办公和协同软件市场正在经历从“AI助手”向“AI Agent”和“AI原生工作流”的剧烈转型。各组织正快速从孤立的AI试点项目,转向企业级规模化部署,从根本上重塑人力与智能系统之间的工作分配模式。这一转型凸显了人类独有的核心能力——判断力、创造力与关系管理能力,而AI智能体则负责处理常规性、重复性及知识密集型任务。进入2026年,这个市场的硝烟已经从简单的对话框延伸到了企业的核心业务链条中。OpenClaw这样的AI Agent开源项目的流行标志着AI原生的智能化工作平台在企业数字化转型中的加速落地:其以多Agent驱动、自动化流程和深度生态集成为核心,极大提升了团队协作效率和创新能力。AI Agent不仅重塑了智能化工作平台的用户体验,还推动了市场对AI驱动工作流、安全合规和开放集成生态的高度关注。

国际数据公司(IDC)于20266月发布的《中国智能化工作平台In-App AI Agent评估, 2026对于智能化工作平台领域中的AI助手/Agent能力进行评估,希望通过对中国市场中主要产品技术提供商的产品评测以及对最终用户的客观访谈来帮助市场更加全面地了解中国智能化工作平台AI能力的发展现状,能力特点和应用情况以及未来的技术发展趋势。IDC定义的智能化工作平台是指通过整合企业通讯、协同办公应用、生产力工作套件以及AI助手/AI Agent等数字化工具,帮助个人和组织高效、有效地完成任务或工作,实现目标的综合性平台。评估的内容包括智能办公能力、AI Agent流程自动化能力、集成和跨平台能力、基安全合规/Agent执行安全能力、行业解决方案和客户、AI生态建设等维度。

IDC通过这次评估观察到,智能化工作平台头部厂商的AI能力已经迈向逐渐成熟的发展阶段。领先厂商已经能够支持智能写作、会议纪要生成、文档总结、知识库问答、数据收集和分析、日程与任务建议、跨应用信息检索,以及基于用户指令调用部分办公工具完成稍有难度的任务。这说明AI能力正在从辅助内容生产,逐步走向办公流程中的任务协同。但是从一些评估维度来看,AI Agent流程自动化能力、Agent执行安全和AI生态建设整体得分偏低,说明该领域仍在快速探索阶段。模型稳定性、任务理解能力、流程编排能力、权限控制机制、执行边界管理以及生态伙伴协同机制尚未完全成熟,不同厂商之间的能力差异也比较明显。总体来看,当前市场的竞争核心正在从传统协同工具转向AI驱动的平台能力。未来市场竞争的关键,将不再只是AI功能的数量,而是厂商能否让AI Agent真正进入企业业务流程,并在可控、安全、可审计的前提下实现规模化应用。

基于本次评估情况和对未来趋势的判断,IDC认为,未来中国智能化工作平台市场将有以下发展趋势:

1. AI原生与多Agent驱动的智能化工作平台将成为主流

未来的智能化工作平台将以AI为核心驱动力,原生集成多模态协作、智能代理(Agent)、自动化、知识管理等能力。AI不仅承担任务自动化、智能推荐、内容生成,还将通过多Agent协作实现流程编排、知识流转和业务决策支持。IDC预测,企业将逐步从“AI插件”过渡到“AI原生”平台,AI与人、AI与AI之间的协作将成为提升创新力和敏捷性的关键。Agentic协作将推动企业从单一任务自动化迈向端到端的智能业务流程重塑。

2. 多模态与可视化协作体验持续深化

随着智能化办公场景的复杂化,智能化工作平台正快速融合文本、语音、视频、视觉画布、手势等多种交互方式。可视化协作(如数字白板、流程画布、可视化工作流)和“多人游戏化”体验成为主流。根据IDC全球调研数据,82%的企业认为视觉协作显著提升了创新与决策效率。未来,平台将进一步支持XR、智能硬件(如AI眼镜)、多语言和无障碍访问,实现全员、全场景的沉浸式协作体验。

3. 数据治理、信任与合规成为平台落地前提

随着AI和多方协作的深入,数据安全、隐私保护、合规和AI治理成为平台采购和部署的核心考量。平台需支持细粒度权限,权限继承自动化,数据驻留,合规认证(如GDPR、ISO等),合规审计可追溯,并通过透明的数据政策和治理机制建立员工与组织间的信任。IDC强调,信任是数字协作的基石,缺乏信任的协作空间难以实现持续创新和高效运营。

4. AI治理与FinOps兴起

随着数字员工增多,企业开始担心“AI乱花钱”或“AI越权”。因此,海外出现了专门审计AI Token消耗和AI安全合规的工具。Token成为财务成本,企业开始像审计水电费一样审计AI的Token消耗量。厂商需要推出更精准的ROI分析工具,用来证明AI助理到底帮员工省了多少时间。

5. 平台化与低代码深度融合

智能化工作平台正在从沟通和流程工具,升级为企业统一的工作入口。它可以通过连接ERP、CRM、HRM、财务和供应链等系统,把组织协作、业务流程和数据处理整合到同一界面中。与此同时,低代码能力将与AI和AI Agent深度结合。业务人员可以用自然语言描述需求,由AI辅助生成表单、流程、报表和轻量级应用;AI Agent则可以根据业务规则自动触发流程、分派任务、同步数据和提醒异常。

分析师观点

IDC
中国助理研究总监李昭表示,AI在办公场景中的演进,正在从“提升个体效率”加速走向“重构办公执行体系”。以OpenClaw为代表的新一代Agent框架,已不再局限于知识问答、内容生成、沟通协作、会议助手、数据和表格处理、日程规划等传统助手能力,而是进一步具备跨应用调用、连续任务执行和自主编排流程的特征,这标志着企业办公AI正从“会说”走向“会做”。软件不再只是工具,而是数字员工。AI Agent不再是被动响应,而是具备了规划能力,能自主跨应用完成复杂任务。与此同时,近期市场变化也表明,Agent的价值正在快速释放,但其规模化落地已不再只是技术问题,而是进入“技术能力、治理能力与成本模型”三重约束并行的新阶段。

进一步交流

如果您希望进一步了解Agent在企业软件中的落地路径、市场演进趋势或对自身业务的具体影响,欢迎与IDC分析师团队联系(点击此处)。IDC将基于持续的市场跟踪与研究,提供更具针对性的洞察与建议,支持企业与厂商在这一轮变革中做出更有前瞻性的决策。

Lizzie Li

Lizzie Li - Associate Research Director

Lizzie Li is Associate Research Director of IDC China's Enterprise System and Software Research that focuses on research and analysis of the China Datacenter, Cloud Computing, and IT infrastructure markets. She also provides intelligence and consulting services in customized projects for…

この記事でわかること:AIスーパーサイクルとは何か、そして日本にとって何を意味するのか。AIベンダーと日本企業のバイヤーの間にある最大のギャップとは何か。AIはB2Bテクノロジーの購買意思決定をどう変えているのか。

2026年6月23日、IDC Directions Tokyo 2026の会場には特別な熱気が漂っていた。日本のAIの転換点は、もはや「近づいている」段階ではなく、「すでに到来した」——それを物語る空気だった。IDCシニアバイスプレジデントのSandra Ngが伝えた中心的なメッセージは、警鐘であり、同時に進むべき道筋でもあった。彼女は率直にこう語った。問うべきは、AIが日本市場を変えるかどうかではない。誰が最短時間で最大の価値を生み出せるか、なのだ。

IDC Directions Tokyo 2026には約400名が参加し、そのうち65%が部長職以上。単なる情報収集の場ではなく、意思決定者が集う場となった。アナリストとの1対1セッションは軒並み満席となり、フロアから相次いだ質問の多さが、議論の核心を突いていたことを物語っている。

スーパーサイクルは現実だ。そして日本は、いま追う立場にある

何が起きているのか、そのスケール感を伝える数字がある。2025年、世界のIT支出は4兆2,000億ドルという市場規模の中で14%成長した。これは1996年以来、最も高い成長率だ。ただし当時、同じ14%成長が適用されていた市場規模はわずか7,000億ドル。桁がまるで違う。(出典:IDC Worldwide Black Book、2026年3月31日)

これこそがAIスーパーサイクルだ。中国・台湾・インドが「AIスーパーパワー・ビルドアウト」や「デジタルネイティブ・スケーラー」として競争を加速させる一方、IDCは日本を韓国とともに「レガシー・モダナイザー」と位置づけている。つまり、AIの価値にたどり着くまでの道筋が、モダナイゼーション(近代化)を必ず経由する市場ということだ。これは不利な条件ではなく、日本ならではの戦略的な立ち位置である。

日本のAI市場も、この勢いをそのまま映し出している。国内のAIインフラ支出は2026年末までに90億ドルに達し、2029年にかけて年平均24%(CAGR)で成長する見通しだ。AIサービスは2030年までに80億ドル、2025年比で3.2倍に伸びると予測されている。さらにアジア太平洋地域のCEOの61%が、エージェント型AIと生成AIの大規模活用を、新規投資の最優先分野に挙げている。(出典:IDC Worldwide AI and Generative AI Spending Guide、Forecast V1 2026;IDC CEO Survey、2026年3月)

ベンダーとバイヤーの間にある断絶が、双方にとって損失になっている

ここから議論は、耳が痛いが避けて通れないテーマへと移った。

Sandra Ngは、日本企業のバイヤーが本当に求めているものと、テクノロジーベンダーが実際に提供しているものとの間にある、4つの根本的なギャップを指摘した。

ギャップ①——ビジネスケースの明確さ
バイヤーが求めているのは、自社の業界に根ざした成果であり、グローバル共通の一般的なベンチマークではない。「日本のメーカーにとって、これは具体的に何を意味するのか見せてほしい」——このニーズは、海外の事例紹介スライドを並べるだけでは満たせない。

ギャップ②——AIの総所有コスト
このテーマは、通常の議論よりもう一段深く掘り下げる必要がある。日本でAIを導入する際の隠れたコスト——データ準備、システム統合、継続的なモデルメンテナンス、METIガイドラインへの準拠状況の監視、チェンジマネジメント——は、表示されている技術価格に対して50〜70%以上も積み上がるのが当たり前になっている。これを事前にはっきり示さないベンダーは、CFOとの商談で信頼を失っている。

ギャップ③——AIガバナンスへの準備
日本の規制環境は、決して名目だけのものではなく、独自の色合いを持つ。金融庁のAIガイダンス、METIのAIガバナンスフレームワーク、個人情報保護法(APPI)——いずれも高い基準を求めている。バイヤー側の動きの速さは、多くのベンダーのガバナンス対応が追いつけていないほどだ。

ギャップ④——「実例を見せてほしい」という瞬間
日本企業のバイヤーが求めているのは、国内のリファレンス顧客、導入までの具体的な期間、そして数字で示せる成果だ。海外の成功事例ではない。それを提示できないベンダーは、検討リストから静かに外されていく。

日本のトップ企業は、実際にどう動いているのか

これらのギャップは、机上の話ではない。2025〜2026年にIDCが実施したCxOへのヒアリングからは、日本を代表する先進企業がこの課題にどう向き合っているか、具体的な姿が見えてくる。

  • トヨタは、ハイパースケーラーのパートナーとともに自社開発のAIプラットフォームを構築し、年間1万時間分の手作業を削減した。AIを活用した品質システムによって塗装の不良率を25%減らしたことも、実績として記録されている。
  • 東京海上は、METIと金融庁のガイダンスを踏まえ、文書・画像分析へのAI導入に先立って、透明性・人による監視・バイアス排除・データ保護・運用面での信頼性という5つの柱からなるAIガバナンスポリシーを先に公表した。ガバナンスを先に固め、導入はその後——これが、日本の規制当局がいま求めている順序だ。
  • ヤマト運輸は、160万社の法人顧客と4,000社を超える物流パートナーからなるサプライチェーン全体に、AIによる最適化を導入した。80ルートに及ぶ中継拠点の最適化を通じて、人件費65%削減、温室効果ガス排出量42%削減を目標に据えている。

いずれも実験段階のパイロットではない。数字として記録された成果を伴う、本番環境での導入だ——まさに、バイヤーが求めているリファレンスストーリーそのものである。

バイヤーは、すでにAIを使ってあなたの会社を調べている

Sandra Ngのプレゼンテーションの中でも、とりわけ印象的だったデータがある。世界のテクノロジーバイヤーの84%が、「今後12ヶ月でAIが自社の技術購買のやり方を変える」と答えているのだ。(出典:IDC B2B Technology Buyer Survey 2025、WW n=406)

購買までの流れそのものが変わりつつある。今の順番はこうだ:AI検索 → ベンダーのウェブサイトで確認 → 同僚やピアによる検証 → チャネルパートナーへの相談 → 候補リストの確定。CIOやCFOがChatGPTやGoogle Geminiを使って調査したとき、あなたのブランドとソリューションが明確に浮かび上がらなければ、競争に参加していたことすら知られないまま、選考から外れているということになる。

だからこそ、AEOAnswer Engine Optimization:回答エンジン最適化)とGEOGenerative Engine Optimization:生成エンジン最適化)は、もはやマーケティングの実験ではなく、事業の存続を左右する能力になっている。2027年までに、日本企業の35%が体系的なAIガバナンスを整備すると予測されている。AIが生成する回答の中で、露出度が高く、信頼され、戦略的に進めているベンダーには、時間とともに積み重なる優位性が生まれる。(出典:IDC FutureScape 2026 – AI-Fueled Business Strategies、Japan)

エージェンティックAI:次の競争フロンティア

業界全体で起きている転換が、なぜスピードが重要なのかを裏付けている。IDC Japanの植村 卓弥が強調したように、AIスーパーサイクルは今、投資の第2波——インフラ構築からエンタープライズアプリケーション・サービスの採用へ——に入りつつある。エンタープライズ向けAIプラットフォーム・アプリ・サービスへの世界の支出は、2026年の4,000億ドルから、2029年には1兆ドルに達する見込みだ。(出典:IDC Worldwide AI and Generative AI Spending Guide V1 2026)

注目すべきは、調査対象バイヤーの83%が「AIエージェントによってサプライヤーの乗り換えが以前より容易になった」と回答していることだ。市場がコモディティ化する前に、成果ベースの関係をいま築いておけるベンダーこそが、次の競争優位を決めることになる。

今すぐ取るべき3つのアクション

Sandra Ngが最後に示したフレームワークは、シンプルで、すぐに動き出せる内容だった。

今すぐ行うこと:
一般的なベンチマークは捨て、日本国内のリファレンス顧客1社、指標1つ、導入タイムライン1つに置き換える。GTMのピッチを、CIOだけでなく購買委員会全体を意識した内容に組み直す。AIの中での自社の見え方(検索可視性)のギャップを、今すぐ修正する。

今年中に行うこと:
ROIの透明性を競争力の武器にする。実際に導入できるエージェント型ワークフローのストーリーをつくる。AIが生成する回答の中で引用されるよう、AIに最適化されたサードパーティコンテンツに投資する。

長期的に投資、実施すること:
AIガバナンス、トラスト、コンプライアンスを、新たな収益の柱として位置づける。あるいは、コンプライアンスを単なる話題ではなく、実際の成果として提供できるベンダーになる。カテゴリー、ペルソナ、地域ごとに「アンサーシェア」の計測を始める。SEO+AEO+GEOを組み合わせた、重層的なディスカバリー戦略を実行する。(出典:IDC C-suite Survey、2025年9月、APJ、n=300)

日本のAIスーパーサイクルは待ってくれない

日本のAI市場でリーダーの座を確立するための窓は、今現在は開いている——ただし、それが開いたままでいる保証はない。実験段階からいち早く成果ベースの導入へ進む企業、4つのギャップを埋めるベンダー、そしてAI主導の購買ジャーニーの中で自社を見つけてもらえる組織。これらこそが、日本のテクノロジー市場の次の時代をつくっていく。

東京で始まったこの対話は、大阪へと続く。

2026728日開催のIDC Directions Osakaでは、日本のAIスーパーサイクルが各業界をどう再編しているのか、次の価値創出の波がどこで生まれるのか、そしてあなたの組織がその一歩先を行くにはどうすればよいのかを、データとともにさらに深く掘り下げます。今すぐIDC Directions Osakaにご登録ください。席数には限りがあります。

Note:本記事は202671日に英語で公開されたブログの抄訳です。原文は以下からご覧いただけます:https://www.idc.com/resource-center/blog/japans-ai-supercycle-is-here-are-you-ready-to-lead-it/

Mike de la Cruz - Corporate Communications Director - IDC Asia/Pacific

Mike de la Cruz is Corporate Communications Director for Asia Pacific at IDC, bringing over 25 years of career experience in marketing and communications for the information technology industry. He shapes and amplifies IDC's research-driven narratives, positions executives and analysts as authoritative industry voices, builds relationships with top-tier technology and business media across the region, ensures consistent brand voice and positioning, develops content that drives audience engagement, and leverages social media and digital communications to extend IDC's reach.

On June 22, 2026, the White House made quantum computing a formal US national priority, releasing two executive orders that position quantum computing and post-quantum cryptography at the center of US industrial policy and national security strategy.

The urgency is evident in the data. IDC surveyed 535 US government organizations and found that 32% are currently running quantum computing pilots; among federal agencies specifically, that figure rises to 49%, with another 40% planning to launch pilots within the next 12–24 months. This is no longer theoretical exploration; it reflects active government commitment to quantum technology development.

Two executive orders target quantum computing

The first executive order, “Ushering in the Next Frontier of Quantum Innovation,” establishes the Quantum Computer for Application Development and Discovery Science (QC-ADDS) program. This initiative directs the Department of Energy, in coordination with Defense, Commerce, and intelligence agencies, to develop quantum computing systems capable of solving problems beyond classical computing capabilities. Potential applications include climate modeling, advanced signal processing, and defense logistics optimization.

The timeline is demanding: the Department of Energy must publish technical specifications within 90 days and explore private-sector partnership models within 180 days. The Pentagon is directed to field at least three quantum sensor projects by September 2028. The order also addresses critical infrastructure gaps, including stronger protections for domestic quantum supply chains, expanded counterintelligence measures for quantum technology, and a mandate for a government-wide quantum workforce strategy.

The second executive order addresses the defensive imperative: “Securing the Nation Against Advanced Cryptographic Attacks.” It confronts a specific and present threat: adversaries may be collecting encrypted US data today, banking on the ability to decrypt it once large-scale quantum computers become operational.

This tactic is known as “harvest now, decrypt later”: the practice of stockpiling encrypted data today in order to decrypt it later, once the technology catches up. It demands immediate action.

The order establishes a comprehensive post-quantum cryptography (PQC) migration strategy with clear deadlines. All high-value federal assets and high-impact systems must adopt PQC for key establishment by December 31, 2030, and for digital signatures by December 31, 2031. Federal contractors face identical compliance timelines, enforced through updates to the Federal Acquisition Regulation. NIST will begin pilot migration on its own systems by the end of 2027.

Quantum computing investment and market growth

These orders arrive at an inflection point for quantum technology.

IDC’s Worldwide Quantum Computing Forecast, 2025–2029 projects total spending will reach $17.3 billion by 2029, representing a 43% compound annual growth rate. Government investment has been a primary driver, with quantum spending increasing 37% between 2024 and 2025—gains attributed to advances in error correction and modular quantum architectures.

These executive orders follow the Department of Commerce’s May 2026 announcement of $2 billion in CHIPS Act funding for quantum manufacturing, signaling a pivot toward industrial-scale ecosystem development beyond early-stage research funding. Priority investment areas include manufacturing, packaging, cryogenics, and control systems: the engineering challenges that will determine whether quantum computing can achieve scalable commercial deployment.

Three factors that will determine quantum policy success

The effectiveness of these executive orders depends on three elements:

  • Funding continuity: Executive orders require sustained appropriations and bipartisan Congressional support to insulate quantum policy from electoral cycles. Organizations must prioritize use cases that deliver measurable return on investment within 12–24 months. Academia and private industry must also establish collaborative partnerships to share investment risks and accelerate development.
  • Organizational readiness and talent: The 90-day deadline for technical specifications will clarify the government’s near-term expectations. More fundamentally, both government and industry face significant quantum talent shortages. The current dependence on a limited number of advanced research centers and specialized technology companies represents a strategic vulnerability that must be addressed through coordinated workforce development.
  • Federal contractor preparation: The PQC compliance mandate for federal contractors, with its December 2030 deadline, has the potential to reshape procurement standards across the broader IT industry. Technology vendors and systems integrators that have already established quantum centers of excellence or actively participated in federal quantum pilots are positioned to realize competitive advantages.

What’s next for quantum computing policy

These executive orders represent a sustained commitment to transition quantum computing from the experimental phase toward operational deployment. The combined focus on innovation (QC-ADDS), defensive security (PQC migration), and supply chain development signals a comprehensive approach to quantum technology as critical infrastructure.

For agencies and contractors, the near-term deadlines are now fixed: publish technical specifications, hit PQC migration milestones, and build the workforce needed to run quantum and classical systems side by side.

Massimiliano Claps

Massimiliano Claps - Research Director

Massimiliano (Max) Claps is the research director for the Worldwide National Government Platforms and Technologies research in IDC's Government Insights practice. In this role, Max provides research and advisory services to technology suppliers and national civilian government senior leaders in…
Ruthbea Yesner

Ruthbea Yesner - Group Vice President, Worldwide Public Sector

Ruthbea Yesner is Group Vice President, Worldwide Public Sector, leading IDC's research teams for the Health and Life Sciences and Government and Education. Her teams deliver research and advisory services on the technologies, strategies, and business models shaping healthcare, life…

Most organizations still pick a security framework the way they did in 2022: find the biggest name, check the box, move on. That approach hasn’t survived contact with the last three years, three new regulations, a finalized quantum-cryptography standard, and an entirely new AI threat surface later.

The current situation: A framework landscape transformed since 2022

Choosing the right security framework has never been more consequential or more complex. When IDC last published comprehensive guidance on this topic in 2022, the landscape was manageable: a stable set of well-known frameworks and a relatively predictable regulatory backdrop. The intervening years have fundamentally changed both dimensions.

Four structural shifts now define what buyers must navigate:

  • NIST CSF 2.0 (February 2024) introduced a formal Govern function, its first major revision in a decade, elevating cybersecurity from an operational discipline to a board governance obligation and broadening scope to all organizations regardless of size or sector.
  • DORA (EU 2022/2554) became fully applicable in January 2025, imposing mandatory ICT risk management and third-party oversight obligations on approximately 22,000 EU financial entities.
  • PQC standards were finalized: NIST released three post-quantum cryptography standards in August 2024 (FIPS 203, 204, 205), transforming quantum readiness from a theoretical concern to an operational imperative.
  • AI has created a new risk surface: NIST published a draft Cyber AI Profile (IR 8596) in December 2025, extending CSF 2.0 specifically to AI-related cybersecurity risks. Organizations that have deployed AI, particularly agentic AI or LLM-integrated workflows, must now factor AI governance into framework selection.

The wrong framework choice, one that exceeds organizational maturity, understates regulatory obligation, or ignores supply chain exposure, produces worse security outcomes than a well-adopted, properly scoped, simpler framework. The good news: there are strong options. The challenge: the decision is more complex than it was three years ago.

Decision-making criteria and methodology

Security framework selection is a risk management decision, not a technical checklist, requiring input from legal, compliance, finance, operations, and the board. IDC’s 2026 methodology starts with crown-jewel data classification, then splits into regulated and non-regulated tracks, now including universal AI governance and quantum-readiness branches. Key criteria include data classification, regulatory obligations, threat landscape, and AI adoption footprint. Organizations deploying agentic AI face risks that general-purpose frameworks don’t address, while harvest-now-decrypt-later (HNDL) exposure demands cryptographic roadmap planning alongside traditional control mapping.

Additional criteria round out the methodology: third-party risk (CSF 2.0’s Govern function and DORA Article 28 set the compliance floor), PQC readiness (FIPS 203-205 are finalized, with 2030 as a planning horizon), and organizational maturity (smaller organizations should start with CIS Controls IG1 rather than overreaching). Budget discipline favors phased, risk-prioritized roadmaps supported by cyber risk quantification. Finally, multi-framework interoperability and GRC technology support are now prerequisites: manual evidence collection across concurrent regimes such as CSF 2.0, ISO 27001, HIPAA, and DORA is no longer sustainable at scale.

Regulated versus non-regulated organizations: Two different journeys

Regulated organizations

For regulated organizations, the regulator largely determines the framework; strategic focus shifts to execution. Key questions: How can multiple simultaneous requirements (DORA + ISO 27001; HIPAA + NIST 800-53) be satisfied without duplicating evidence work? Which GRC platform best automates cross-framework mapping? How should gap closure be sequenced within the budget? Have ICT third-party providers been assessed against CSF 2.0 Govern, DORA Article 28, and NIST 800-161? Mature organizations typically adopt a “framework stack”: CSF 2.0 or ISO 27001 as backbone, NIST 800-161 for high-risk vendors, plus industry-specific overlays.

Non-regulated organizations

Non-regulated organizations must perform more active analysis, with the right starting point depending on maturity. Early-stage or SMB organizations should adopt CIS Controls v8 Implementation Group 1, 56 safeguards achievable with limited staff, mapped to NIST CSF 2.0 for growth. Mature programs or those facing elevated threat exposure should adopt NIST CSF 2.0 or ISO 27001:2022, both of which include a Govern function that supports board-level accountability and SEC disclosure readiness. All non-regulated organizations, regardless of status, should evaluate the AI Governance and Quantum Readiness branches given their present-day risk implications.

AI and quantum: Two criteria that didn’t exist in 2022

Artificial intelligence: risk surface and governance obligation

AI has added two urgent dimensions to the framework selection process. Offensively, adversaries use LLMs for convincing phishing, automated vulnerability discovery, and direct attacks via model poisoning and prompt injection. Assess whether your framework addresses AI-enabled detection and response. Defensively, organizations deploying AI in production, especially autonomous agentic AI, face authorization, auditability, model integrity, and supply chain risks that general-purpose frameworks don’t address. The NIST Cyber AI Profile (IR 8596, December 2025 draft) extends CSF 2.0 across three risk areas and should serve as a supplementary governance layer. Shadow AI and SaaS-embedded AI remain largely unmeasured exposures requiring dedicated governance tooling.

Post-quantum cryptography: from theory to operational imperative

PQC standards are finalized: NIST published FIPS 203, 204, and 205 in August 2024, with a fourth HQC-based standard selected in March 2025. The harvest-now-decrypt-later threat is present today: adversaries are collecting encrypted data now to decrypt once quantum computing matures, creating real exposure for organizations holding financial, healthcare, or critical infrastructure data. NSA’s CNSA 2.0 mandates 2030 migration for National Security Systems; commercial organizations should treat this as a planning horizon, not a start date. Immediate actions include completing a cryptographic inventory, prioritizing long-lived data systems, evaluating vendor PQC roadmaps, and considering hybrid cryptographic approaches.

Essential guidance for the technology buyer

A well-adopted, properly scoped framework always outperforms a theoretically superior one that exceeds organizational capacity. Buyers should structure stakeholder conversations, including legal, compliance, finance, and the board, around the 10 decision criteria, treating AI governance and PQC readiness as first-order, not future-state, considerations. Conduct a cryptographic inventory now and evaluate AI footprint against the NIST Cyber AI Profile. Address third-party risk per CSF 2.0, NIST 800-161, and DORA Article 28. Favor frameworks with strong cross-mapping, invest in automated GRC technology, and adopt cyber risk quantification for CFO-ready budget conversations. Build a phased, five-year roadmap, closing highest-risk gaps first, and recalibrate annually.

“The right security framework is a critical factor in adequately managing security risks, not only those present today, but also those that could emerge in the future. The 2026 landscape demands that organizations evaluate AI governance and post-quantum cryptography readiness as first-order criteria, not future-state considerations.”Philip D. Harris, Research Director, Cybersecurity GRC Solutions, IDC

Guidance for the technology supplier and services provider

The security framework market is in structural transition, and suppliers calibrated to 2022 are selling into a market that no longer exists. Multi-framework compliance automation is now the dominant selection criterion, making unified control libraries across CSF 2.0, ISO 27001, HIPAA, and DORA essential. Suppliers should build DORA as a named capability, establish AI governance credibility now while the Cyber AI Profile remains in draft, and develop a PQC advisory practice anchored to cryptographic inventory services. Value propositions should be reframed financially for CFO-influenced procurement, while midmarket buyers increasingly favor managed compliance wrappers. Roadmap priorities span CSF 2.0/DORA now, Cyber AI Profile alignment in 2027, and full PQC/EU CRA support by 2030.

Philip D. Harris, CISSP, CCSK

Philip D. Harris, CISSP, CCSK - Research Director, Governance, Risk, and Compliance (GRC) Solutions

Phil Harris is Research Director for GRC Solutions at IDC, where he develops and promotes IDC's point of view on risk, advisory, privacy, and compliance services and software. He conducts research on business strategies and the impact of relevant offerings…

Global spending on digital transformation (DX) software is on pace to hit $640 billion by 2029, and where that money goes is shifting fast. IDC’s latest Worldwide Digital Transformation Spending Guide shows AI pulling value out of infrastructure and into applications, with the pace of that shift varying sharply by sector.

Software is becoming the primary engine of digital transformation

Digital transformation (DX) is the broad shift by organizations to embed technology into every layer of their operations, customer experiences, and business models. It spans hardware, services, and software. Data shows software is increasingly where DX investment is concentrating.

Among the three technology groups that make up DX spending, software is the fastest growing. Its share of total DX spend rises from 32% in 2026 to 36% by 2029, with a 21.7% CAGR, leading both services (10.6% CAGR) and hardware (19.8% CAGR). This momentum is increasingly driven by the scaling of AI, which is rapidly shifting value from infrastructure into applications and accelerating demand for AI powered software capabilities.

By 2029, AI will account for roughly 40% of worldwide DX software investment, which is a significant shift from today. The other 60% still flows into business applications, system infrastructure, and development and deployment platforms. The AI investments showing up across every sector in this analysis do not stand alone. They run on top of foundational layers already in place, and in many cases they depend on those layers to deliver value at all. Across the six sectors, the pace of AI adoption varies considerably, and that variation tracks closely with how mature the underlying software stack already is. AI is coming everywhere, but it is arriving on top of what organizations have already built.

Six sectors. Significant scale. Different priorities.

IDC’s Worldwide Digital Transformation Spending Guide (V1 2026) tracks DX software spend across six major industry sectors. Overall software investment is growing strongly in all six, and AI’s share within it is rising – but the pace, scale, and use case priorities differ considerably by sector. What drives DX spending in retail and services is not what drives manufacturing or financial services. Sector context matters.

“The growth in DX software is broad-based. This is not simply an AI spending wave. Organizations are investing across the full software stack to transform how they operate, and AI is an accelerating part of that — not the whole of it.”Mariya Yahnyuk, Research Analyst, Data and Analytics

#1 Retail and services

The Retail and Services sector leads all sectors in total DX software spend. This sector covers both retailers and a range of activities such as hospitality, travel, services, and more. Customer management tops the list of investment priorities: engaging customers across every channel, in real time, with personalized and consistent experiences. Omnichannel commerce and service delivery platforms come second, as both retailers and service providers now operate across physical, digital, and mobile simultaneously.

The third priority is operational intelligence: energy management, workforce scheduling, and efficiency tools that apply equally to store environments and service operations. Many of these investments sit in the applications and infrastructure layers of DX software – the customer data platforms, workforce systems, and integration layers that make AI useful when it arrives. AI is projected to grow to 42% of this sector’s DX software spend by 2029, building directly on that operational foundation.

#2 Financial services

Financial services have the highest existing AI share of any sector, and the broader DX investment picture explains why it got there. Security leads the use case list: detecting cyber threats and preventing fraud are areas where software investment delivers clear, measurable returns, making them natural anchors for DX spending early. Automating core business operations comes next, as financial institutions replace manual, rules-based processes with systems that can adapt and scale.

The pattern here is integration. AI is being built into existing DX programs, the same core applications and infrastructure platforms institutions have been modernizing for years. That is why the spending is sustaining: it is tied to operational outcomes that financial institutions already care about.

#3 Manufacturing and resources

Manufacturing currently has the lowest AI share of any sector in DX terms, which also makes it one of the most interesting to watch. DX investment here is driven by very practical pressures: aging equipment, fewer experienced engineers, and increasingly complex supply chains. The largest spending area is autonomic operations, production environments moving toward systems that can monitor and adjust on their own, recovering from most faults without needing a person in the loop.

Self-healing assets and augmented maintenance follow closely. A third priority, less obvious but growing, is customer and client management: manufacturers are increasingly investing to understand and serve end customers directly, not just to optimize internal operations. Together, these priorities describe a sector using DX investment to reduce operational risk, extend asset life, and build closer market relationships.

“Manufacturing’s lower AI share today should not be read as lower ambition. The use cases driving DX investment are autonomic operations, asset health, customer proximity, that’s exactly where AI delivers durable, measurable value. The growth trajectory reflects that.” —Mariya Yahnyuk, Research Analyst, Data and Analytics

Three actions for technology providers

The DX software data carries clear implications for technology vendors and platform providers. These are not passive trends to monitor – they are signals that should shape how you position, sell, and support your customers.

Help your customers understand where they stand. Most organizations do not have a clear view of how their software investment compares to peers in their sector. Use sector-level DX spending data to show them where investment is concentrating, where they may be behind, and what use cases are driving results for similar organizations. Then recommend tools that fit where they are in their journey.

Make the case for platform integration using evidence. Organizations that rely on fragmented point solutions face growing complexity as DX programs scale, and slower AI outcomes as a result. Many of your customers have not made that connection yet. Use the spending data trends to show them concretely that platform integration is delivering better results and why consolidation is the more practical path forward.

Treat data readiness as a customer success issue, not a prerequisite. DX outcomes and AI outcomes depend on the quality and accessibility of underlying data. Offer data quality and governance support as part of the engagement, so customers build data readiness while the work is underway.

Want the full picture?

This analysis draws on IDC’s Worldwide Digital Transformation Spending Guide (V1 2026), which covers DX investment across all six industry sectors, further detailed by 27 industries, 12 technology markets, and geographies through 2029. The full research covers Healthcare, Infrastructure and Energy, and Public Sector in depth alongside the three sectors featured here.

Mariya Yahnyuk

Mariya Yahnyuk - Research Analyst, Data and Analytics

Mariya Yahnyuk has been a research analyst in IDC’s Worldwide Data and Analytics team since 2022. Yaknyuk supports the development of IDC's Spending Guide portfolio, assuring alignment with technology and market changes, relevancy, and business value for customers Mariya directly…

Key questions answered in this article: What is the AI supercycle and what does it mean for Japan? What are the biggest gaps between AI vendors and enterprise buyers in Japan? How is AI changing B2B technology buying decisions?

The energy in the room at IDC Directions Tokyo 2026 on June 23 said it all: Japan’s AI moment is no longer approaching — it has arrived. And the central message from IDC’s Senior Vice President Sandra Ng was both a wake-up call and a roadmap. The question, as she put it bluntly, is not whether AI will reshape the Japan market. It is: who will create the most value in the shortest time?

IDC Directions Tokyo 2026 brought together close to 400 attendees, with 65% at Director level or above, a room full of decision-makers, not just observers. Analyst 1-on-1 sessions were fully booked, and the volume of questions from the floor made one thing clear: the conversation was hitting a nerve.

The Supercycle Is Real and Japan Is Playing Catch-Up

To understand the scale of what is happening, consider this: global IT spending in 2025 is growing at 14% on a $4.2 trillion market base, the strongest IT spending growth since 1996. Back then, that same 14% growth rate applied to a $700 billion market. The numbers are in a different league entirely. (Source: IDC Worldwide Black Book, 31 March 2026)

This is the AI supercycle. While regions like China, Taiwan, and India are racing ahead as “AI Superpower Built-Outs” or “Digital Native Scalers,” IDC positions Japan alongside Korea as a Legacy Modernizer, a market where the path to AI value runs directly through modernization. That is not a disadvantage. It is a specific strategic imperative.

Japan’s AI market reflects this momentum. Domestic AI infrastructure spend is expected to reach $9 billion by end of 2026, growing at a 24% CAGR through 2029. AI services are projected to hit $8 billion by 2030 or 3.2x the 2025 figure. Moreover, 61% of CEOs in Asia Pacific named agentic and generative AI at scale as their number-one new investment priority. (Source: IDC Worldwide AI and Generative AI Spending Guide, Forecast V1 2026; IDC CEO Survey, March 2026)

The Vendor-Buyer Disconnect Is Costing Everyone

Here is where the conversation got uncomfortable and important.

Sandra Ng identified four persistent gaps between what enterprise buyers in Japan need and what technology vendors are actually delivering:

Gap #1 — Business case clarity. Buyers want outcomes grounded in their industry, not generic global benchmarks. “Show me what this looks like for a Japanese manufacturer” is a very different ask from a global case study slide.

Gap #2 — Total cost of AI ownership. This one deserves more attention than it typically gets. Hidden AI implementation costs in Japan — covering data preparation, system integration, ongoing model maintenance, compliance monitoring with METI guidelines, and change management, routinely add 50–70% beyond the headline technology price. Vendors who do not address this upfront are losing trust at the CFO conversation.

Gap #3 — AI governance readiness. Japan’s regulatory environment is real and specific. The FSA AI guidance, METI’s AI governance framework, and the Act on Protection of Personal Information (APPI) set high standards. Buyers are moving faster than many vendors’ governance answers can keep up.

Gap #4 — The “show me” moment. Japanese enterprise buyers want a local reference customer, a deployment timeline, and measurable outcomes, not a global success story. Vendors who cannot produce this are losing shortlist positions.

What Japan’s Leading Enterprises Are Actually Doing

These gaps are not theoretical. IDC’s CxO conversations in 2025–2026 point to concrete examples of how Japan’s most forward-thinking companies are navigating them:

  • Toyota built a proprietary in-house AI platform with a hyperscaler partner, saving 10,000 hours of manual work annually and documenting a 25% reduction in paint defects through AI-assisted quality systems.
  • Tokio Marine responded to METI and FSA guidance by publishing a 5-pillar AI governance policy first, covering transparency, human oversight, bias elimination, data protection, and operational reliability, before deploying AI for document and image analysis. Governance policy first, deployment second—this is the sequence Japan’s regulators now expect.
  • Yamato Transport used AI-driven supply chain optimization across 1.6 million corporate customers and 4,000+ logistics partners, targeting a 65% reduction in labor costs and a 42% cut in GHG emissions by optimizing relay points across 80 routes.

These are not pilots. These are production deployments with documented outcomes—exactly the reference stories buyers are demanding.

Your Buyers Are Already Using AI to Research You

One of the most striking data points from Sandra Ng’s presentation is that 84% of global technology buyers agree that AI will change how their companies buy technology in the next 12 months. (Source: IDC B2B Technology Buyer Survey 2025, WW n=406)

The buying journey has already shifted and it looks like this: AI discovery → vendor website check → peer and colleague validation → channel partner consultation → shortlist. If your brand and solutions do not surface clearly when a CIO or CFO uses ChatGPT or Google Gemini to research a purchase decision, you are being eliminated before you even know you are in the running.

This makes Answer Engine Optimization (AEO) and Generative Engine Optimization (GEO) business-critical capabilities, not just marketing experiments. By 2027, 35% of Japanese organizations are projected to have unified, coordinated AI governance in place. The vendors who are visible, credible, and well-structured in AI-generated answers will have a compounding advantage. (Source: IDC FutureScape 2026 – AI-Fueled Business Strategies, Japan)

Agentic AI: The Next Competitive Frontier

The broader industry shift reinforces why speed matters. As highlighted in the presentation by IDC Japan’s Takuya Uemura, the AI supercycle is now entering its second investment wave — moving from infrastructure build-out to enterprise application and services adoption. Enterprise AI platform, app, and services spend globally is projected to grow from $400 billion in 2026 to $1 trillion by 2029. (Source: IDC Worldwide AI and Generative AI Spending Guide V1 2026)

Critically, 83% of buyers surveyed agreed that AI agents are lowering switching barriers between suppliers. The vendors who lock in outcome-based relationships now, before the market commoditizes, will be the ones defining the next competitive moat.

Three Moves That Matter Right Now

Sandra Ng’s closing framework was direct and actionable:

Do now: Replace generic benchmarks with a single Japan reference customer, one metric, one timeline. Map your go-to-market pitch to the full buying committee, not just the CIO. Fix your AI discoverability gap immediately.

Do this year: Make ROI transparency your competitive weapon. Build deployable agentic workflow stories. Invest in AI-optimized third-party content that earns citations in AI-generated answers.

Bet on this: Position into AI governance, trust, and compliance as a revenue line, or become the vendor that makes compliance an outcome rather than a conversation. Start measuring your share of answer by category, persona, and geography. Execute a layered discovery strategy: SEO + AEO + GEO. (Source: IDC C-suite Survey, September 2025, APJ, n= 300)

Japan’s AI Supercycle Will Not Wait

The window to establish leadership positioning in Japan’s AI market is open—but it will not stay open indefinitely. The enterprises that move from experimentation to outcome-based deployment, the vendors who close the four gaps, and the organizations that make themselves discoverable in AI-powered buying journeys are the ones who will define the next era of Japan’s technology market.

The conversation that started in Tokyo continues in Osaka.

Join us at IDC Directions Osaka on July 28, 2026 to go deeper on how Japan’s AI supercycle is reshaping your industry, what the data says about where the next wave of value will be created, and how your organization can get ahead of it. Register for IDC Directions Osaka today. Seats are limited.

Mike de la Cruz - Corporate Communications Director - IDC Asia/Pacific

Mike de la Cruz is Corporate Communications Director for Asia Pacific at IDC, bringing over 25 years of career experience in marketing and communications for the information technology industry. He shapes and amplifies IDC's research-driven narratives, positions executives and analysts as authoritative industry voices, builds relationships with top-tier technology and business media across the region, ensures consistent brand voice and positioning, develops content that drives audience engagement, and leverages social media and digital communications to extend IDC's reach.

国际数据公司(IDC)最新研究显示,AI超级周期带来的不仅是技术代际的更迭,更是一场深层次的组织范式变革。以智能体为核心驱动力的新型生产力形态——OPX(One-Person X),正在悄悄影响传统的企业组织逻辑,并对PC等终端硬件市场产生结构性的拉动效应。

AI超级周期下的组织重构

以智能体为代表的AI超级周期带来的不仅是技术迭代,更是一场生产力供给的结构性变革。当智能体从“辅助工具”进化为能够独立执行与产出的“生产力单元”,组织形态的底层逻辑必然随之重构。国际数据公司(IDC)最新研究将这一变化命名为OPX(One-Person X)——AI时代以单人为核心决策单元的新型组织范式,其核心共识可概括为:少数人负责决策与价值判断,AI智能体集群承担执行与产出

基于这一框架,OPX下辖两种落地形态:OPC(One-Person Company),即以个人为核心决策者、AI智能体为生产力引擎的企业级组织形态;OPE(One-Person Expert),面向个人专业服务领域。其中,OPC是AI重构商业组织最具颠覆性的形态。IDC对其的定义是:以个人为核心决策者,AI智能体/工具为核心生产力,至少可在单一任务环节独立完成业务,人工负责审核、判断与异常处理,最终实现传统需多人团队才能完成的商业闭环。从组织规模看,OPC可以是1个人,也可以是10人以内的轻量化团队

OPX的运作模式催生了一套独特的硬件需求:7×24小时不间断运行(无专职运维,中断即停业)、全生命周期高性价比(设备与算力成本均需可控)、端到端自动化闭环(任务接收、拆解、执行无需人工干预)、多设备无缝协同(跨终端业务连续性刚需)、无人值守下的远程管理与自进化,以及数据隐私与端侧主权保护。在这一需求图谱中,手机承担移动交互,云端提供弹性算力,而本地AI推理、多任务编排与端侧数据保护均依赖本地算力这使得PC设备,尤其是AI PC、AI主机设备,正成为OPX工作流的本地中枢。

OPX驱动的PC出货量近160万台

根据 IDC 最新预测,2026 年中国 PC 市场总出货量约为 4,038万台,同比微降 6.1%,整体大盘小幅收缩。但在结构层面,以 OPX(一人公司/超级个体)范式驱动的 PC 需求正在形成一个独立的增长极。IDC预测,2026年由OPX驱动的PC市场出货量近160万台。

IDC 认为,OPX 对 PC 市场的拉动效应正在经历从 “边缘增量”到”结构性力量” 的转变。尽管 160万的出货量仍属早期,但考虑到 OPX 的增速逻辑,即存量企业业务扩张、设备换机周期叠加、AI Native 工具链持续成熟。这一力量的上升值得重点关注。我们判断,OPX 有望在未来 3-5 年内成为 PC 市场增长叙事中不可绕过的结构性变量。

OPX对PC市场的影响集中体现在两个核心维度:采购逻辑的多元化和换机节奏的两极分化。首先,采购逻辑正在走向多元化。传统企业PC采购长期由人员规模驱动,OPX则打破了这种一一对应关系,设备需求开始更多地由任务负载而非人头数定义,一个创业者同时运行多台设备处理不同任务流的情况在特定场景下已成为现实。其次,OPX的换机节奏呈现两极分化态势。驱动快速换机一侧的力量来自OPX业务机会的不可预见性,当核心设备直接承载产出,AI技术的代际迭代持续抬高硬件准入门槛,不具备新能力的旧设备将在特定场景中丧失竞争力,都可能迫使OPX为抓住机会而在短期内高频地更新设备。另一方面,OPX极致追求效率与性价比敏感又构成了强大的延缓效应。

IDC建议

采购触达方面,OPX群体的决策路径迥异于传统企业客户——信息获取以与AI结合的评测内容、技术社区和同行口碑为核心,购买决策由个人独立完成,无需IT部门审批,购买渠道偏向个人消费路径。这意味着,针对OPX组织形态,厂商需要结合AI相关的内容做精准触达,可能包括评测合作、社区运营、KOL口碑建设,将成为撬动OPX客群的杠杆。

产品策略方面,OPX在消费级价格、企业级可靠性与AI原生能力之间提出了三元要求。面向OPX群体的产品设计,不宜简单地在现有企业或消费产品线上做加减法,而应从OPX的实际业务场景和工作流出发,重新定义产品配置与服务组合。

服务创新方面,OPX用户缺乏IT部门的缓冲,设备故障直接等同于业务中断。这种脆弱性定义了一个高付费意愿的服务市场,快速响应维修、备用机备援、数据恢复保障等“个人级紧急兜底”服务,有望成为PC后服务市场增长最快的板块之一。

AI超级周期带来的不仅是技术迭代,更是生产力供给的结构性变革。OPX之所以值得PC产业持续关注,根本在于生产力的最小单元正从“企业”转变为“个人+AI智能体”。产业的价值链条正从以硬件为中心的线性供给,转向围绕“个人生产力闭环”的软硬服一体化。

进一步交流

IDC中国终端系统研究团队将持续追踪OPX范式下的硬件需求演变与市场机会。如需获取更详尽的PC市场季度追踪数据、OPX用户画像深度分析,或定制化的市场策略咨询,欢迎随时与IDC中国团队取得联系。让我们共同探讨,如何在“一人即公司”的时代,重新定义计算终端的价值边界。如对其他研究内容感兴趣,也可与我们保持沟通。

引言:算力变局下的真实回响,拨开智算云迷雾

在日新月异的AI时代,中国智算云市场正经历着前所未有的狂飙突进。从大模型百模大战的喧嚣,到智能体应用的全面爆发,算力需求的重心正在发生深刻转移。然而,在千亿级市场规模的宏大叙事下,企业真实的使用现状却往往被层层迷雾所掩盖。当行业逐渐褪去‘唯算力规模论’的狂热,转向对效能与价值的理性审视时,真实的AI业务落地究竟面临着怎样的需求和挑战?

为了探寻这一命题,国际数据公司(IDC)与中国信通院分别从企业使用智算云服务与专属智算云建设两大视角,面向涵盖互联网、政府、金融、制造、通信、交通、能源等多个行业的250家企业用户,开展了一场深入调研,旨在还原当前中国企业智算云使用的十大真相。

核心数据总览:中国企业智算云整体产业格局

中国智算云市场正迎来从百亿级迈向千亿级的历史性跨越,整体格局呈现出规模扩张与结构优化并进的态势。

IDC数据显示:2025年中国智算云基础设施市场(AI IaaS)规模达到486.7亿元,同比迅猛增长128%,预计未来2年超千亿。互联网、大模型服务、汽车等行业是当前该市场的采购主力,部分制造、医疗、政务数智化项目从试点采购转向常态化租赁,成为拉动该市场的新增力量。

与此同时,企业的大模型推理服务需求呈指数级爆发,推动专属智算云建设市场进入高速增长通道。联合调研测算数据显示:截至2026年2月,中国企业自建智算云推理算力规模已超过350EFlops,能够支撑日均Token产能超70万亿,成为中国大模型推理服务的核心承载形态。

真相一:混合部署成产业标配,单一形态已不再是企业算力战略的最优解

在数据安全、成本效益、业务弹性的多重考量下,企业智算云部署已告别单一模式,“智算云服务+专属智算云建设”组合成为当前多数企业的默认形态和最优解。

IDC所调研的智算云服务用户中,85.8%的企业选择了 “外部AI算力服务+自建数据中心” 的混合架构,表明企业既希望解决突发峰值算力、短期项目研发、临时模型微调等弹性需求,又希望保持对核心数据、高频稳定业务的控制权。

中国信通院调研数据也显示:超过80%的省市级政务单位通过政务专有云平台部署大模型服务;头部银行机构中96%的算力规模采用自建模式,而中小金融机构则更多采用“专属云+公有云服务”的混合架构。此外,能源、制造等行业普遍遵循分级部署,按需采购原则。

真相二:通算与智算各司其职,共同构成企业算力供给底座

此前在智算中心大热的背景下,“通用算力中心是否会被取代”成为行业热议话题,本次调研数据给出了明确的答案。

从实际应用来看,通算凭借低成本、高兼容优势持续占据企业算力底盘,支撑 日常办公、ERP 、数据库、大数据离线统计等 IT 业务;智算作为增量支撑模型训练、多模态推理、Agent 智能应用等AI 业务,二者形成互补分工格局。

在问到企业当前智算和通算的使用比例时,最高比例(40.0%)的企业选择了“智算超过40%,通算少于60%”,但69.2% 的企业认为未来2年智算占比仍在50%以内。

2026年,随着AI迈入智能体时代,CPU迎来了新的发展机遇。过去大模型训练高度依赖GPU强大的并行计算能力,但随着智能体应用的规模化落地,智能体工作负载不仅需要GPU提供的大规模并行推理能力,也需要CPU所擅长的复杂任务编排、多步逻辑控制、系统级调度与状态管理能力。这一变化正推动 “CPU-GPU协同均衡”发展。

真相三:智算云规模增势迅猛,整体利用率均衡,但行业资源利用率分化明显

从供给侧看,智能算力利用率呈现节点与区域分化特征。枢纽节点的智能算力利用率高于非枢纽节点;东部地区整体利用率与西部地区相近,西部地区承接大量东部外溢的低时延、高计算密度、弱交互性的智算需求,显示出跨区域调度对提升整体算力效率的积极作用。从需求侧看,不同行业的智算资源利用率呈现分化格局。IDC调研数据显示:近半数企业的智算资源利用率维持在 51%-70% 的中等区间,且仍有 6.7% 的企业利用率仅为 31%-50%,处于较低水平。

头部云厂商、互联网企业、AI 自研大厂、新能源汽车等凭借多年技术积累,通过算力调度平台、精细化资源管控系统与全栈协同优化体系,实现了算力资源的高效复用,部分头部企业核心集群利用率突破80%。

制造、政务等行业的利用率低于平均水平,实现 70% 及以上高利用率的企业占比不足 35%。不少自建智算项目存在 “重硬轻软” ,仅完成了服务器、GPU 等硬件设备的上架部署,但缺乏配套的统一算力调度平台、资源池化管理系统与动态资源分配机制,造成了算力资源的浪费。

真相四:推理场景消耗AI算力过半,线下推理算力贡献1/3以上

百模大战时期,资本市场聚焦千亿参数大模型自研训练,训练算力需求一度占主导,但随着 Agent 应用规模化落地,轻量化小模型 + 端云协同推理进一步拉动推理算力需求,推理场景成为智算云市场的基本盘。

IDC调研数据显示,当前八成以上的企业用于训练工作负载的AI算力占比低于50%,如果按样本数加权平均计算来看,企业用于训练和推理工作负载的AI算力四六开。泛互(60%)、金融(46.7%)用于训练的AI算力占比高于平均值,而政府和制造行业2/3的AI算力都用于推理。

中国大模型 Token 调用量的爆发式增长,进一步推动了线下推理的崛起态势。根据发改委数据:截至2026年3月,中国智能算力规模超过1882EFLOPS。调研显示:推理算力占据六成,而其中企业自建智算云贡献的线下推理算力占比超过34%。政务、金融、能源等强监管行业的核心业务推理需求正不断通过物理隔离或逻辑强隔离实现线下部署,既能保障数据安全合规,又能实现毫秒级低时延响应,适配核心业务的严苛要求。

真相五:硬件与公有云IaaS仍是中国企业IT最大支出项

2025年企业总体IT支出中, IT软件与第三方服务花费占比不足三成,七成用于采购IT硬件和公有云IaaS。具体到AI相关支出,中国企业也高度集中在AI硬件与AI IaaS上,在 AI 应用和服务层的投入仍处于早期阶段。

IDC调研数据显示:当前智算云企业用户IT硬件支出中用于GPU服务器的占比为58.4%,公有云IaaS支出中用于AI IaaS的占比为54.3%。泛互行业作为AI投入力度最大的行业,AI支出占比超过上述平均水平。

真相六:100P算力是企业智算云部署模式偏好的关键分水岭

IDC调研数据显示:当前超过85%的企业使用了26-100 PFLOPS规模的AI算力。这一区间足以支撑中小参数大模型后训练、企业级场景推理及数据密集型AI应用,是兼顾成本与实用性的主流选择。对于该算力区间的企业而言,AI业务往往存在阶段性、波动性特征,全年算力使用起伏较大,自建投入的回报率偏低,因此更倾向于采购公有智算云服务。

中国信通院调研显示,100P FP16等效算力是企业选择规模化自建和上云的核心分界指标。当算力需求超过100P时,企业倾向于自建以掌握核心资产并实现长期成本摊薄。这类主体主要为大型央企、头部银行机构、头部互联网企业。调研发现,当前国有六大行自建智算云规模均超过100P,实现了完全物理隔离的私有云环境。而当需求在几十P级别时,企业更倾向使用服务商提供的共享或独享的云服务,这一分化布局现象在能源行业尤为突出。

真相七:制约智算云产业规模化落地的并非算力短缺,而是全链路配套能力不足

AI 基础设施落地是系统性工程,单一环节短板会形成 “木桶效应” ,拉低整体算力效率。IDC调研数据显示:企业 AI 基础设施落地的瓶颈已从 “算力短缺” 转向 “全链路配套能力不足” ,机房环境、网络互联、软件适配、存储吞吐等共同构成了制约 AI 业务规模化落地的系统性障碍。

真相八:使用 MaaS 和 Agent 的挑战是“将 AI 真正用进业务里”

模型即服务(MaaS)与Agent的持续落地,让AI的获取变得像水电一样简单,但“用好”依然是巨大挑战。调研显示:企业核心瓶颈并非算力供给不足,真正的挑战在于:如何与现有业务集成打通业务孤岛?如何让Agent理解复杂的业务逻辑?如何管理数量庞大功能各异的智能体?只有跨越了从 “技术工具” 到“业务引擎” 的鸿沟,AI才能真正创造商业价值。

真相九:全栈智算云在线下推理场景凸显效能优势

对于采用智算云服务的企业而言,推理效能取决于云服务商;而对于自建智算云的企业而言,AI落地涉及从底层芯片、算力调度、云平台,到大模型部署、场景微调、应用开发的每个环节,任一环节的短板都会拉低整体效能。

本次联合调研及实测数据显示,相较于零散拼凑多厂商软硬件的项目,采用全栈AI云服务厂商产品组合的项目词元产出效能提升超过20%(词元产出效能:相同场景下单位算力日均Token调用量),全栈厂商在芯片、云平台与模型层面可实现端到端协同优化,降低跨产品兼容损耗。

真相十:国芯国模适配逐步完善,全栈国产化进入规模化部署窗口

IDC调研数据显示:总体而言,当前国产AI芯片的使用率已高于国外品牌,但不同行业差异较大,互联网、模型和IT服务行业国外品牌的使用率明显高于国产品牌;汽车和金融服务行业两者的使用率相当;政府、能源和传统制造行业中国产品牌的使用比例更高,政务行业新建智算项目已100%采用全栈国产化方案,大型银行机构的核心业务系统逐步完成了国产化适配,能源、制造等行业的国产化替代进程也在不断加速。

随着国产芯片市场占有率逐步提升,国产AI芯片与国产大模型的协同适配正从“单点突破”走向“联合体推进”,基于“国芯+国模”的全栈国产化解决方案已迈入实质性规模化部署周期,这点在政府、银行、能源等关键行业表现得尤为明显。

结语:中国智算云告别粗放,迈入精细化运营时代

2026年的企业智算云市场,正在经历一场深刻的 “成人礼” 。随着市场认知回归理性,粗放式堆算力的发展逻辑彻底落幕,逐渐进入精细化运营时代。在这个时代,赢家不再属于单纯囤积算力的人,而是属于那些能够精准调度异构资源、深刻理解行业场景、并将AI无缝融入业务血脉的长期主义者。

IDC相关研究

围绕智算云相关研究内容、技术能力与市场格局,IDC 将持续开展系统性研究,包括但不限于:

  • 《中国智算云基础设施服务(AI IaaS)市场跟踪,2025下半年》(2026年4月发布)
  • 《IDC Market Forecast:中国整体云计算市场预测,2025-2029》(2026年5月发布)  
  • 《IDC Survey:中国智算云服务市场企业用户调研,2026》(即将发布)
  • 《IDC Tech Assessment:面向智能体的混合云智算基础设施技术能力评估,2026》(即将发布)
  • 《IDC Market Share:中国AI云存储市场份额,2025》(即将发布)
  • 《IDC PeerScape:中国市场异构算力调度和管理实践洞察,2026》

如需进一步了解智算云相关研究内容,或咨询 IDC 在云计算、AI 基础设施及数字化转型领域的其他研究成果,欢迎与我们保持联系

Rachel Liu

Rachel Liu - Research Director

Rachel Liu is a research director for China’s Cloud and Services group. Her research covers public cloud, private cloud, edge cloud, industrial cloud, intelligent computing, and IT services. She is responsible for research plans, research execution and management, data tracking…

2025年,中国金融行业站上了从”数字化”迈向”智能化”的关键转折点。银行、保险、证券这类一向以稳健著称的机构,在合规与风控的硬约束之下,对新技术的引入历来审慎;但生成式AI带来的效率跃升与体验重塑,又是任何一家金融机构都不愿错过的红利。行业演进的核心驱动力,已经从最初”要不要用大模型”的试探和观望,转向场景的快速扩张、应用的规模化落地,以及对底层算力、解决方案与合规能力的系统性需求。金融AI云的竞争逻辑由此发生了一次根本性跃迁——从比拼算力”资源供给”维度,转向以”算力底座、平台调度、行业模型与合规可控”为核心的全栈能力竞争。

国际数据公司(IDC)最新发布的《中国金融云市场跟踪研究Add-on_AI全栈云》报告首次对中国金融云市场中公有云AI算力服务、私有云软硬件AI基础设施和GenAI解决方案子市场做出全栈式市场营收评估,中国金融AI全栈云市场2025全年市场规模207.6亿人民币,较2024年同比增长50.0%,远超金融云总盘23.9%的同比增速。这个数据对比显示出金融机构对AI相关领域的重视程度不断升级,预算投入持续加码,市场潜力快速兑现。

金融全栈布局下,头部云厂商的五种路径

经过几轮市场调整,金融AI全栈云的第一梯队逐渐清晰。五家代表性厂商在算力、平台、模型与应用上的侧重各有不同,拼出一张完整的能力地图。

公有云算力+私有化交付双轨并进:代表厂商 阿里云

阿里金融云,在公有云一侧依托灵骏智能计算集群,在大规模算力调度与推理优化上持续投入;金融自主私有云一侧,平头哥自研的真武AI芯片已进入规模化部署阶段,并从2025年下半年开始逐步起量。叠加面向金融场景打磨的通义点金行业大模型,以及可落入客户机房的一体机交付形态,阿里云事实上把”芯片—模型—应用”的链路在金融场景里走通了一遍,并将在2026年持续深耕,加速复制。

软硬协同稳扎稳打:代表厂商 华为云

华为昇腾系列智算服务器,多年来在国产算力替代的进程中扮演压舱石角色。对金融机构而言,华为的吸引力恰恰在于”软硬一体、自主可控”带来的确定性——在数据不出域、供应链安全成为硬约束的当下,一套经过大规模验证的昇腾底座,往往比单纯的性能参数更有说服力。

AI应用和智能体先行:代表厂商 火山引擎

背靠字节跳动内部高并发推理场景的长期锤炼,火山引擎在推理成本控制与资源利用率方面建立了差异化优势,豆包C端的成功也给火山解决方案带来足够的曝光度和品牌效应,豆包大模型与火山方舟平台构成其对外的主要解决方案抓手。在智能营销、智能客服等高频交互场景中,这种”应用先行、智能体多点开花”的策略,更容易让金融客户在数据非敏感领域优先尝到AI红利甜头。

让金融AI“融汇贯通”起来:代表厂商 腾讯云

依托在平台层数据库和大数据产品的行业优势,腾讯云从AI通用大模型到上层AI应用,从平台产品到底层异构算力管理,试图把这条链路做成一个连续整体,而非彼此割裂的模块。对于拥有庞大存量系统、又要稳步引入AI能力的金融客户来说,这种平滑过渡、整体交付的能力,本身就是一种稀缺竞争力。

“芯片+平台”组合拳:代表厂商 百度智能云

昆仑芯P800已完成规模化验证,2025年以来已交付多个万卡集群,并支撑了文心大模型新版本的训练。配合百舸AI计算平台在异构调度上的能力和千帆平台对金融应用场景的支撑,百度为金融客户从模型基础训练到推理开发上线,铺设了一条相对完整的国产化通道。

五家厂商路径不同,指向的判断却高度一致:在金融这样对”稳健”近乎苛求的行业里,单点能力很难构成壁垒,唯有把算力、平台、模型、应用与合规能力打通,才能真正站稳脚跟。

金融ISV服务商:增量市场背后的挑战

在AI云厂商与金融机构之间,往往活跃着各个子赛道金融ISV服务商的身影。他们既是AI方案的集成者,也是AI应用落地的”最后一公里”。2025年,金融ISV服务商感受到了明显的市场变化,AI市场带来新增营收机会的同时,也带来了全新的机遇和挑战。

一方面,金融AI场景在2025年正式跨过了项目落地的门槛,并呈现出明显的量价齐升势头。在传统金融预算大环境整体承压的当下,这样一块实打实的市场增量殊为不易,它意味着新的项目机会、新的合作入口,以及与客户重建深度连接的契机。不过,这份增量需要冷静看待,AI相关收入对整体金融云解决方案营收的直接贡献占比仍然偏小,更多扮演的是”敲门砖”与”引流器”的角色——通过AI项目切入客户、增强黏性,再向定制化开发、平台软件乃至底层基础设施资源的销售导流,而这些方向的受益者往往是云厂商占优。真正的商业价值,可能兑现在AI之外。

另一方面,金融机构对AI方案的选型普遍呈现迷茫状态。用哪家大模型,走开源还是闭源,部署在公有云还是落到本地机房,是否涉及业务系统解耦和微服务改造,数据安全合规如何满足监管要求,未来方案兼容性与可持续性又如何保证——这些问题几乎没有现成经验可以照搬。金融ISV服务商需要陪伴客户共同摸索试错,在反复的POC与调优中消耗大量人力成本。这种”陪跑”固然能加深信任,却也推高了交付成本,挤压了本就不宽裕的利润空间。

IDC市场调研发现,金融ISV AI服务商正从”项目制交付”向”持续运营”转型,从单纯的技术提供者,转变为深度参与客户流程重构的”业务转型伙伴”。金融AI可能对未来ISV服务商竞争格局带来颠覆性变化,金融服务商不仅要懂业务,还要懂AI、养AI、用AI,扛的住长期AI运营的玩家才能在未来竞争中立于不败之地。

IDC洞察:金融AI要以全栈能力和生态协同赢得长期战斗的胜利

对于金融云服务商和云厂商而言,是否全栈式布局已不再是选答题,而是必答题。金融客户要的不是某一颗更快的芯片或某一个更准的模型,而是一套数据模型平台应用服务端到端打通、并能自我迭代和强化的能力体系。在这一点上,自主算力与行业大模型的协同尤为关键,它既回应了自主可控的政策诉求,也构筑起别人难以复制的护城河。与此同时,”合规”应当被前置为产品能力而非事后补丁——数据不出域、决策可追溯、幻觉可约束,在通用市场或许是加分项,在金融市场却是入场券。

随着智能体应用从单点走向全流程,推理调用量将呈指数级放大,按需租赁、弹性扩容与混合部署在未来三到五年内,有望逐步替代传统的重资产采购模式,成为金融AI基础设施建设的新常态。能在保障性能的同时把综合成本压下来的厂商,将在长期竞争中占得先机。

此外,建议金融云厂商加大对金融ISV服务商的支持力度。ISV是云厂商触达金融客户的重要渠道,也是行业Know-How的历史沉淀者,并在AI拓展方面成本承压。云厂商在算力补贴、平台开放、收益合作与人才培养上给予生态伙伴更深层的支持,可加快金融AI整体升级节奏。

金融行业要的,从来不是非此即彼的”创新”或”稳健”,而是两者之间的平衡。能否帮客户既迈得开步子、又站得稳脚跟,正是这场金融AI全栈云竞速中,最终拉开差距的地方。

如需进一步了解IDC相关研究,或就中国金融云AI市场发展趋势进行深入交流,欢迎与IDC联系,获取更多洞察与数据支持。

北京,202665——国际数据公司(IDC)最新发布的《2026年第一季度全球手持智能相机市场跟踪报告》显示,2026Q1全球手持智能相机市场出货量达到414万台,同比增长33%。销售额超过105亿元人民币,同比增长20%。IDC预计,至2030年,全球手持智能相机市场规模将会超过4000万台,五年复合增长率接近18%。

2026Q1全球手持智能相机发展状况

  • 运动相机市场:26Q1全球运动相机出货量接近201万台,同比增长39%。平均单价2149人民币,同比下降11%。推动市场高速增长的主升浪来自于大疆和影石爆款明星产品热销,同时上一代产品的强势调价促销也为厂家贡献了良好业绩。大疆和影石出货量市场份额同比攀升13个百分点,老牌厂商GoPro生存空间继续被挤压。从运动相机市场细分品类来看,可拆卸运动相机即“拇指相机”出货量同比增长惊人超过350%,不可拆卸运动相机同比增长16%。
  • 云台相机市场:26Q1全球云台相机出货量同比增长超过18%,平均单价2840人民币,同比下调12.5%。此市场大疆继续保持独占领先优势。手持云台相机市场在产业上下游供给侧和最终用户需求端都保持了高热度,26年将迎来产品突破升级和新竞争厂商出现。
  • 全景相机市场:26Q1全球出货量超过50万台,同比增长高于50%。影石在全景相机领域排名第一且大幅领先,出货量份额接近七成,大疆Osmo360火爆单品保持热度拿下超过两成市场。

2026Q1全球手持智能相机市场主要厂商概况:

26Q1全球手持智能相机市场,大疆以65%出货量市场份额保持第一,同比增长38%。影石出货量市场份额达到22%,同比增长66%排名第二。GoPro受到大疆和影石的全球激烈竞争出货量同比持续萎缩。

大疆

大疆凭借无人机领域全产业链成功多年积累的硬件核心技术(多轴云台稳定系统,影像及多传感器融合技术,飞行及运动控制系统)向手持智能影像设备迁移,打造高可靠性和易用性的影像产品全系列矩阵。品牌拥有完善供应链体系,成本抗压弹性,议价能力,上游核心部件定制化能力均较强。26Q1云台相机自身出货量占比超过一半,Pocket4新品预售火爆断货,3代产品调价后继续受到市场追捧。不可拆卸运动相机出货量市场份额同比上升3和百分点增长至54%,可拆卸运动相机即“拇指相机”单品Osmo Nano在25年发售后保持热销,26Q1出货量拿下过半市场份额。在全景相机领域,26Q1大疆凭借Osmo360单品赢得22%出货量市场份额占稳这一市场。

影石

26Q1全球出货量接近90万台(结果基于IDC研究方法论统计得出,涉及上市公司数据最终请以厂商财报披露为准)。影石持续加大投入研发资源,凭借在全景相机领域作为全球领军厂商的多年技术积累,已经在差异化应用场景和应用软件方面形成独特优势和壁垒。其系统应用软件在拼接算法,多场景防抖,深度主体追踪以及在AI助手一键剪辑方面获得庞大用户群的广泛认可。影石在全景相机领域虽然遭遇挑战,但26Q1仍旧保持了接近七成出货量市场份额。在不可拆卸运动相机领域,AcePro2街拍套装促销取得成功,出货量市场份额创新高超过17%,同比吃掉7个百分点。在可拆卸运动相机市场,尽管面临强势竞争,产品出货量同比增长翻倍且Go Ultra表现亮眼。此外云台相机高热度新品将在第二季度全球出货也为厂商锚定新的增长点。

GoPro

遭遇中国厂商全球激烈竞争,面临供应链成本上涨,削减库存等多方面压力同比大幅收缩。26Q1全景相机出货量同比跌幅大于运动相机,运动相机内部Lit Hero以及Hero低价老品自身比重增加,整体均价同比下调17%。虽然发布M1系列新一代运动相机以及继续推广全景Max2新品但市场表现有待观察。

IDC分析师洞察与市场未来发展预测:

  1. 手持智能相机从硬件堆叠到全产品矩阵,进一步深度挖掘场景覆盖。回顾手持智能相机发展历程,关键硬件研发诸如大底传感器,可变光圈系统,8K超高分辨率等技术仍旧是推动市场发展的核心因素。从产品端侧看已经由全景相机,不可拆卸运动相机,手持云台相机到可拆卸运动相机形成手持智能影像全产品矩阵。与此同时覆盖用户人群由小众极限运动极大拓展到广大运动爱好者到Vlog创作者和大众记录日常。厂商仍旧致力于深挖用户场景需求乐此不疲。
  2. AI融入手持智能影像终端最终能为用户带来什么价值。目前AI功能主要集中于AI场景深度感知,AI智能降噪,多帧合成算法,暗光画质增强,AI后期智能剪辑成片等功能型应用。AI融入手持智能影像终端的未来愿景是成为智能影像机器人终端的底座和大脑,由被动的影像记录终端成长为基于海量数据采集可以按照用户各种需求主动生成创作内容的超级智能体。
  3. 中国是手持智能相机的第一大市场。从2025年度出货量来看,中国贡献了超过一半的市场份额,26Q1延续了这种趋势且贡献进一步有所加大。除中国外,美国,西欧,亚太区和日本是全球前四大市场。海外市场产品渠道营销是国内厂商投入资源的核心议题之一。
  4. 手持智能相机市场的渗透率仍旧较低,未来增长空间可期。手持智能相机市场无论从产业资本和上下游供给侧还是最终用户消费者层面仍旧保持高热度。无论哪一条细分产品线市场的未来五年复合增长率均高于15%。IDC预计,至2030年,全球手持智能相机市场规模将会超过4000万台,总体复合增长率接近18%。

IDC手持智能相机定义

IDC对手持智能相机的定义是指具备计算处理能力,搭载电子或光学等防抖能力,分辨率2K及以上且可手持使用的消费级便携影像设备,包括运动相机(包含可拆卸和不可拆卸运动相机两种形态),全景相机和云台相机。

IDC更多相关详细内容,敬请关注:IDC WW Quarterly Handheld Smart Camera Tracker

请点击此处与我们联系

Jacky Xue

Jacky Xue - Research Manager

Jacky Xue is a research manager with the Client Systems Research team at IDC China, responsible for projector, IWB and enterprise client devices research and analysis. His responsibilities include tracking the industry data, monitoring the market development, analyzing the future…