2025年中国企业级MaaS市场经历了从试点到规模化应用的关键转折。无论是Token调用量还是实际营收,各个统计口径都呈现出倍数级的增长,同时Token消耗的快速攀升正在重新定义企业使用人工智能的方式。然而,在这一轮高速增长的背后,市场仍然面临性能、安全合规、回答质量等多重实际约束。IDC认为,MaaS厂商的竞争焦点正在从过去单纯的价格比拼,转向“价格、性能与工具链支持”的综合能力竞争。

市场总体规模:调用量增长16倍,但需理性看待基数效应

2024年,中国企业级MaaS市场按调用量统计的规模仅为114万亿Tokens,而到2025年,这一数字跃升至1944万亿Tokens,同比增长约16倍。在营收口径下,2025年中国公有云MaaS市场的规模达到30.7亿元人民币。与此同时,面向传统政企客户的大模型私有化部署市场也保持活跃,尤其在政务、金融、能源等领域形成了独立的采购与交付体系。

IDC预计2026年全年Token消耗量约为40,000万亿次,较2025年进一步增长约20倍。这一加速增长的主要驱动因素,是多模态大模型的逐步成熟以及Agent类应用的规模化落地。换句话说,市场正在从“文本生成”向“多模态理解与自动执行”扩展,每一次交互所消耗的Token量级也相应大幅提升。

Token消耗快速增长,但不同场景之间差异巨大

Token消耗的整体走势非常明确:2025年1月全市场日均消耗约为1.6万亿Tokens,到2025年12月底,这一数字已经攀升至9.6万亿Tokens。随着多模态大模型的进一步成熟,日均Token消耗的增长曲线将会变得更为陡峭。

不过,不同应用场景之间的Token消耗量级存在巨大差异,这一点往往被笼统的市场总量数字所掩盖。例如,一家投资机构在合同摘要场景中,一次处理几十个合同文档,单次消耗的Tokens可达200万。一家教育机构在其在线培训场景中,高峰使用期日均消耗高达2000亿Tokens。而使用多模态产品生成一个视频,也会消耗上千万Tokens。这些差异意味着,企业在评估MaaS服务时,不能只看单位Token的报价,还必须结合自身场景的实际消耗特征来判断总成本。

从应用场景的分布来看,当前公有云上MaaS的主要应用群体集中在泛互联网行业(游戏、娱乐、教育)、智能办公赛道、智能硬件赛道(智能汽车、手机、智能眼镜)以及大消费赛道。主要的应用场景包括角色扮演、短剧生成、市场营销、搜索、数据处理、数据分析和文档处理。而在传统政企的私有化部署项目中,应用场景相对收敛,主要集中在智能办公、数据处理与分析、市场营销等几个方向。

竞争格局:公有云头部集中,私有化部署市场更为分散

在公有云MaaS市场,按调用量计算,2025年火山引擎占据了接近一半的份额,其次是阿里云、百度智能云、硅基流动以及移动云。其他值得关注的厂商还包括腾讯云、商汤科技、华为云、天翼云等。

如果按营收口径来看,火山引擎仍然占据40%以上的市场份额,阿里云、百度智能云、智谱以及移动云位列前五。不同厂商在单位Token定价、折扣策略以及高价值场景的占比上存在显著差异。例如,某些厂商可能在低单价、高调用量的场景中占优,而另一些厂商则在高单价、专业场景中获得了更高的营收贡献。

在私有化部署市场上,格局则完全不同。传统政企客户出于数据安全、合规可控等考虑,仍然将私有化部署作为第一选择,也因此培育了众多的大模型平台私有化厂商。2025年IDC追踪到的头部厂商包括百度智能云、商汤科技、电信AI、中关村科金、创新奇智、星环科技以及中国电子云。相比公有云市场,私有化部署市场的集中度更低,这与政企采购的区域性、行业性特征有直接关系。

企业落地大模型时,性能、合规与质量优先于成本

高速增长的数字背后,企业在大模型落地过程中仍然面临一系列现实挑战。根据IDC的调研数据,影响大模型落地的Top5因素依次是:模型性能、安全合规要求、回答质量、在AI平台可用性以及成本效益。其中,性能指的是企业部署上线后的稳定性、并发数和可靠性;安全合规既要求模型在中国市场可以合规使用,也要求模型生成的内容本身合规;回答质量则直接决定了模型能否被大规模上线应用。

值得注意的是,成本效益在当前阶段排在第五位,并非企业最优先考虑的障碍。这一结果在一定程度上说明,企业当前更关注“模型是否能用、是否安全、是否稳定”,在这些前提条件满足之后,才会进入精细化的成本比较。然而,原文也明确提示了一种中长期风险:随着大模型应用场景的持续渗透,以及算力紧缺状况的延续,大模型的投资预算终将成为关键挑战。换句话说,成本问题目前还没有爆发,但它迟早会来。

MaaS市场的竞争要素正在发生结构性的变化

随着Token经济的兴起和市场参与者的迅速增加,MaaS市场的竞争规则也在发生变化。IDC认为,未来制胜的核心要素将集中在三个方面:价格与成本优化、性能、以及工具链与应用开发的支持。

在价格与成本方面,企业Token的日均消耗正在快速增长,尤其是各类Agent类产品的出现,进一步放大了Token的使用规模。随着规模化智能的到来,厂商需要关注的不是单纯降低单位Token的价格,而是帮助用户降低整体成本,同时确保输出的是高质量、有效用的Tokens。

在性能方面,大量行业场景对并发数和吞吐量有很高要求,与此同时算力紧缺的现状短期内难以根本缓解。能够通过底层算力优化来提升模型训练和推理效率的厂商,将在用户选型中获得明显的优势。

在工具链与应用开发支持方面,过去两年的MaaS市场中,买方选型的首要因素几乎都是价格,而且由于大模型迭代速度太快,买方通常选择直接调用大模型API,而不进行二次调优。但未来,企业将有更多垂直场景落地,这些场景离不开智能体的开发。因此,MaaS平台在工具链(如调试、评测、智能体编排)上的完备性,在用户选型中的重要性将不断提高。

展望:2026年预计20倍Token增长,但需注意情景依赖

基于现有数据和趋势,IDC对MaaS市场给出了明确的预测:2026年中国MaaS市场的Token消耗量将达到约40,000万亿次,按营收规模计算将达到约186亿元人民币。2024年至2030年的年复合增长率约为1154.9%,但这一数字是基于高增长情景得出的。

需要特别提示的是,如此高的复合增长率高度依赖于一系列前提条件:多模态模型的持续成熟、Agent类应用的大规模落地、算力供给不出现严重瓶颈,以及合规政策保持稳定。上述任何一个条件发生变化,实际增速都可能显著低于高增长情景。对于行业参与者和企业买方而言,在看到巨大市场机遇的同时,也有必要对风险保持清醒的认知。

对厂商与买方的建议

基于上述分析和IDC的调研数据,我们可以对MaaS厂商和企业买方分别提出几点建议。

对于MaaS厂商而言,在价格竞争的同时,应当优先解决高并发场景下的性能稳定性和合规推理能力。工具链能力(调试、评测、智能体编排)正在成为差异化竞争的关键,不可忽视。此外,可以考虑在垂直场景(如合同处理、视频生成、教育陪练)中建立深度优化能力,而不是在所有场景中采取同质化的竞争策略。

对于企业买方而言,在当前阶段应当优先验证模型的回答质量、安全合规性和并发稳定性,成本可以放在次优级的位置上。在选择MaaS厂商时,建议关注其工具链的成熟度,而不仅仅是API的单次调用价格。从长远来看,建议尽早建立Token成本的评估机制,避免业务规模扩张后出现成本失控的局面。

关于数据:本文所有数据和图表均来自IDC《中国AI软件市场半年度追踪,2025H2》。调用量口径统计企业通过公有云MaaS平台调用大模型API产生的Token数量;营收口径统计企业实际支付的公有云MaaS服务费用(不含私有化部署);私有化部署市场统计面向政企客户的大模型平台软件及相关授权收入。

进一步交流

MaaS市场的爆发才刚刚开始,IDC将持续追踪这一市场的格局变化与技术演进。如需获取完整报告、订阅后续研究,或针对特定厂商与行业进行定制化分析,欢迎与IDC中国人工智能研究团队联系。

请点击此处与我们联系。

Yanxia Lu

Yanxia Lu - Research Director

  Yanxia Lu is a research director, focusing on big data and artificial intelligence (AI). Her responsibilities include big data information management platform, and big data analytics and applications. She is also involved in research on AI technology and enterprise…

The foundational question for any vendor or service provider operating in Japan’s technology market today is no longer whether AI infrastructure will grow. That is settled. The question is how fast, in what form, and most importantly who will capture the value as Japan’s AI infrastructure market, consisting of servers and storage for AI, approaches and surpasses the ¥1 trillion threshold.

IDC’s latest data and forecasts provide a clear-eyed answer. What follows is the strategic view every vendor in this ecosystem needs to understand.

The road to ¥1 trillion: Three milestones that define the opportunity

IDC projects Japan’s AI infrastructure market will follow a clear and compelling trajectory over the next five years. Three milestones stand out:

These are not optimistic projections. They are grounded in structural forces – government policy, enterprise digitalization pressure, hyperscaler commitments, and the irreversible integration of AI into Japan’s economic fabric – that show no signs of reversing.

Where we are today: The foundation that makes this possible

To appreciate the scale of what lies ahead, it helps to understand how quickly Japan’s AI infrastructure market has already moved. In 2025, the domestic market reached ¥670 billion. For accelerator-equipped servers alone, the three-year CAGR from 2023 to 2025 approached 200%, significantly outpacing the global average.

This growth was not accidental. It was driven by a deliberate convergence of government economic security policy and domestic capital mobilization. Under cloud-related policy frameworks designed to advance Japan’s technological sovereignty, domestically capitalized service providers and telecommunications carriers launched large-scale AI infrastructure buildouts at a pace rarely seen in the Japanese market.

Physically, the transformation is equally striking. Rack-scale systems and multi-rack liquid-cooled configurations are becoming the new standard. What was once a 2U GPU server is evolving into an infrastructure architecture where entire data centers are purpose-designed around AI workloads. The ¥670 billion market of 2025 is built on this foundation, and the ¥1 trillion market of 2030 will be built on what comes next.

The structural dynamics vendors must understand

Japan’s AI infrastructure market in 2025 is heavily concentrated in service providers, who accounted for 90.6% of total market spend. With that figure, three dynamics will shape the competitive landscape through 2030:

  • Hyperscalers will continue to expand. Their investment share surged from 39.8% in 2022 to 58.9% in 2025, a 19-point gain in three years. Alignment with hyperscaler platforms and roadmaps will remain a prerequisite for relevance in this market.
  • Domestic service providers will remain strategically critical. Other service providers including policy-backed Japanese service providers and telcos held 31.6% share in 2025, essentially unchanged from 31.8% in 2022 despite the hyperscaler surge. This is a resilient customer segment that values local trust, regulatory compliance, and sustained partnership over pure price competition.
  • Enterprise direct investment is the next growth frontier. At 9.4% of the market today, enterprise direct AI infrastructure investment has more than doubled in absolute terms since 2022, yet it remains nascent. Most large enterprises currently consume AI through generative AI services and SaaS platforms. As enterprises deepen their AI ambitions and move from consumption to ownership, direct infrastructure investment will accelerate. Vendors who build enterprise relationships now will be best positioned to capture this wave.

The semi-sovereign AI model: Japan’s defining strategic architecture

In April 2026, Microsoft announced plans to invest approximately ¥1.6 trillion in Japan between 2026 and 2029. Alongside this commitment came a plan to make AI infrastructure operated domestically by two partner companies available through Microsoft Azure, domestically built and owned infrastructure, connected to a global hyperscaler’s service layer.

This is what IDC describes as “Semi-Sovereign AI” and it is rapidly emerging as the defining model for AI infrastructure strategy in Japan. It represents a pragmatic and politically viable middle path: neither complete dependency on foreign hyperscalers, nor the prohibitive cost of fully independent domestic AI capability.

For vendors, this model is not a constraint. It is a structural opportunity. Semi-Sovereign AI creates a rich and expanding set of market roles in infrastructure design, systems integration, managed services, compliance, and the development of Japan-specific AI platforms. The vendors and integrators who understand this model deeply, and position themselves within it deliberately, will define the competitive landscape of Japan’s AI infrastructure market through 2030 and beyond.

What this means for vendors: Act now, not later

The policy-driven supply wave that built Japan’s AI infrastructure base is transitioning into a demand-driven growth phase. The central question is no longer whether the infrastructure exists, it does. The question is who helps enterprises put it to work, at scale, in ways that generate measurable business value. Three imperatives stand out for vendors competing in this market:

  • Accelerate enterprise engagement. The 9.4% enterprise share of today is the growth story of tomorrow. Vendors who invest in enterprise relationships, Japan-specific use cases, and ROI demonstration frameworks now will be positioned to ride the most significant demand wave of the decade.
  • Align with the Semi-Sovereign AI model. Understanding the interplay between domestic infrastructure owners, hyperscaler service layers, and government policy frameworks is not optional background knowledge. It is the strategic map for winning in this market.
  • Build for depth, not just scale. Japan’s market rewards vendors who demonstrate sustained local commitment, deep technical expertise, and an understanding of Japanese enterprise culture. The ¥1 trillion opportunity will not be captured by volume alone.

Ongoing vendor share and demand analysis is available via the IDC Worldwide Quarterly AI Infrastructure Tracker. For more detailed insights and market trends, please contact our analysts by completing this form IDC | Identifying Market Opportunities – Contact Us.

Shinya Kato - Senior Research Manager, AI and Automation - IDC Japan

Shinya Kato is a Senior Research Manager at IDC Japan and is responsible for the data analysis and forecasting team of Japan enterprise infrastructure market. He analyzes the impact of product technology, service offerings, and marketing strategies on enterprise infrastructure market and provides market forecasts, focusing on the domestic enterprise storage systems market. Through understanding technology adoption trends, he also provides insight into emerging devices such as flash, accelerators, and quantum computing. In addition to researching the HPC and AI infrastructure markets, he is also investigating new consumption models such as Hardware-as-a-Service, to help stimulate the market. Prior to joining IDC, he spent more than 10 years at Silicon Graphics, which was later acquired by HPE, where he held various domestic positions in sales, marketing, and business development. He has covered a wide range of businesses, from infrastructure hardware and container-based data center facilities to digital asset management, industrial virtual reality, and software for media & entertainment. He also served as a product manager for enterprise internet security software and appliances at the emerging vendor. He holds a Bachelor of Economics degree from Rikkyo University.

The Middle East War is no longer just a regional risk for technology leaders. It is becoming a global economic stress test — affecting energy prices, supply chains, inflation, business confidence, and IT budget decisions across Asia Pacific.

For CIOs, CFOs, and technology suppliers, the question is not simply whether IT spending will continue to grow in 2026. The sharper question is: which investments will still earn approval when volatility hits the budget?

That is the central signal from IDC’s latest webinar, Asia Pacific IT Spending Outlook 2026: Where to Win Amid Market Volatility. The market is not freezing. It is filtering.

This matters because the disruption is emerging against a market with strong underlying momentum. The opportunity is still there. The rules for winning it are changing.

The Middle East War is the shock

IDC’s earlier point of view on the Middle East War’s impact on IT spending outlined the broader global pressure points, from energy price volatility to supply chain disruption and cloud resiliency.

In Asia Pacific, those pressures are showing up in practical ways. Higher oil and electricity costs can raise operating expenses. Supply chain disruption can increase hardware costs and delay availability. Currency and pricing pressure can make approvals harder. As uncertainty rises, organizations move faster into contingency planning.

The first effect is not always a budget cut. More often, it is hesitation. Organizations start asking which projects are necessary, which can be phased, and which need to prove value faster.

As I noted during the webinar: “IT spending is likely to remain more resilient than in previous downturns because AI investment continues to support both the IT industry and broader economic growth.”

Resilience matters. It keeps the outlook from becoming a slowdown story. But resilience does not remove risk. The longer energy prices and supply chain disruption remain elevated, the more likely selective delays become targeted cuts.

Budget rotation is the signal

The most important shift is not how much organizations spend, but what they are willing to defend.

In a more volatile environment, buyers are becoming more risk-averse. ROI thresholds are rising. Procurement cycles involve more stakeholders. Vendors are seeing longer deal cycles, greater pricing pressure, and less predictable pipelines.

The visual above captures the core market signal: APAC IT spend is rotating, not disappearing.

Large transformations, experimental initiatives, and hardware-heavy refreshes are more likely to be deferred or reduced unless they are directly tied to productivity gains. Enterprise platform upgrades, GenAI pilots, and data and analytics programs are still in play, but they face tighter scrutiny and phased rollouts.

What remains protected? Investments tied to measurable outcomes. AI with clear ROI. Cloud optimization that supports flexibility and cost control. Cybersecurity that reduces risk and supports regulatory compliance. Infrastructure resiliency that helps organizations absorb disruption.

My colleague Vinayaka Venkatesh, summarized the shift directly: “The deals are not disappearing, but they are taking longer and require more effort to close.”

For technology suppliers, that line is the market cue. In 2026, vendors will not win by selling transformation in broad terms. They will win by helping buyers make the case for action now.

AI remains resilient, but proof matters more

AI remains central to the Asia Pacific technology agenda. IDC’s Asia/Pacific AI and GenAI spending outlook shows continued growth momentum as organizations invest in automation, productivity, and new business models.

But resilience does not mean every AI project gets a free pass.

One of the more important nuances from the webinar is the gap between AI infrastructure investment and enterprise AI execution. Hyperscalers and service providers continue to invest aggressively in AI infrastructure. At the same time, some enterprise AI initiatives are being de-scoped, delayed, or challenged on measurable outcomes.

This is where the “year of reckoning” for AI becomes real. AI use cases tied to cost savings, automation, customer experience, risk reduction, and operational efficiency will be easier to defend. Weak GenAI pilots, unclear operating models, or projects without a credible path to production will face more scrutiny from CFOs and cross-functional decision-makers.

AI for AI’s sake will not win the next budget cycle. AI with measurable business impact will.

What to remember for H2 2026 planning

As technology leaders plan for the second half of 2026, three takeaways matter most:

1. APAC IT spending is being disrupted, not derailed
The Middle East War is creating real pressure through energy costs, inflation, supply chain disruption, and weaker business visibility. But priority investments continue to support the market, especially AI infrastructure, cybersecurity, resiliency, and cloud optimization.

2. Budgets are rotating toward measurable value
Lower-priority initiatives, long-payback projects, and unclear business cases are facing delays. Investments that reduce risk, improve efficiency, support compliance, or deliver near-term ROI are better positioned for approval.

3. AI remains resilient, but the proof bar is rising
AI investment remains strong, but enterprise projects need clearer outcomes, stronger business cases, and a faster path from pilot to production.

The path forward is not about spending more for the sake of momentum. It is about knowing where to spend, what to protect, and what to pause.

Register for the on-demand webinar

The Middle East War is the shock. Budget rotation is the signal. Measurable value is the next move.

Register for the on-demand webinar, Asia Pacific IT Spending Outlook 2026 to see IDC’s latest scenario analysis, buyer data, and guidance on where IT spending is holding, where it is slowing, and what it means for H2 2026 planning.

Stephen Minton - Group Vice President, Data & Analytics - IDC

Stephen Minton is a group vice president with the IDC Data & Analytics group, focusing on ICT spending and macroeconomics. Mr. Minton is responsible for Worldwide ICT Spending programs, including the Worldwide Black Book, Worldwide 3rd Platform Spending Guides, and Worldwide Telecom Services Tracker. Mr. Minton's research expertise includes global ICT and economic analysis, and he tracks market data across hardware, software, services, telecom and emerging technologies. He is the author of papers that focus on the economic impact of IT, and is a regular speaker on the subject of IT spending. In 2002 he addressed the United Nations in New York, speaking to UN ambassadors on the subject of the Information Society. Mr. Minton previously worked with Digital Equipment Corporation (DEC), before joining IDC in 1998. Originally from Hartlepool in the North of England, he graduated from the University of Salford in 1995. He has also worked in the field of consumer market research with Millward Brown International.
Vinayaka Venkatesh

Vinayaka Venkatesh - Senior Market Analyst

Vinayaka Venkatesh is a senior market analyst for the IDC Asia/Pacific IT Spending Team based in Bangalore, India. He currently works as a vertical analyst, covering multiple spending guides for Asia/Pacific (excluding Japan and China) (APEJC). He joined IDC in…

レガシーシステムが稼働し続けるたびに、競合他社が優位を築いていく。日本の2.1兆円規模のITモダナイゼーション市場は待ってくれない—変革を急ぐ企業も同様である。

主要指標

1,304億円 — ITモダナイゼーションサービス市場規模(2025年)

10.2% — 年平均成長率(2025〜2030年)

2,123億円 — 市場規模予測(2030年)

約80% — 依然としてレガシーシステムを稼働させている大企業・中堅企業の割合

なぜ日本は世界を上回るペースで成長しているのか

日本のITサービス市場は2024年から2029年にかけて年平均6.6%成長すると予測されており、世界平均の3.6%のほぼ2倍にあたる。その背景には構造的な要因がある。日本は特有の重いレガシー資産を抱えている——長年にわたる汎用機(メインフレーム)やオフィスコンピュータなどへの投資、複雑な個別開発システム、そしてそれらを長年維持してきた人材がある。今、これら三つに起因する課題が重なり合う中、ITモダナイゼーションが避けられないものになっている。

富士通メインフレームのサポート終了

2022年、富士通はメインフレームおよびUNIXサーバー製品の販売・サポートの2030年前後の終了を発表した。この発表により、1,000社以上の企業が後戻りのできないカウントダウンに入り、日本市場全体でITモダナイゼーションの取り組みが加速している。

AIへの対応という至上命題

AIの活用には、緊密に統合されたデータパイプラインと近代的なビジネスプロセス基盤が前提となる——まさにレガシーシステムはこれらの実現を阻む要素となっている。AI競争力を維持したい企業にとって、ITモダナイゼーションはもはや選択肢ではない。

人口動態の圧力

日本のレガシーシステムを構築・維持してきた世代のエンジニアが退職しつつある。そのノウハウや技術が失われてしまう前に、知識とインフラを移行できる時間は着実に縮まっている。

モダナイゼーションへの三つのアプローチ

IDCはITモダナイゼーションサービスを三つの実行タイプに分類しており、それぞれがサービス企業に異なる意味をもたらす。

リホスト

既存のアプリケーション資産を維持しながら、レガシー以外のプラットフォームへリフト&シフトする。予算や移行期間に制約を抱える企業にとっての入口となる手法である。

リライト

ビジネスロジックを変えずに、レガシーのソースコードを現代的な言語に変換する。管理された変革のための中間的なアプローチである。

リビルド

プロセス、データモデル、アーキテクチャをゼロから再定義する。最も高い価値をもたらす一方、最も複雑なアプローチでもある。

短期的には、リホストはリビルドに次ぐ2番目に大きなセグメントであり、メインフレームなどのEOL(End of Life)に対し早急な対応を要するに企業による支出が市場を牽引している——ただし既に成熟期を迎えており、今後はマイナス成長が予測されている。中長期的な成長機会は、アプリケーションのモダナイゼーション——リライト、リファクタリング、マイクロサービス化やクラウドネイティブアーキテクチャの採用——にある。

国内ITモダナイゼーションサービス市場 支出額予測: 2025年~2030年

Source: IDC Japan, 2/2026

企業がサービスプロバイダーに本当に求めているもの

IDCの調査では、レガシー依存度が相対的に高い大企業・中堅企業は、単なる技術的な実行だけを求めているのではなく、変革のパートナーを求めていることがわかった。セキュリティは基本的な前提として期待される一方、上位のニーズにはビジネスプロセス変革の支援やクラウド活用支援が挙がっている。

需要のシグナルはセクターによっても明確に異なる。

金融サービス

クラウドネイティブなアプリケーション開発能力、すなわち近代的なインフラ上で素早くイノベーションを起こす能力を優先している。

製造・流通

ビジネスプロセスの変革を優先している。基盤となる技術を刷新するだけでなく、業務に効率性とインテリジェンスを組み込むことを重視している。

全セクターを通じて、IDCは企業の期待に一貫した変化を観察している。ビジネス上の成果が主要な購買基準になりつつある。技術的な能力は当然のこととして見なされ、価値の創出が差別化要因となっている。

今、勝てるポジションを築くために

サービス企業にとって、競争上の必要性は明確だ。この市場で勝利する最良のポジションにある企業は、次の三つを実行する。

1. レガシーモダナイゼーションの実績を体系化する

過去の案件は活用されていない資産だ。サービス企業は、達成したビジネス成果——コスト削減、リードタイムの改善、AI対応力の解放——を体系的にまとめた資料を構築し、これを市場への訴求の核にすべきである。

2. AIの時代に向けた業種別のリファレンスアーキテクチャを開発する

汎用的なモダナイゼーションの提案は説得力を失いつつある。企業は自社のセクター、規制環境、そしてAIへの志向に合わせたシステムアーキテクチャと実装ロードマップを求めている。

3. 需要に先行してアプリケーションモダナイゼーション能力に投資する

リホストの波は既にピークに差し掛かりつつある。高い利益率をもたらす機会——リライト、リファクタリング、リビルド——がその後に続いている。クラウドネイティブとマイクロサービスの深い能力を培ったサービス企業こそが、2030年に向け企業から選ばれる存在となる。

IDCが提供するレポートのご紹介

IDCでは、国内ITモダナイゼーション市場の動向を詳細に分析したレポートを発行しています。

本調査レポートは、IDCの国内サービス市場予測における主要な成長促進要因の一つであるレガシーシステム(老朽化・陳腐化、肥大化・複雑化、ブラックボックス化したシステム)のITモダナイゼーションについて、市場規模の中期予測を示すと共に、国内企業(ITバイヤー)の取り組み動向や、それを支援するサービスベンダーの動向を分析しています。国内ITモダナイゼーションサービス市場予測では、サービスセグメント別、実行タイプ別(リホスト、リライト、リビルド)、システムタイプ別、産業分野別に予測しています。これらの分析から、国内企業のITモダナイゼーション支援におけるニーズ変化や市場機会、サービスベンダーの支援サービスの特徴や戦略を包括的に把握できます。

関連する調査やご相談について

より詳細なインサイトや市場動向については、当社アナリストへお気軽にご相談ください。

Masaru Muramatsu - Senior Research Analyst, Software, Services, and IT Spending, IDC Japan - IDC Japan

Masaru Muramatsu is a senior research analyst, responsible for research and analysis of the Japanese IT services market, including IT consulting, systems integration, business services. Prior to joining IDC, Masaru worked to help digitalize local government in Japan, implementing software as a service (SaaS) in the education and taxation sectors. He also acquired experience in domestic and international sales/marketing with his work for a company that provided materials for electronic devices like smartphones, PCs, and printers. Masaru Muramatsu earned a master’s degree in engineering from Chuo University, Japan.

For years, digital accessibility, the practice of ensuring that digital products and services can be perceived, understood, and used by everyone regardless of ability, was treated as a compliance checkbox. That framing is no longer adequate. AI is reshaping accessibility into a strategic capability, one that is adaptive, continuous, and embedded in how people work, interact, and innovate.

As AI-enabled work becomes the norm, accessibility is no longer about supporting a small subset of users. It is about ensuring that everyone, across physical, sensory, cognitive, and neurodiverse dimensions, can fully participate in increasingly digital and AI-mediated environments. In this context, accessibility becomes foundational to productivity, inclusion, and ultimately business performance. Accessibility is also part of company culture: involving disabled and neurodiverse individuals in co-design, not just testing, creates more robust and adaptable systems. Sustaining long-term impact also requires investment in skills and culture, training employees, fostering inclusive design practices, and making accessibility a shared responsibility across teams.

The opportunity: AI as a scaler of inclusion and innovation

AI introduces a powerful opportunity to rethink accessibility at scale.

First, it enables real-time content adaptation. Capabilities such as automatic captioning, transcription, translation, and alternative text generation allow organizations to dynamically tailor content to different user needs. AI can also adjust reading levels, restructure complex information, and personalize interaction styles, supporting a broader range of cognitive and sensory preferences.

Second, AI supports continuous accessibility operations. Traditionally, accessibility has relied on periodic audits and remediation efforts. AI-driven testing tools now allow organizations to embed accessibility checks directly into development pipelines, transforming accessibility into a continuous, iterative process aligned with DevOps cycles.

Third, AI helps democratize innovation. By making tools and workflows more accessible, organizations can engage a wider and more diverse talent pool, including neurodiverse individuals and those historically underserved by traditional work environments. This expands creative input, improves problem-solving, and strengthens organizational resilience.

Finally, AI enables data-driven accessibility insights. Organizations can use AI to analyze accessibility barriers, monitor usage patterns, and measure outcomes, linking accessibility directly to business metrics such as productivity, employee engagement, and customer satisfaction.

The pitfalls: Bias, complexity, and the risk of scaling barriers

Despite its promise, AI also introduces significant risks that organizations must actively manage.

One of the most critical challenges is bias in AI models. Many AI systems are trained on data and designed by teams that lack diversity. This can result in outputs that unintentionally exclude or disadvantage certain groups, particularly people with disabilities or non-standard interaction patterns. Without deliberate inclusion in design and testing, AI can reinforce existing barriers or create entirely new ones. Feedback loops that combine AI-driven insights with real user experiences are essential to countering this risk.

Another risk lies in inaccessible AI-generated content. While generative AI can produce fluent and polished outputs, these may still fail accessibility standards through improper structure, missing semantic cues, or formats that are difficult for assistive technologies to interpret. Auto-generated captions, for example, are often not accurate enough for compliance purposes.

The rise of agentic AI systems (autonomous AI that acts across workflows and applications without direct human instruction at each step) adds further complexity. If poorly designed, they can propagate inaccessible processes at scale, embedding friction into core operations rather than eliminating it.

There is also a governance challenge. As AI becomes embedded across systems, organizations must ensure clear accountability, transparency, and control over how accessibility preferences are handled, how decisions are made, and how user data is used.

Recommendations: Turning intent into impact

Organizations that want to lead in AI-enabled accessibility should focus on four key actions:

  • Prioritize accessibility as a design principle. Move from reactive compliance to proactive, accessible-by-design systems embedded in AI-enabled platforms and services.
  • Establish proactive AI accessibility governance. Integrate accessibility into AI governance frameworks early, ensuring inclusive workflows and avoiding costly retrofits.
  • Design for workforce adaptability and inclusion. Extend accessibility strategies beyond compliance to support diverse employee needs, including neurodiversity, aging workforces, and varying cognitive styles.
  • Act early to mitigate risk and maximize value. Early investment reduces remediation costs, strengthens trust, and positions accessibility as a strategic differentiator rather than a regulatory burden.

AI is redefining digital accessibility as a core element of how organizations operate, innovate, and compete. Those that embrace accessibility as a strategic priority will not only meet regulatory requirements but also unlock broader talent, improve user experiences, and build more resilient AI systems.

Erica Spinoni

Erica Spinoni - Senior Consultant, IDC4EU IDC European Government Consulting

Erica Spinoni is a Senior Consultant at the IDC4EU European Government Consulting unit. She led IDC’s EMEA-focused research for the Future of Work group, examining how AI, automation, and emerging workplace models are reshaping employee experience and productivity across the…
Amy Loomis, Ph.D.

Amy Loomis, Ph.D. - Group Vice President, Workplace Solutions

Amy Loomis is Group Vice President for IDC’s worldwide Workplace Solutions.  Amy leads a team of analysts focused on the evolving nature of human resources, skills development, collaboration, and leadership across the employee lifecycle. Her research into the Future of…
Melinda-Carol Ballou

Melinda-Carol Ballou - Research Director, AI Assurance, ALM, Quality & Portfolio Strategies

Melinda Ballou delivers insights into the future of AI assurance, the impact of AI, ML and agentic adoption on agile and digital work, resilience, quality, product and software engineering, the role of technology in business and culture, and the evolution…

If you’ve spent the last few years talking to enterprise IT leaders about cost efficiency, you weren’t wrong. That was the conversation. But over the past few months, things have clearly shifted.

The outbreak of war in the Middle East, with its direct impact on people and organizations in the region, as well as broader effects on energy costs and IT manufacturing supply chains, is a primary driver. At the same time, early AI buildout pressures on memory supply were already raising concerns.

Today, when CIOs and their teams make technology decisions, the question is no longer, “How do we optimize spend?” It’s, “How do we keep the business running when things break?”

This shift shows up clearly in data from two major surveys on IT priorities and spending plans conducted in February and again in March. Concerns about hardware supply constraints have increased by more than 15%, and geopolitical risk is rising quickly. Meanwhile, traditional cost pressures, while still present, are starting to take a back seat.

This is not because cost no longer matters. It is because cost is now seen as downstream. If systems go down, supply chains stall, or cyber incidents escalate, cost becomes secondary very quickly.

What are IT leaders most concerned about in 2026?

When you talk to IT leaders today, the tone is different. There is more urgency, more realism, and more skepticism. They are thinking about exposure:

  • Where are we too dependent on a single cloud region?
  • What happens if a supplier cannot deliver?
  • How quickly can we recover from a cyber event?

Increasingly, they recognize that these risks are interconnected. A geopolitical event can disrupt supply chains, which impacts infrastructure, which affects applications, and ultimately hits revenue.

That is why IDC is seeing a clear pivot toward resilience.

Cybersecurity has moved to the top of the investment list globally, not just as a defensive measure but as a core part of keeping operations running. At the same time, organizations are accelerating investments in multi-region cloud architectures and backup strategies. Cloud security and multi-region resilience are now leading priorities across every major region.

IDC is also hearing from CIOs about a growing push to reduce dependency. CEOs are placing more focus on diversifying suppliers across all parts of the business. CIOs are responding by exploring sovereign cloud options and rethinking how and where infrastructure is deployed.

AI has not disappeared from the agenda, but it is being reframed. It is no longer just about innovation. It is about using automation and intelligence to keep systems stable under pressure.

Put simply, IT buyers are trying to build systems that can bend without breaking.

What does this shift mean for IT suppliers?

For suppliers, this shift creates both risk and opportunity.

The biggest risk is continuing to sell the way you did before. Leading with performance benchmarks, cost savings, or incremental features will not resonate the same way.

The opportunity is much bigger. Buyers are actively looking for partners who can help them navigate uncertainty. They are asking tougher questions:

  • What happens if this service goes down in one region?
  • How quickly can workloads move?
  • Where are the hidden dependencies?
  • How exposed am I if conditions worsen?

If you can answer these questions clearly and credibly, you move from being a vendor to becoming a strategic partner.

How should IT suppliers respond to rising resilience demands?

The challenge is that resilience means something different depending on where you sit in the ecosystem. The common thread is this: you must show how your offering performs under stress, not just under ideal conditions.

Cloud providers: How to prove resilience beyond scale

For cloud providers, this is a moment to rethink the narrative.

Scale and efficiency still matter, but they are no longer enough. CIOs want to know how your platform behaves when a region is disrupted, connectivity is constrained, or workloads need to move quickly.

This means making multi-region resilience the default, not an add-on. It also requires transparency about risk exposure and greater flexibility around sovereignty and localization.

In short, you are not just selling capacity anymore. You are selling survivability.

SaaS providers: Why continuity is now a core differentiator

SaaS providers are increasingly part of the critical path of operations. If your application goes down, the business feels it immediately.

Buyers want reassurance. They want to understand your disaster recovery posture, regional architecture, and dependencies. They want to know how their data is protected and how quickly services can be restored.

The vendors that stand out will clearly articulate how they maintain continuity, not just deliver functionality.

IT and professional services firms: From transformation to readiness

For services firms, the conversation has shifted from long-term transformation to immediate readiness.

Clients still care about transformation, but right now they need help answering urgent questions: Where are we exposed? What should we fix first? How do we prepare for multiple scenarios?

There is a real opportunity to lead with practical, actionable support. Rapid assessments, scenario planning, and resilience design are where clients need help now.

Speed matters. Clarity matters even more.

Communications providers: Why network resilience is now critical infrastructure

Connectivity has always been important. Now it is critical infrastructure in the truest sense.

Organizations are looking for redundancy, alternative routing, and, in some cases, entirely new connectivity models, including satellite and hybrid networks.

The differentiator is reliability under pressure. If you can demonstrate that your network keeps people and systems connected when other options fail, that becomes a powerful advantage.

Infrastructure vendors: Delivering certainty in uncertain supply chains

Hardware vendors are facing a different kind of scrutiny.

Availability and certainty in delivery are becoming as important as performance. Buyers want to know not just what the system can do, but whether they can actually get it, deploy it, and rely on it.

Transparency into supply chains, flexibility in configurations, and the ability to adapt to constraints are becoming key differentiators. In this environment, certainty is value.

Why IT buying decisions are shifting from optimization to assurance

Stepping back, what we are seeing is a shift in how technology decisions are made.

It is less about optimization and more about assurance. Less about peak performance and more about consistent operation.

The suppliers that win over the next six months will be the ones that can answer a simple but critical question:

What happens when things do not go according to plan?

From an enterprise IT leader’s perspective, that is no longer a hypothetical. It is the reality they are planning for every day. Resilience is no longer just a capability. It is the basis for trust.

What should IT suppliers do next?

If you are an IT supplier, now is the time to recalibrate how you engage with customers.

Start by pressure-testing your value proposition:

  • Can you clearly articulate how your offering performs under disruption?
  • Can you quantify how you improve resilience, not just efficiency?
  • Can you help customers understand and reduce their exposure?

Just as importantly, ground your strategy in real buyer insight.

IDC’s latest Future Enterprise Resiliency & Spending Survey (March 2026, Wave 2) provides a detailed view into how enterprise IT leaders across regions are reprioritizing risk, resilience, and investment decisions in response to geopolitical and supply chain disruption.

We encourage you to explore the survey findings to better understand:

Suppliers that align early with these shifts will be better positioned to engage, differentiate, and win. Because in this market, insight isn’t just helpful.

It’s your competitive edge.

Rick Villars

Rick Villars - Group Vice President Worldwide Research

Rick is IDC's leading analyst guiding research on the future of the IT Industry. He coordinates all IDC research related to the impact of Cloud and the shift to digital business models across infrastructure, platforms, software, and services. He helps…

AI is not just changing job descriptions; it is actively rewiring how work is coordinated, controlled, and created, and it is doing so on multiple fronts at once, inside the same organization.

AI Is Transforming Work on Multiple Fronts Simultaneously

Some of our IDC Future of Work predictions bring this into sharp focus: by 2027, 40% of current job roles in large organizations will be redefined or eliminated, accelerated by GenAI adoption. At the same time, by 2030, around 70% of new job roles in Europe are expected to be directly enabled by AI technology. This is not a neat “old jobs out, new jobs in” swap. It is a systemic reconfiguration of how value flows through the enterprise. Yet most leadership frameworks still present AI scenarios as if they were mutually exclusive: automate to cut headcount, augment to boost productivity, redesign work for agility, or push toward autonomous operations.

When Automation, Augmentation, and Autonomy Collide

On the ground, those dynamics do not arrive one by one; they collide. In the same business unit, you may be cutting FTEs as routine tasks are automated and taken over by “digital colleagues,” while simultaneously hiring AI orchestrators, prompt engineers, and automation product owners to keep up with demand for AI-adjacent skills. You may be tearing up long-standing workflows as agentic systems reshape a significant share of knowledge work, at the same time as parts of your operation drift toward near-autonomous execution, powered by employees building personal agents and conversational workflows that quietly absorb whole segments of the process. These are not options on a slide; they are concurrent forces acting on the same organizational fabric. Treating them like menu choices is not workforce planning. It is misdiagnosing an organizational phase transition, a fundamental shift in the underlying architecture of how work happens.

From Role-Based Models to Capability-Based Architectures

The uncomfortable truth is that many leaders are still planning for roles, new and “to be eliminated,” while AI is reshaping the landscape at the level of capabilities and architecture. You can see the tension in three simple signals. A clear majority of European organizations have already deployed or are piloting automation to offset chronic labor shortages. A growing share of executives openly discusses replacing positions with automation, and many plan to substitute a measurable portion of their workforce with “digital colleagues.” Meanwhile, by the end of this year, a meaningful slice of frustrated knowledge workers with no formal development background will be building their own agentic workflows to change how they work, regardless of what HR’s role catalog says. When people can spin up an agent in a week, any static role taxonomy you publish today is out of date tomorrow. The center of gravity moves from “what roles do we have?” to “what capabilities can we compose, and how fluidly can we recombine them as AI matures?”

Why Traditional Role Models No Longer Hold

Role-centric models allow for some seriously wrong assumptions: that tasks are stable enough to bundle into jobs, that jobs are stable enough to plan around for three to five years, and that hierarchies are stable enough to govern how value flows. Agentic AI quietly breaks all three. Tasks fragment, recombine, and migrate between humans and machines in near real time. Work starts to look less like a tidy org chart and more like a living graph of capabilities, human, machine, and hybrid. In that context, planning headcount against static job descriptions is like trying to architect a cloud-native platform using only server rack diagrams.

Architecture Determines the ROI of AI

However, IDC’s Future of Work research also shows that when enterprises invest in digital adoption and automated learning technologies, they can unlock substantial productivity gains. The pattern across these findings is consistent: it is the architecture that determines the yield of AI, not just the tools themselves. If your workflows are fragmented, AI struggles to “see” the end-to-end journey it needs to transform. When critical data is locked in legacy systems, it cannot provide the rich, contextual recommendations you were promised. When governance is tuned for stability rather than experimentation, it throttles the learning cycles AI needs to be useful. Layer on top the reality that many organizations openly acknowledge they lack the capability support to implement automation effectively, and a clear picture emerges.

AI Amplifies Existing Organizational Weaknesses

In that environment, throwing more AI at the problem does not fix anything. It amplifies what is already there. Bad processes simply run faster. Poor decisions scale further. Shadow automation blooms in the gaps, as frustrated employees script around the constraints of the operating model. AI becomes an accelerant, not a cure.

Reframing the Strategic Question for Leaders

This is why the strategic question has to change. Instead of asking, “Which jobs will we automate?”, leaders need to ask, “Is our organization structurally able to absorb intelligence at scale?” Answering that requires moving from headcount planning to capability mapping, designing work around the interplay between human strengths, judgment, domain expertise, relationship-building, and machine strengths such as pattern recognition, generation, and orchestration. It means treating architecture as a product: standardizing interfaces, workflows, and data contracts so AI can plug into work without bespoke integration every single time. It means tracking how many workflows, decisions, and customer journeys are genuinely enhanced by AI, not just how many licenses have been bought. And it means steering reduction, augmentation, redesign, and autonomy as one coherent portfolio of change, not four disconnected projects.

Conclusion: The Real Stress Test Is Your Operating Model

AI is already changing jobs. The real test is whether your operating model can evolve quickly enough to harness that change, or whether AI will simply accelerate you toward the limits of the system you already have.

If you would like more information, drop your details in here.

Meike Escherich - Associate Research Director, European Future of Work - IDC

Meike Escherich is an associate research director with IDC's European Future of Work practice, based in the UK. In this role, she provides coverage of key technology trends across the Future of Work, specializing in how to enable and foster teamwork in a flexible work environment. Her research looks at how technologies influence workers' skills and behaviors, organizational culture, worker experience and how the workspace itself is enabling the future enterprise.

Key figures at a glance

  • ¥1,304B – IT modernization services market size, 2025
  • 10.2% – Projected average annual growth rate, 2025–2030
  • ¥2,123B – Forecast market size by 2030
  • ~80% – Large and mid-sized enterprises still running legacy systems

Why Japan is outpacing the world

Japan’s IT services market is forecast to grow at a CAGR of 6.6% from 2024 to 2029, nearly double the global average of 3.6%. The answer is structural. Japan carries a uniquely heavy legacy burden, decades of investment in proprietary mainframe environments, complex bespoke systems, and a workforce that has long maintained them. Now, three forces are converging to make modernization unavoidable:

  • Fujitsu Mainframe Sunset – In 2022, Fujitsu announced the end of sales and support for its mainframe and UNIX server products around 2030. This single announcement put more than 1,000 enterprises on an irreversible countdown, accelerating timelines across the entire Japanese market.
  • AI Readiness Imperative – AI adoption presupposes tightly integrated data pipelines and modern business process architectures, exactly what legacy systems make impossible. Modernization is no longer optional for companies that want to remain AI-competitive.
  • Demographic Pressure – The generation of engineers who built and maintained Japan’s legacy systems is retiring. Organizations face a narrowing window to migrate knowledge and infrastructure before institutional memory disappears entirely.

Three paths to modernization

IDC segments IT modernization services into three execution types, each with distinct implications for services firms:

  • Rehost – Lift-and-shift to non-legacy platforms. Preserves existing application assets. The near-term entry point for enterprises constrained by budget or migration timelines.
  • Rewrite – Convert legacy source code to modern languages without changing business logic. A middle path for controlled transformation.
  • Rebuild – Redefine processes, data models, and architecture from the ground up. The highest-value, highest-complexity path.

Near-term, rehost is the second-largest segment after rebuild, driven by enterprises responding urgently to mainframe end-of-life deadlines — though it has already reached maturity and is forecast to decline. The mid-to-long-term growth opportunity lies in application modernization, rewriting, refactoring, and the adoption of microservices and cloud-native architectures.

What enterprises need from services firms

IDC surveyed large and mid-sized Japanese enterprises and found that organizations with significant legacy exposure do not simply want technical execution, they want transformation partners. Security remains a baseline expectation, but top-ranked needs now include business process redesign and cloud architecture strategy.

Demand signals also diverge meaningfully by sector:

  • Financial Services – Prioritizes cloud-native application development capabilities, the ability to innovate rapidly on modern infrastructure.
  • Manufacturing and Distribution – Prioritizes business process transformation, embedding efficiency and intelligence into operations, not just upgrading the underlying technology.

Across all sectors, IDC observes a consistent shift in enterprise expectations: business outcomes are becoming the primary purchase criterion. Technical competence is assumed; value creation is the differentiator.

How to build a winning position now

For services firms, the competitive imperative is clear. The service providers best positioned to win this market will do three things:

1. Codify your legacy modernization track record

Past engagements are an underutilized asset. Service providers should build structured libraries of business outcomes achieved, cost reductions, cycle time improvements, AI readiness unlocked and make these the core of their go-to-market narrative.

2. Develop industry-specific reference architectures for the AI era

Generic modernization pitches are losing traction. Enterprises want system architectures and implementation roadmaps calibrated to their sector, their regulatory environment, and their AI ambitions.

3. Invest in application modernization capabilities ahead of demand

The rehost wave is already approaching its peak. The high-margin opportunity –  rewrite, refactor, rebuild – is building behind it. Service providers who develop deep cloud-native and microservices capabilities now will be the ones enterprises turn to in the second half of this decade.

About the IDC Report

IDC has published a comprehensive analysis of Japan’s IT modernization market: 2026 Japan IT Modernization Market Analysis. The report provides a medium-term market forecast for IT modernization of legacy systems — a primary growth driver in IDC’s Japan IT services market outlook. Legacy systems are characterized by aging and obsolescence, excessive complexity and scale, and a lack of transparency. It covers enterprises’ IT modernization trends and an analysis of the service vendors’ services trend. Market forecasts are segmented by service type, execution type (rehost, rewrite, rebuild), system type, and industry vertical. Together, these analyses offer a comprehensive view of shifting enterprise needs, emerging market opportunities, and the strategies and service offerings of leading vendors in Japan’s IT modernization landscape.

For more detailed insights and market trends, please contact our analysts by completing this form IDC | Identifying Market Opportunities – Contact Us.

Masaru Muramatsu - Senior Research Analyst, Software, Services, and IT Spending, IDC Japan - IDC Japan

Masaru Muramatsu is a senior research analyst, responsible for research and analysis of the Japanese IT services market, including IT consulting, systems integration, business services.

中国公有云市场正在经历一场由AI驱动的结构性转折。IDC最新发布的《中国公有云服务市场跟踪报告,2025 下半年》数据显示,2025年下半年公有云IaaS市场人民币同比增速恢复至20.0%,整体公有云市场(IaaS+PaaS+SaaS)半年总值首次突破2000亿元。这一增长并非简单的市场回暖,其背后是AI需求对云计算产业底层逻辑的重塑:市场认可度的提升正在从“资源规模”转向“全栈AI能力”,市场份额正加速向“算力+大模型”双强厂商收敛,出海成为增长第二曲线,而行业间的需求分化也将进一步拉大。IDC认为,未来三到五年,公有云市场的竞争将不再是价格战与规模战,而是算力、模型、行业方案、生态与全球化能力的综合较量。谁能在这一轮AI红利中真正从“云厂商”升级为“AI服务商”,谁就将主导下一轮市场格局。

AI 能力将重构市场份额 ,大模型为云厂夺回定价权

在传统IaaS市场中,云主机租赁长期占据主导,价格战也因此成为竞争的主旋律。然而,随着生成式AI和大模型需求的爆发,客户在选择云服务商时的核心依据正在发生根本性变化。算力、大模型、芯片与平台的全栈能力,正取代过去的资源规模与价格优势,成为新的竞争焦点。相应地,定价逻辑也在从资源计费转向价值付费。

这一转变最直接的体现,是市场增长动力从存量博弈转向增量创新。传统IaaS价格战逐步退潮,取而代之的是智算需求驱动的新一轮扩张。大模型产品的定价模式正向按Token计费倾斜,云基础设施产品,尤其是高性能算力的需求大幅提升。在这一新竞争环境中,云服务商是否具备AI原生能力与高效的算力调度能力,将直接决定其能否占据有利位置。

全栈能力成为竞争壁垒,市场向“算力+大模型”双强厂商收敛

智算集群、液冷数据中心、自研芯片、大模型训练等领域均属于高投入赛道,只有具备完整闭环能力的厂商,才能实现算力规模化变现并穿越投入周期。根据IDC MarketScape评估,阿里云、百度智能云等具备全栈能力的云厂商,其市场份额正持续提升。这些厂商凭借硬件芯片、异构算力兼容、集群调度、MaaS平台及行业生态等综合优势,已在政企、互联网等多个行业中积累起丰富的落地案例。

相比之下,单纯依赖资源出租的厂商,由于缺乏技术与生态支撑,增速已开始放缓。未来市场将进一步向具备全栈AI能力、生态协同能力和行业深度适配能力的头部厂商集中。

行业分化加剧,高适配行业领跑

不同行业在公有云需求上的差异正在拉大。自动驾驶、电商、游戏、互联网金融、协同办公等与AI结合度高且资本红利充足的领域,公有云需求保持高速增长。IDC预测,2025年泛科技行业的AI公有云渗透率将持续提升,行业间的市场增速差距将进一步扩大。而政务、制造、传统金融等行业,则受合规要求和系统改造周期等因素限制,上云节奏相对缓慢。这些行业客户在选型时,更加关注模型的训练与推理性能、行业精调落地情况、数据主权与合规、成本治理等多维度能力,这也在推动云服务商加快行业化与场景化产品的布局。

海外资源加速布局,出海成为增长第二曲线

海外背景的云厂商,如AWS和微软Azure,在云算力出海领域依然保持强劲竞争力,持续服务于中资企业的全球化部署与跨境AI业务。与此同时,阿里云、腾讯云、华为云以及运营商阵营也在积极瞄准中资企业出海需求,在跨境电商、游戏发行、AI应用出海等领域加速落子。根据IDC跟踪数据显示,中过企业出海用云市场规模五年复合增长率超过30%,远高于国内市场增速。出海正成为中国云厂商寻找增长“第二曲线”的重要方向。

AI红利花落谁家:三类厂商各显其能

在这一轮AI驱动的市场重构中,不同类型的云服务商正走出截然不同的增长路径。

以阿里云和百度智能云为代表的全栈型厂商,凭借“模型+芯片+平台+公有云基础设施”形成的产品线闭环,构建起公有云竞争的护城河。其公有云IaaS同比增速已从2024年的个位数增长提升至2025年的25%以上,市场份额持续提升,并在大模型平台、MaaS、行业精调等领域不断加码。

以腾讯云和火山引擎为代表的场景驱动型厂商,则更加聚焦AI的商业化落地,推动将AI真正“用起来”。腾讯云在2025年首次实现规模化盈利,而火山引擎则凭借高性价比的智算方案与灵活计费模式,在2025年再次实现超100%的同比增速,市场份额快速提升。

以中国电信天翼云、移动云和华为云为代表的算力运营型厂商,依靠自建与跨平台调度能力的结合,灵活适配第三方方案,为政企、金融等高安全性行业的AI应用提供保障。凭借多年积累的机房资源、属地化服务和央企背景,这些厂商的企业级服务优势逐步显现,市场排名稳居前五。

从算力到模型:未来四阶段演进路径正在形成

展望未来,中国公有云市场的竞争将沿着清晰的阶段路径演进。

在2026年的第一阶段,算力投入依然是AI发展的核心方向。智算集群、自研芯片、数据中心建设推动资本开支持续攀升。IDC预测,到2027年,超过85%的中国组织将把传统云环境转型为适配AI工作负载的新型平台。

到2027年前的第二阶段,商业化将迎来突破。云服务商的营收模式将从单纯算力出租,转向“Token+算力”的双营收结构,收费模式从资源侧向场景侧迁移,云厂商的盈利格局将随之重塑。大模型驱动的AI云服务市场格局正在形成,MaaS、行业精调和Agent平台等新型商业模式加速落地。

进入2028年的第三阶段,竞争焦点将从算力底座转向大模型应用场景的的训练、微调和推理优化。“云上模型好用度”将成为企业客户选择云服务商的决定性因素。企业买家将优先评估多模态模型覆盖能力、推理准确性、行业适配能力以及生态工具链的完备性。

而纵观未来五年,行业分化将成为长期特征。泛科技行业将持续领跑,传统行业则在政策引导与国产化适配推动下逐步而坚定的推进上云进程。市场将向具备全球化能力、行业方案能力和生态协同能力的综合AI服务商集中,头部厂商将通过全栈能力与行业深耕构建起长期壁垒。

分析师结语:从“云厂商”到“AI服务商”的升级之战

2025年下半年中国公有云市场重回高增长,其本质是AI产业爆发所带来的公有云基础设施红利。IDC中国研究经理崔婷婷表示,IaaS增速重回20%以上,标志着中国公有云行业已从存量博弈转向增量创新。未来三到五年,AI能力的竞争将进入白热化阶段,市场不再是简单的价格与规模比拼,而是算力、模型、行业方案、生态与全球化能力的综合较量。云计算,尤其是公有云服务,作为AI竞争的核心载体,其资源铺设广度、能力韧性、安全性、营收增长与利润转换率的动态提升,也将直接反映出AI阵营的发展状态,成为AI竞争态势的晴雨表。在这场升级之战中,谁能更快将AI能力转化为客户可感知的业务价值,谁就有望在下一轮格局洗牌中占据先机,真正完成从“云厂商”到“AI服务商”的跃迁。

如需进一步了解IDC相关研究,或就中国公有云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。

请点击此处与我们联系。

The global semiconductor market is undergoing a seismic transformation. IDC’s latest forecast projects the industry will surge past the $1 trillion revenue threshold in 2026, significantly ahead of prior expectations. The growth will be driven overwhelmingly by AI infrastructure investment, which is reshaping the entire market.

Total semiconductor revenues are forecast to reach $1.29 trillion in 2026, up 52.8% year over year from $842.8 billion in 2025. The memory segment is at the epicenter of this shift: DRAM revenues alone are projected to nearly triple in 2026 to $418.6 billion, driven by demand for high-bandwidth memory (HBM) and DDR from hyperscalers and AI infrastructure providers. Meanwhile, non-memory semiconductors are growing at a robust but more measured pace, reaching $693.5 billion in 2026.

In this post, we break down three forces reshaping semiconductors right now: why AI infrastructure has become the industry’s new center of gravity, what’s happening in memory markets and why it matters beyond the data center, and how other markets from automotive and IoT to mobile and PCs are navigating a market increasingly defined by AI.

Global Semiconductor Market: Selected Forecast (USD Billions)

Source: IDC Semiconductor & Semiconductor Applications Forecast, April 2026. A = Actual, E = Estimate, F = Forecast.

AI infrastructure: The engine of the supercycle

The single most consequential shift in the semiconductor market is the emergence of AI infrastructure as a structurally dominant end market. What began as a cyclical uplift in data center spending has evolved into a self-reinforcing investment cycle that is reshaping demand patterns across the semiconductor value chain.

Hyperscale capital expenditure exceeded $100 billion for the first time in Q3 2025, and the i4 are expected to increase capex by 70% year over year to approximately $600 billion in 2026. IDC forecasts data center semiconductor revenues to reach $477.1 billion in 2026. By 2030, data center semiconductors will account for $843.2 billion, nearly half the total semiconductor market.

Datacenter Semiconductor Revenue Decomposition

Source: IDC Semiconductor Applications Forecast, April 2026.

The $281 billion “intelligent” datacenter segment, encompassing CPUs, AI accelerators, GPUs, custom ASICs, and networking silicon, now constitutes the largest identifiable category within non-memory semiconductors. Spending is heavily concentrated among top-tier hyperscalers and a growing set of sovereign AI infrastructure programs, many of which have secured long-term supply agreements with leading chip manufacturers.

Three factors are keeping this growth self-sustaining rather than cyclical:

  • Compute intensity continues to rise. Generative AI and agentic workloads require far more compute density per rack than prior architectures, increasing the overall silicon footprint
  • Inference demand compounds on itself. Each new model generation increases the volume of inference, requiring ongoing hardware upgrades
  • AI is spreading beyond the data center. As enterprises, edge deployments, and client devices begin running AI workloads locally, demand becomes more distributed

Memory: From cyclical commodity to strategic constraint

If you want to understand what’s really happening in semiconductors right now, start with memory.

Total memory revenues rise from $226 billion in 2025 to $594.7 billion in 2026, and then to $790.4 billion in 2027. This is not simply a recovery cycle, it reflects a market that is being structurally repriced.

DRAM is where the shift is most visible. IDC forecasts $418.6 billion in DRAM revenues for 2026, up 177% year over year. This is not primarily a volume story driven by consumer devices. Hyperscalers are buying a fundamentally different, more expensive class of memory and are willing to pay a premium to secure supply. Each HBM chip also requires significantly more silicon real estate, further tightening the availability of other types of DRAM.

The HBM bottleneck

High-bandwidth memory has become the primary constraint in the AI accelerator supply chain. Most capacity is already pre-committed through 2026, with forward allocations extending into 2027. That capacity is concentrated in NVIDIA and AMD GPU platforms, along with a growing set of hyperscaler custom silicon programs.

The production economics are also very different. HBM relies on advanced packaging and stacking technologies, resulting in per-bit costs that are several times higher than standard DRAM.

Suppliers are investing aggressively to expand capacity, but the technical complexity and capital intensity mean meaningful new supply will not reach the market until late 2026 at the earliest.

NAND: AI drives storage demand

NAND Flash revenues are forecast to reach $174.1 billion in 2026, up 138.5% from 2025. AI infrastructure is again the dominant driver, with demand coming from training datasets, checkpoint storage, and high-performance inference environments.

Unlike DRAM, the NAND market is seeing broader repricing. Enterprise SSD prices have surged as hyperscalers secure supply, which is tightening availability across consumer and OEM channels.

Other markets: Navigating the shadow of the AI supercycle

While AI infrastructure dominates the headlines, the broader semiconductor market is facing a more nuanced environment.

Non-memory, non-datacenter revenues are projected at $406.3 billion in 2026. Several end markets are dealing with margin pressure, supply allocation challenges, and macroeconomic headwinds.

In mobile, semiconductor revenues are forecast to decline to $89.8 billion in 2026. The issue is not consumer demand, particularly for AI-capable devices, but cost pressure. Memory now represents a larger portion of the bill of materials, forcing OEMs to make difficult tradeoffs between margin, pricing, and product specifications.

Automotive is being shaped more by macro factors than AI. Tariffs, interest rates, and energy prices are weighing on demand. While the long-term outlook remains strong, 2026 reflects a period of near-term softness.

IoT shows a similar pattern. The segment is projected at $136.6 billion in 2026, with near-term pressure from inventory digestion and cautious spending. However, edge AI is beginning to create a new, higher-value demand category that will become more meaningful over time.

Source: IDC Semiconductor Forecast, April 2026.

Outlook: Path to $1.75 trillion

IDC’s base case projects semiconductor revenues reaching $1.75 trillion by 2030.

Several dynamics will shape that trajectory:

  • Memory pricing will normalize, but remain structurally higher than pre-AI levels
  • Non-memory semiconductors will continue steady growth, driven by AI adoption across devices and industries
  • Macro and geopolitical risks will remain important variables

What is clear is that the semiconductor market has undergone a fundamental shift.

IDC will be tracking how AI infrastructure investment continues to reshape semiconductor demand at Computex 2026.

Jeff Janukowicz

Jeff Janukowicz - Vice President, Computing Systems Platforms and Technologies, Enterprise Infrastructure

Jeff Janukowicz is a Vice President within IDC’s enterprise infrastructure global research domain, and the global subdomain lead for the Computing Systems Platforms and Technologies subdomain. Jeff and his team deliver data-driven analysis, technology insights, market trends, and strategic guidance…
Nina Turner

Nina Turner - Research Director, Computing Systems Platforms and Technologies, Enterprise Infrastructure

Nina Turner is Research Director within IDC's enterprise infrastructure global research domain, part of the computing systems, platforms, and technologies subdomain. Nina and her team cover the breadth of processors and architectures, from datacenters to client devices, including embedded use…