和安全证据链服务。

软件层的一切优化最终都要回到物理世界中验证，而硬件层正是这个闭环的起止点。

硬件基础设施层：连接真实数据与物理执行

硬件层包括能源电力、算力和IoT/OT连接基础设施。云端承担模型训练、批量仿真和数据处理，边缘节点负责现场协同与模型管理，设备端完成实时感知、推理和控制。

传感器、摄像头、雷达和设备运行系统提供真实环境数据，控制器和执行器负责把模型判断转化为物理动作。硬件层既是数据闭环的起点，也是系统执行结果的出口。

涉及车辆制动、机器人避碰和工业安全的关键任务，还需要保留本地执行、故障降级、冗余控制和人工接管能力。

中国在设备制造、通信网络、能源系统和工程实施方面具备较好基础，高端AI芯片、工业级传感器、核心零部件和复杂环境可靠性仍需持续提升。

三层架构完整就位之后，真正的挑战不在于单点突破，而在于如何让闭环从技术层面延伸到行业应用层面。

展望：从技术闭环走向行业闭环

未来几年，世界模型仍将保持多条技术路线并行，数字孪生、空间智能、仿真平台、合成数据和真实运行数据之间的连接会继续加强。模型能力的提升将扩大系统可以完成的任务范围，验证体系则负责控制进入真实环境的速度和风险。

中国拥有丰富的制造、城市运行和机器人应用场景。下一阶段需要提高软件平台的产品化水平，推动场景数据跨项目复用，加强设备接口协同，并建立覆盖模型、硬件和运行过程的评测与验证体系。

物理AI的市场差距将更多体现在系统学习效率、验证可信度、长期运行能力和跨场景复制能力。企业能否把现场问题快速转化为场景资产、验证用例和模型更新，将影响其产品迭代速度和规模化能力。

IDC将持续跟踪物理AI技术、市场和生态变化，并通过“物理AI+行业场景”的研究方式，重点关注制造业、城市物理系统和具身智能等方向，进一步分析技术架构、市场机会、厂商格局和产业化路径，为技术供应商和行业用户提供持续参考。

进一步交流

如需了解物理AI架构评估、场景落地路径或数据闭环能力诊断等研究方向，欢迎联系IDC中国物理AI与行业智能化研究团队。我们将安排对应行业分析师与您深入沟通，提供定制化决策参考。

Lily Li

Lily Li - Research Manager

Analyst Profile
Lily is the Research Manager for China Robotics and Embodied Intelligence, specializing in market research on embodied intelligent robots. She has long focused on the development trends of China’s embodied intelligence robotics industry, systematically studying the evolution of robot hardware,…

This is the second post in Meet IDC Quanta, a short series showing what the product actually does, starting with the portal, then your inbox, then wherever you’re already working in Claude.

It’s 4:47 PM. You’re prepping for tomorrow’s board meeting, and a question just landed that wasn’t on your list: how does your cloud infrastructure position compare to what’s shifted in the market over the last two quarters. You could open a portal, log in, run a search, filter by date, read three reports, and synthesize an answer. Or you could pull up your phone in the elevator.

Most executives don’t have a research workflow. They have an inbox. IDC Quanta was built on that premise: the fastest tool for a time-pressed leader is the one they never have to open, because it’s already open.

What it actually does

IDC Quanta in email lets you ask a question and get a sourced IDC answer back in about 60 seconds. It works the way you’d already expect email to work, which is exactly the point.

Emailing Quanta
  1. Compose an email to brief@quanta.idc.com, from your phone, your laptop, whatever’s in front of you.

  2. Ask the question in plain language. No query syntax, no keyword tricks. “Which cloud infrastructure vendors gained share in EMEA in the past 12 months?” is a complete request.

  3. Send it, and keep moving. A structured, sourced answer arrives in about 60 seconds, fast enough that you can send the question walking into a meeting and have the answer before you sit down.

  4. The reply carries its own receipts. Every claim in the answer traces to a specific IDC source. If you want to go deeper, “View in IDC Quanta” drops you straight into the full portal conversation, context intact.
A sample email response from IDC Quanta

No login screen to remember, no new app for IT to vet. And no tab-switching mid-meeting. The intelligence comes to the inbox. The inbox doesn’t change to accommodate it.

Where this earns its keep

The mechanic is simple. What makes it valuable is what it replaces across a week of an executive’s actual work:

  • Before the call.
    A prospect meeting is in twenty minutes and you need a current read on their competitive positioning. Email the question on your way to the conference room. The answer’s there before you sit down, cited and ready to use.

  • Benchmarking your own thinking.
    Attach a competitive deck or account plan to your email along with your question. IDC Quanta reads it alongside its own research base and flags where your internal view and IDC’s tracked data disagree. You’re getting your own analysis checked against the numbers.

  • The follow-up nobody has time to chase.
    Someone asks a sharp question in a meeting and the honest answer is “let me get back to you.” Now that follow-up takes one email and about a minute. Who else on your team wishes they had that?

  • Board and investor prep, compressed.
    The night-before scramble for one more data point doesn’t need the whole deck reopened. One email, one sourced answer, dropped straight into the slide.

Why this isn’t just a fast AI reply

A lot of tools will answer an email question with confidence. Confidence isn’t the same as being right, and it isn’t the same as being defensible in a room full of people who will ask where the number came from. Every IDC Quanta answer draws on IDC’s research base: 1,000+ analysts across 100+ countries, tracking 15B+ data points and 800K+ companies annually. That’s the citation trail behind every answer. It’s what makes an answer dropped in your inbox worth repeating in the room.

Get it in your inbox

If you’re already an IDC Quanta customer, brief@quanta.idc.com is live. Send the question you didn’t have time to research properly and see how fast a sourced answer actually moves.

If you’re not yet a customer, request a demo, and we’ll show you what it looks like when your inbox starts acting like a research team.

Ryan Smith - Content Marketing Director - IDC

Ryan Smith is the Director of Content Marketing at IDC, where he leads brand-level content and social media strategy, aligning research insights with compelling storytelling to engage technology decision-makers. With a background in both IT and marketing, Ryan brings a unique blend of technical understanding and creative strategy to his work. He’s also a seasoned storyteller, speaker, and podcast host who believes the right message, told the right way, can drive both trust and transformation.

Physical AI时代开启：具身智能机器人重构新一代生产力

7月20日，2026世界人工智能大会迎来收官之日。值此大会尾声，IDC同期举办了“Physical AI时代开启：具身智能机器人重构新一代生产力”线上专场会议，IDC中国分析师团队基于全球追踪数据及工业用户调研，系统分享了具身智能机器人的技术演进、市场格局与商业化路径。本文为会议核心内容纪要。

本届WAIC特设具身智能专属展区，汇聚众多机器人整机、核心零部件及具身智能大模型企业，集中展示量产人形机器人、四足机器人、轮臂复合机器人以及具身智能大模型的最新进展。展馆内，人形机器人在展台前挥手致意、四足机器人在人群中灵活穿行、机械臂在模拟产线上精准操作——围观的人群里，既有举起手机拍照的普通观众，也有拿着笔记本仔细记录参数的工程师和采购决策者。

热闹背后，一个更值得关注的信号正在浮现：今年的展品不再只是“能动的演示”，而是围绕3C电子、汽车制造、仓储物流、商用服务、家庭陪伴等真实场景展开应用展示，反映出中国具身智能机器人正由技术验证迈向商业化落地，产业实用价值持续释放。

这一变化背后，是AI产业正由生成式AI迈向Physical AI。生成式AI赋予机器认知能力，Physical AI进一步赋予机器在物理世界中自主感知、规划和行动的能力。国际数据（IDC）最新用户调研结果显示，继生成式AI之后，Physical AI将成为未来两年企业AI布局的重点方向。

具身智能机器人是Physical AI最重要的产业载体。IDC预计，未来五年中国具身智能支出将由14亿美元增长至770亿美元，年复合增长率达94%，中国有望持续引领全球具身智能机器人产业发展。

机器人原生智能的三大技术底座

围绕机器人原生智能的持续提升，正在推动模型、数据和算力三大技术底座加速演进。这三者环环相扣，共同决定了机器人的智能化上限。

在模型层面，世界模型与VLA（视觉-语言-行动）的协同正成为核心架构。世界模型负责物理世界建模、任务规划及仿真数据生成，VLA模型负责感知、理解与动作生成，两者协同推动具身智能机器人迈向理解、预测、规划、执行的全面能力。

在数据层面，“仿真优先、虚实融合”已成为行业共识。互联网视频、第一视角数据、仿真数据和真实机器人运行数据共同构建数据飞轮，持续驱动机器人能力迭代。数据飞轮转得越快，机器人的进化效率就越高。

在算力层面，AI原生计算正驱动机器人架构全面升级。随着模型规模持续扩大，推理算力需求快速增长——IDC预测，到2027年推理将占智能算力需求的70%以上。云边协同带动计算平台、感知系统、运动控制及机器人操作系统全面升级，为规模化部署提供支撑。

从关注热度走向落地探索，多形态具身机器人需求形成

技术突破之外，市场的反应同样关键。IDC用户调研显示，超过80%的企业已开始关注并评估具身智能技术，部分企业已进入试点验证阶段。不同应用需求正在推动机器人向多形态发展，各品类的商业化节奏也呈现出明显差异。

人形机器人是增长最快的品类。2025年全球出货量接近1.8万台，同比增长800%，正由展示验证走向工业应用。中国厂商保持领先，中尺寸机型贡献主要出货量，全尺寸机型支撑高价值市场。IDC预计，2030年全球出货量将突破51万台。

四足机器人形成“消费驱动规模、商用驱动价值”的发展格局。2025年全球出货量预计约6万台，工业巡检、应急救援等行业应用持续扩大，情绪陪伴与专业作业需求共同推动市场增长。

商用服务机器人进入全球化扩张阶段。2025年全球出货量超过15.5万台，配送、清洁等应用持续向室外、多功能及多机器人协同方向发展，中国厂商继续保持全球领先。

外骨骼机器人加速向医疗康复、消费助力和工业辅助等多场景拓展。2025年中国市场出货量约2.6万台，多模态感知、轻量化设计及AI算法持续提升产品能力，医疗康复场景已形成成熟商业模式，消费与工业场景正处在从0到1的突破期。

家庭清洁机器人：家庭具身智能发展的最佳入口

相比通用家庭机器人仍处于早期探索阶段，家庭清洁机器人已率先实现规模化商业落地，成为家庭具身智能发展的最佳入口。

2025年全球家庭清洁机器人出货量达到3273万台，同比增长20.1%，市场正由单一室内清洁迈向覆盖扫地、割草、擦窗、泳池等场景的全场景智能清洁。其中，扫地机器人仍是最大的细分市场，2025年全球出货量达到2412万台；割草机器人和擦窗机器人保持高速增长，增速分别达到64%和70.4%，无线化、智能化成为产品升级的重要方向。

随着AI能力持续提升，行业竞争正由清洁能力转向智能能力。空间理解、自主导航、多模态感知、智能决策以及家庭IoT生态协同，正在成为产品差异化的核心竞争力。依托AI算法、供应链和全球化布局优势，中国厂商持续扩大在全球家庭清洁机器人市场的领先地位。

展望未来，家庭清洁机器人将逐步由单一功能设备演进为家庭具身智能体，并进一步融合家庭大模型、智能家居和IoT生态，成为未来家庭服务的重要智能终端。

商业化进入规模复制阶段

当技术逐步成熟，产业竞争的核心正由单点技术突破转向规模化落地能力。谁能在真实场景中跑通闭环、实现复制，谁就能在下一阶段占据主动。

从应用场景来看，分层演进的节奏已逐步清晰。服务场景率先完成市场培育；工业场景进入规模化导入阶段，码垛、搬运、拾取、检测等应用持续落地；家庭场景则处于技术迭代和应用培育阶段，蓄势待发。

从工程化能力来看，落地周期持续缩短。机器人正从单一产品走向工程化交付，模块化软硬件提升了场景适配能力，全栈解决方案与云边端协同架构加快应用部署，多品类机器人协同作业持续拓展应用边界，推动具身智能从单机智能迈向系统智能。

从商业模式来看，行业正由卖产品走向卖服务。“硬件+软件+服务+AI”的融合模式正在成为主流，一次性销售、RaaS租赁/订阅、服务增值、本体与模型协同等模式并行发展。其中RaaS有效降低了用户使用门槛，加速了机器人的普及推广。

IDC四个判断：未来五年的竞争焦点

基于持续的全球市场跟踪，IDC对具身智能机器人产业提出四点判断：

  • 技术竞争进入系统能力竞争阶段。 模型、数据和算力将持续协同演进，世界模型、VLA和数据飞轮将成为机器人智能能力提升的核心驱动力。单点技术的领先不再足以构建壁垒，系统能力才是决胜关键。
  • 商业价值将由场景验证走向规模复制。 未来竞争重点将由技术突破转向场景适配、工程化交付和商业模式创新，机器人将进入规模化部署阶段。谁能率先跑通场景闭环，谁就能占据先机。
  • 生态能力将成为产业竞争的分水岭。 平台能力、开源生态、数据闭环和产业协同，将决定企业能否在五年后依然留在牌桌上。生态建设不再是可选项，而是必答题。
  • 中国厂商有望持续引领全球产业化。 依托完善的制造体系、丰富的应用场景、完整的供应链和数据规模优势，中国企业有望继续引领全球具身智能机器人产业的发展。

进一步交流

具身智能机器人产业正处在从技术验证走向规模化落地的关键转折期。IDC持续追踪全球机器人市场动态，覆盖人形、四足、商用服务、家庭清洁、外骨骼等全品类，为企业提供数据驱动的市场洞察与战略决策支持。如需获取完整报告、行业数据或与分析师团队深入交流，欢迎联系IDC中国。

Lily Li

Lily Li - Research Manager

Analyst Profile
Lily is the Research Manager for China Robotics and Embodied Intelligence, specializing in market research on embodied intelligent robots. She has long focused on the development trends of China’s embodied intelligence robotics industry, systematically studying the evolution of robot hardware,…
Predictions Jul 20, 2026 Erin Lin

IDC最新报告：医院核心业务系统市场规模破100亿大关，智能体与医共体成为“十五五”时期新增长极

2025年国内医院核心业务系统市场的智能体、医共体、本土替代的改造需求将稳步释放，成为新的增长极。头部厂商相继完成云原生底座迭代与医疗大模型原生融合布局，AI 智能体、全域数据治理、全栈信创持续迭代，未来市场的竞争将更加聚焦于技术深度与、服务运营与生态协同能力的综合较量。

Read full release
Marketing Jul 20, 2026 Siri Si

IDC最新报告：揭秘金融生成式AI规模化卡点，端到端落地能力决定胜负

未来市场竞争不再比拼单一产品功能，而是由分层产品体系、金融专属工程落地能力、多智能体协同架构、长期持续运营服务四大维度共同决定。数据治理、模型全生命周期管控、智能体治理等核心议题将贯穿行业发展全程，能够深度融合金融业务规则、解决规模化落地工程难题、绑定业务长期价值的解决方案，将主导下一阶段金融行业大模型市场格局。

Read full release

2026年7月17日，上海世界人工智能大会（WAIC）现场，全球首款 AI 智能体手机、23 克轻量化 AI 眼镜集中首发，人形机器人走出演示场景落地工业、服务一线。

这些现场实景与IDC最新终端市场数据形成双向印证：全球智能终端产业已正式告别概念化技术叙事，AI从附加功能升级为重构硬件、交互与产业格局的底层核心。对于不同角色的参与者而言，这场变化的含义与应对路径各不相同。

数据层面：AI终端全线增长，分化格局已现

IDC研究数据显示，在国内整体终端市场承压背景下，AI硬件成为唯一增长主线，各细分赛道呈现差异化扩张态势：

  • 手机：2026年国内生成式AI手机渗透率突破50%，正式进入普及周期。端边云协同算力、原生智能体系统、全场景生态是行业突围核心方向。
  • 具身智能机器人：中国持续引领全球发展，预计2026年市场规模接近50亿美元。其中人形机器人作为高阶载体，全球出货同比增长超过150%，工业场景为重点突破领域。
  • 家用大清洁机器人：家用机器人仍是出货量最大的机器人品类，2026年第一季度全球出货893万台，同比增长36.7%。扫地机器人出货656.3万台，同比增长29.4%，头部厂商在欧洲市场布局力度持续加大。
  • PC/平板：2026年中国Gen AI PC出货587万台，市场占比11.9%，销量同比提升166.4%；Gen AI平板出货112万台，占比35.6%，销量同比提升68.4%。AI已成为平板摆脱低价同质化竞争的核心抓手之一。
  • 可穿戴/智能眼镜：轻量化硬件与AI大模型融合已成新品标配，1500-3000元价位为核心竞争区间，AI集成与多元场景生态是增长核心驱动力。

产业增长驱动力源于大模型成熟、国产零部件配套升级、实景需求持续释放。同时，跨应用权限互通、智能体长效记忆、数据隐私合规仍是当前三大核心瓶颈。

趋势层面：WAIC实景揭示的四重变革

本届WAIC集中呈现了四个方向的结构性变化，这些变化正在重新定义终端产品的竞争逻辑：

趋势一：AI原生软硬件闭环成为行业硬性标准

阶跃星辰STEPX Neo搭载自研Step AOS智能体原生系统，在安卓底层增设智能体调度层，搭建”大模型-原生系统-硬件”一体化架构；努比亚量产AI手机依托MCP/A2A协议实现跨应用自主执行；Rokid原生操作系统YodaOS落地消费级智能眼镜。行业竞争重心已从硬件参数比拼转向软硬一体生态建设。

趋势二：交互逻辑从“人操控设备”向”机器理解意图”跃迁

传统触控、语音交互正向意图感知交互迭代：STEPX Neo依靠双域长效记忆简化重复指令；智能眼镜打造”被动感知、主动服务”模式，依托视听感知自动识别需求并后台响应，实现无操作主动服务。交互效率的提升，直接决定了终端设备在真实场景中的可用性。

趋势三：端云协同架构标准化，具身智能搭建数据闭环

主流智能眼镜统一采用”端侧轻量化模型+云端大模型”分层算力架构，空间计算芯片与自研光学引擎成为头部厂商高端产品趋势标配。人形机器人赛道重点布局世界模型、数字孪生、合成数据体系，具备”模型训练-数据采集-落地反馈”完整闭环的企业，竞争优势已显著拉开。

趋势四：单品智能向多设备场景协同升级

大会首发的”HGR人·镜·犬混合智能协同系统”，依托智能眼镜联动机器狗完成现实服务任务；工业、仓储、商用、家庭多场景同步落地多机器人联动方案。行业发展方向已从单一设备智能化，转向全域场景协同智能。

综合以上趋势，IDC得出三个核心判断：

判断一：行业已跨越尝鲜期，进入普及阶段。AI手机渗透率过半、机器人市场增速翻倍，叠加量产机型集中发布，表明消费者需求已从猎奇尝鲜转向实用刚需，市场增长由厂商单向教育驱动转为用户主动采购拉动。

判断二：场景落地能力正在成为企业核心分水岭。厂商宣传逻辑从”硬件功能堆砌”转向”垂直场景解决方案”——人形机器人深耕制造、物流商用场景；智能眼镜差异化布局本地生活、通用办公、个性化服务。真实场景的落地深度，是拉开企业差距的关键指标。

判断三：国产产业链完成从跟跑到自主协同的转型。本届WAIC集中亮相的国产底层技术底座——自研智能体操作系统、本土端侧芯片加速渗透、机器人核心零部件国产化持续提升——表明硬件、系统、大模型、生态四位一体已成为AI终端规模化落地的必要条件。

未来12-18个月产业核心变量预判

基于当前数据趋势与技术实景，以下变量值得持续关注：

  • 技术端：终端专用NPU、轻量化端侧大模型、智能体原生OS将成为旗舰标配；无底层重构的浅层AI设备将逐步被市场淘汰；国产眼镜芯片加速渗透以高通AR1为主的市场格局。
  • 市场端：AI智能体手机价格将持续下探覆盖全价位；1500-3000元智能眼镜将开启价格战；内容电商线上种草加速眼镜消费转化，线下专业验配渠道潜力待释放。
  • 政策生态端：AI终端安全、隐私合规规范持续收紧，合规能力不足的中小品牌将加速出清；政企、办公、出行垂直场景成为智能体商业化核心赛道，产业扶持政策持续降低消费门槛。

不同角色的行动参考

以上数据与判断，对不同类型参与者的含义各有侧重：

终端厂商及产业链企业：竞争逻辑已从参数比拼转向软硬一体生态建设。需重点关注：是否具备完整的”模型训练-数据采集-落地反馈”迭代闭环；在垂直场景中是否形成了可验证的解决方案；端侧芯片、操作系统等底层能力是否自主可控。

行业投资者：赛道分化趋势已明确。需关注具身智能机器人、智能眼镜等增速领先品类的结构性机会；同时留意合规收紧带来的行业洗牌，具备生态壁垒的头部企业抗风险能力更优。

应用开发者与场景方案商：多设备协同与场景智能方向正在打开增量空间。HGR人·镜·犬等混合协同方案的落地，预示着设备联动与服务闭环的想象空间远大于单品智能化。

与IDC进一步交流

以上洞察基于IDC覆盖全球终端市场的持续数据追踪与WAIC现场实景调研。当前时间窗口的关键特征在于：当前时间窗口的特殊性在于：数据转向已经发生，但市场共识尚未完全形成。对于正在进行AI终端战略规划、产品定义或市场进入决策的机构而言，这意味着一个关键的时间窗口。

如您希望就具体赛道数据进行深度探讨，或针对您的业务场景获取定制化的数据与分析支持，欢迎联系我们(请点击此处）进一步交流。

Fiona Wu

Fiona Wu - Associate Vice President, Client System Research and IPDS, IDC China

Analyst Profile
Fiona Wu, associate vice president for IDC China's Client System Research, has 10 years of experience working in the IT industry and is focused on research and analysis of the China IT hardware market. She leads the team to provide…

At the IDC Quanta launch webinar, Jamie Fiorda, SVP of Product Marketing at IDC, laid out why the gap between how fast markets move and how fast organizations understand them is becoming the defining challenge for companies trying to act before conditions shift again. He also laid out where AI-powered intelligence tools fit into closing it.

Fiorda’s starting point was access: who actually gets to use the intelligence a company already has. In most organizations, the market intelligence a company already invests in tends to serve a narrow slice of the business (a research team, maybe a strategy analyst or two), while the insights that could sharpen decisions elsewhere in the company never reach the people making them.

That’s the pattern Fiorda described: intelligence that exists, but sits in a portal most of the organization has no reason to open. The problem isn’t the quality of the research. It’s that the value of it is being rationed by who has access, right as every function in the business is under pressure to move faster on planning that used to have more runway.

Positioning AI as an Intelligence Layer, Not Another Tool

Fiorda’s case for IDC Quanta wasn’t framed as a new subscription or a new dashboard to learn. It was framed as infrastructure: a layer that sits beneath the tools an organization already uses, making the intelligence within them actionable and defensible.

Organizations already have plenty of AI tools. What most of them can’t do is prove their answers are trustworthy. That gap shows up largest in finance: nearly a third of global enterprises still run their finance analytics function with no formal structure at all, department by department, which is exactly where an M&A assumption or a market-sizing input gets challenged first by a board, an investor, or an auditor. Fiorda’s argument centers on embedding intelligence directly into the workflows where strategy is developed, backed by a source a leadership team can point to when that challenge arises.

What This Looks Like Across a Business

Fiorda walked through what that shift means function by function, not as a feature list, but as a picture of how fast an organization can move when intelligence isn’t gated to one team.

A market intelligence team doesn’t change what it does, just how quickly it does it: sizing markets and benchmarking with intelligence flowing directly into the tools they already use.

Finance picks up a use case that’s often entirely new: market sizing inside their own models, risk analysis, revenue validation, all backed by third-party data at the point of decision. IDC research points to real upside here: cognitive technologies applied across due diligence and predictive analytics are projected to drive a 25% increase in M&A returns by 2027, largely because AI-enabled screening surfaces viable targets faster than traditional methods can.

Product gains competitive benchmarking and roadmap validation the same way. Marketing builds segmentation and messaging grounded in actual demand trends instead of assumptions a real shift for CMOs navigating today’s pressure to justify every dollar: IDC research found 52% are already leaning harder into scenario-based planning and 41% report increased pressure to justify marketing ROI, both signs that assumption-driven messaging is no longer good enough. And sales walks into a room with validated, current talking points instead of a battlecard built last quarter.

The strategic point is that all five are pulling from the same source of truth, at the same time, without waiting on each other.

The Question Leaders Must Consider

Fiorda’s close reframed the pitch as a single question for leadership to sit with: if the intelligence an organization already values could reach five functions instead of one, what would that be worth? It’s worth sitting with the scale of that gap: IDC research shows nearly 60% of Chief Data Officers say their organizations need to rethink how they use analytics in decision-making, and only about a third of executives are active users of the intelligence tools already in place. That’s not just a research quality problem. It’s strategic value sitting untouched, simply because of where information lives.

In a landscape changing this quickly, the organizations that move fastest will be the ones where that research actually reaches every team that needs it.

Learn More

Organizations already working with IDC can talk to their account team about which teams could get access next. Those exploring IDC Quanta for the first time can request a demo at idc.com/quanta to see where this fits for your team.

Ryan Smith - Content Marketing Director - IDC

Ryan Smith is the Director of Content Marketing at IDC, where he leads brand-level content and social media strategy, aligning research insights with compelling storytelling to engage technology decision-makers. With a background in both IT and marketing, Ryan brings a unique blend of technical understanding and creative strategy to his work. He’s also a seasoned storyteller, speaker, and podcast host who believes the right message, told the right way, can drive both trust and transformation.

当Token计费账单逐月攀升、实时响应需求倒逼本地推理、数据合规红线日益收紧——端云集中算力的“三座大山”已让企业不堪重负。国际数据公司（IDC）最新预测数据显示，2026年全球Gen AI PC出货量将达0.5亿台、Gen AI手机将达4.32亿台，边缘算力正从概念走向规模化落地。本文为你拆解算力下沉的底层逻辑、终端变局与新兴赛道，并提供三类角色的可执行行动清单。

AI终端产业正告别云端集中算力时代。随着Token计费模式普及、端侧算力芯片性能跃升，以及企业对数据隐私和实时响应要求的持续提高，纯粹依赖云端处理AI任务的模式已触及天花板。IDC最新发布的2026年全球PC和手机市场预测显示，2026年全球Gen AI PC出货量有望达0.5亿台，占整体PC出货量的19.6%，Gen AI手机出货量预计达4.32亿台、占整体手机出货量的39.7%——AI终端已进入规模化普及期。

但终端出货量激增只是表象。更深层的变化在于：端-边-云算力替代端云扩容，正在成为终端产品差异化竞争、企业智能化降本增效的核心路径。理解这场算力架构的深层切换，才能精准抓住AI终端赛道的结构性机遇。


然而，出货量数字背后的驱动力是什么？为什么厂商和用户不约而同地将算力需求从云端“拽”回本地？答案藏在端云模式的三个深层矛盾中。

一、算力下沉的底层逻辑：端-边-云模式的三大优势

当前AI终端产业正从端云集中算力向端-边-云协同架构转型。这一转型不是技术潮流使然，而是端云模式的三大短板已触达临界点：边缘侧产生的成本节约与安全性。

第一，节约成本。 随着Token计费模式普及，企业AI算力支出持续攀升，轻量化AI任务尚可承担，高频次推理场景下云端调用成本已让多数企业难以负荷。当每月Token费用从数千元飙升至数十万元，算力成本的陡峭曲线正在倒逼企业寻找替代方案，边缘侧大大节约成本。

第二，减少延迟。 7×24小时待机、本地化智能响应已成为终端刚需，而端云模式下的网络延迟无法满足高实时性场景——从智能家居的瞬时响应到工业现场的实时决策，云端往返的毫秒级延迟正在成为生产力损耗。边缘侧本地计算减少数据延迟问题。

第三，数据安全。 企业高频次、高隐私的业务数据通过云端处理，合规风险持续放大。尤其是在金融、医疗、政务等敏感领域，数据不出域已是刚性约束。边缘侧完美的解决了以上诸多问题。

二、终端变局：AI终端正演变为边缘算力核心

算力下沉的第一波红利，落在终端设备的形态重构上。

算力下沉带动AI智能硬件迎来高速增长，AI终端市场已进入规模化爆发周期。算力下沉重塑了手机、PC等传统终端定位——它们不再是单纯的数据采集和展示工具，而是具备本地推理能力的边缘算力节点。目前海内外轻量化AI终端密集迭代，AI眼镜、智能耳机等新品持续涌现，AI终端正从边缘侧支持AI原生应用，成为算力支持中心。

这一变局的意义在于，数亿台AI PC和AI手机构成的不是零散的设备群，而是一个分布式边缘算力池。每台设备都是算力的生产者和消费者，端侧算力从“闲置资源”变为“可用产能”。

各厂商将AI PC定位为边缘算力中心，覆盖mini PC、AI手机、高性能笔记本、塔式工作站等多元形态，兼顾普通用户端-边-云算力需求与企业本地化高强度算力部署场景，搭建起分层边缘算力体系。行业竞争重心已从纯云端算力比拼，转向端-边-云算力调度、生态适配与协同优化，为产业带来全新机遇的同时，也对算力分配、硬件适配、算法迭代提出更高要求。

三、新兴赛道：AI小型算力中心激活边缘新潜力

架构切换的窗口期，往往是新赛道爆发的黄金期。

在边缘算力生态中，AI小型算力中心成为新兴优质赛道。传统NAS、服务器只是数据存储和计算设备，而AI小型算力中心彻底升级了这一品类——集成强算力、算法与本地AI处理能力后，AI小型算力中心可独立完成本地模型推理、数据分析，无需将数据上传云端即可实现智能处理。

这一变化带来三重价值：一是降低云端压力，高频本地任务无需消耗Token费用；二是提升响应效率，本地推理消除网络延迟；三是保障数据安全，敏感数据全程不出本地设备。AI小型算力中心打通了云边端一体化算力链路，在家庭场景中可支撑智能相册管理、本地知识库检索，在中小企业场景中可满足私有化文档处理、轻量级模型微调等需求，适配范围广泛且落地门槛低。
终端变了，赛道出来了——但这些趋势对三类不同角色意味着什么？答案不是统一的，而是分层分场景的。

四、行动指南：三类角色的差异化策略

端侧迁移是AI发展的必然趋势，边缘计算机遇与挑战并存。未来能否搭建成熟的边缘算力体系，将成为厂商产品突围、企业降本增效、从业者把握赛道红利的关键。以下针对三类核心角色，给出具体可执行的行动建议。

硬件厂商：用分层产品替代同质化内卷

厂商需摒弃同质化产品研发，针对三类用户群体打造分层产品：

  • 消费市场：侧重轻量化、低功耗、高适配的AI手机、mini PC、AI小型算力中心设备，适配云端订阅算力模式。
  • 中小商户：提供一体化的边缘算力解决方案，以AI小型算力中心为核心节点，降低本地AI部署的技术门槛和运维成本。
  • 大型企业：重点布局高性能AI PC、塔式工作站，强化本地算力与算法集成能力，满足高强度本地推理场景需求。

具体可落地的第一步： 盘点现有产品线中哪些具备“本地推理能力”标签，将其从“功能卖点”升级为“独立算力节点”来定义产品定位，围绕端-边-云协同重构产品路径。

企业用户：用混合算力替代盲目云端扩容

企业无需盲目采购高端算力设备，可采用“云端通用算力+边缘端专属算力”的混合模式：

  • 常规轻量化AI任务依托云端Token订阅处理。
  • 高频次、高隐私的业务数据通过本地AI设备处理。
  • 以此平衡算力成本、响应速度与数据安全三重目标。

具体可落地的第一步： 梳理企业当前所有AI应用场景，按“实时性要求（高/中/低）”和“数据敏感度（高/中/低）”两个维度画出四象限矩阵，将落在“高实时+高敏感”象限的任务优先迁移至端侧处理，直接砍掉对应的云端Token预算。

最终用户：聚焦边缘生态，把握细分赛道红利

消费者可重点关注AI终端适配、边缘算法优化、算力调度、AI小型算力中心落地等细分领域，依托终端硬件爆发红利，布局个人AI场景化应用解决方案。

IDC洞察

AI产业下半场的核心红利不在端云扩容，而在端侧边缘算力的落地与生态完善。AI PC、AI手机的规模化爆发重构了终端算力需求，AI小型算力中心等新赛道持续补齐场景短板，端-边-云协同将彻底替代纯端云模式。短期来看，行业仍需磨合端-边-云算力配比与落地标准；长期来看，边缘算力的普及将进一步降低AI应用落地门槛，释放全场景智能化升级红利。

IDC中国助理研究副总裁武止戈认为，未来随着端侧生态持续完善、算力硬件不断迭代，边缘计算将成为AI规模化落地的核心底座，重塑整个智能产业的发展格局。

进一步交流

IDC中国终端系统研究团队将持续提供AI终端研究、端-边-云行业发展洞察、为厂商、渠道、企业用户提供可落地的边缘算力布局决策方案。如果你需要最新IDC终端产品市场规模预测、赛道竞争格局分析、厂商专属咨询、或是想要精准评估自身产品客户满意度（NPS）可联系我们分析师团队(点击此处），我们可提供一对一深度咨询服务，帮你规避端云算力内卷，精准抓住端-边-云布局确定性增量红利。

Fiona Wu

Fiona Wu - Associate Vice President, Client System Research and IPDS, IDC China

Analyst Profile
Fiona Wu, associate vice president for IDC China's Client System Research, has 10 years of experience working in the IT industry and is focused on research and analysis of the China IT hardware market. She leads the team to provide…

What Happened in China’s Smartphone Market in Q2 2026?

China’s smartphone shipments came in at roughly 66 million units in the second quarter of 2026, down 4.3% year over year and the fifth straight quarter of decline. Rising memory and component costs pushed most Android vendors to raise prices, which cooled upgrade demand. Huawei and Apple were the exceptions, growing close to 20% and 25% respectively.

China Smartphone Market, Top 10 Companies Market Share, and YoY Growth, Q2 2026 (Preliminary results, shipments in millions of units)
Company2Q26 Market Share2Q25 Market ShareYOY Growth
1. Huawei22.6%18.1%19.4%
2. Apple18.1%13.9%24.4%
3. OPPO16.0%17.3%-9.7%
3. vivo16.0%17.0%-11.4%
5. Xiaomi12.4%15.1%-21.7%
6. Honor11.3%11.9%-9.5%
7. Wiko1.1%2.1%-49.2%
8. Lenovo0.3%0.4%-10.4%
9. ZTE0.3%0.4%-33.5%
10. Samsung0.1%0.4%-60.8%
Others1.8%3.4%-50.3%
Total100.0%100.0%-4.3%
Source: IDC Quarterly Mobile Phone Tracker, July 14, 2026
Note:
• Data are preliminary and subject to change.
• All figures are rounded off.
• IDC declares a statistical tie in the Smartphone market when there is a difference of one-tenth of one percent (0.1%) or less in the shipment shares among two or more companies.
• vivo includes vivo and iQOO; OPPO includes OPPO, OnePlus, and realme.

 What Drove the Quarter

The pressure this quarter traced back to a few causes.

  • Memory and other core component costs climbed from late March onward, and most Android vendors responded by raising prices or trimming configurations. That directly dampened consumers’ willingness to upgrade.
  • The lift from government subsidies faded, removing a prop that had supported demand in earlier quarters.
  • Both forces showed up plainly during the “618” shopping festival, where overall smartphone sales fell close to 15% year over year.

Huawei and Apple moved in the opposite direction for reasons that were just as concrete. Both held prices steady while the rest of the Android field raised them, and both layered on targeted promotions. Huawei kept widening its lineup to cover more of the market, while Apple’s early signaling of price increases on its second-half products pulled some buyers forward into the iPhone 17 series sooner than they might have bought otherwise. Strong brand pull helped as well.

The top six vendors now hold about 96% of the market, which keeps squeezing smaller and mid-sized brands and pushes concentration higher.

China Smartphone Market at a Glance, Q2 2026

  • Total shipments: about 66 million units, down 4.3% year over year
  • Fifth consecutive quarter of year-over-year decline
  • First-half 2026 shipments: about 134 million units, down 4.2% year over year
  • Huawei and Apple: each up roughly 20% year over year
  • “618” festival smartphone sales: down close to 15% year over year
  • Vendor tiers: Huawei and Apple lead, followed by OPPO and vivo, then Xiaomi and Honor close behind
  • Top six vendors: about 96% of total market share

Analyst Insight

“Huawei and Apple held their prices steady while competitors were raising theirs, and that gave hesitant buyers a reason to go ahead and purchase in a quarter when most of the market was giving them a reason to wait,” says Arthur Guo, Research Analyst, Client Devices Research at IDC China.

What Is IDC’s Outlook?

Vendors have been cushioned so far by earlier low-cost component inventory, and as that runs down, cost pressure should land more heavily in the second half of the year. On current trends, the year-over-year decline in China could widen to around 20% in the second half of 2026.

Looking into 2027, the market faces continued headwinds, since storage pricing is unlikely to correct in a big way. Even so, there is room for measured optimism. Consumers are postponing upgrades rather than walking away from smartphones, so that delayed demand should return in time. The smartphone remains the one device almost everyone carries and relies on daily, so its core position holds. A recovery looks likely around 2028–2029 as a fresh replacement cycle comes due.

One scenario worth watching is how deeply software developers and large-model AI companies get involved, and whether some enter the market directly. The “AI agent” phone could become the direction that breaks the current impasse, though real AI adoption depends on four elements working together: hardware, operating system, ecosystem, and large models. Only when those four are integrated closely does AI move from concept to daily use, and that is what could eventually spark the next wave of upgrades.

FAQs

Why did China’s smartphone market decline even with new launches and AI features? The main drag was price. Rising memory and component costs led most Android vendors to raise prices from late March, and with government subsidy support fading, hesitant buyers chose to wait rather than upgrade.

Which vendors benefited most in Q2 2026? Huawei and Apple, growing about 20% and 25% year over year respectively. Both held prices steady while competitors raised theirs, ran targeted promotions, and drew on strong brand loyalty. Apple also saw purchases pulled forward ahead of expected second-half price increases.

What could change the market’s direction after 2026? Component costs, particularly storage, will shape the near term, with pressure likely running into 2027. Longer term, capable AI agent experiences that integrate hardware, software, ecosystem, and large models could revive upgrade demand around 2028–2029.

Learn More

Read the latest press release on preliminary data from the Worldwide Quarterly Mobile Phone Tracker or for the full data set and vendor-level detail, visit here.

Kiranjeet Kaur - Associate Research Director - IDC

Analyst Profile
Kiranjeet Kaur is an associate research director for IDC Asia/Pacific. She is involved in building IDC's successful research tracker programs and producing core market data used across all of IDC's Asia/Pacific research reports and consulting projects. Based in Singapore, Kiranjeet is part of a team that manages the quarterly Mobile Device Tracker where she is responsible for sizing and forecasting the Smartphone and mobile phone markets on a quarterly basis, examining competitive trends and studying end user trends.