AI-driven development has started to reshape profit margins for Japanese IT service providers

Overview

This report focuses on the Japanese IT and software industry, especially large system integrators, and productivity improvement in AI-driven and AI-native development. It argues that corporate willingness to invest in AI remains high, but many companies are still at the pilot stage; Japanese firms are also constrained by legacy systems, low tolerance for failure, and strong dependence on SIers. Therefore, leading SIers that can help clients move from pilots to scaled deployment and implement AI automation earlier in their own development processes are expected to create new efficiency and margin divergence.

Core views

The core view is that AI is becoming a foundational layer spanning requirement definition, design, implementation, testing, and operations, rather than just point tools. Because outsourcing cost as a share of revenue in Japan's information services industry has risen from about 20% in FY1991 to a weighted average of 33.5% in FY2024, and above the same-year labor cost rate of 27.3%, AI automation makes it feasible for primary contractors to insource part of outsourced work. If efficiency benefits are not fully passed on to clients and firms can shift to value-based pricing, there is substantial margin expansion potential. However, the report also emphasizes that the industry view remains In-Line because pricing pressure, upfront AI investment, and organizational inertia will determine the variation in earnings realization across companies.

Analysis framework

The report combines industry cost-structure analysis, AI deployment maturity analysis, and company case comparisons. It starts with the bottlenecks to AI deployment in global and Japanese enterprises to explain why the next stage of competition is moving from pilot to scaled deployment. It then analyzes outsourcing-cost ratios and multi-layer subcontracting in Japan's information services industry to establish the economic basis for AI insourcing. Finally, it compares four major SIers—Fujitsu, NEC, NRI, and NTT DATA—in terms of AI development platforms, implementation goals, productivity cases, and profit-model transformation paths.

Methodology notes

• Industry cost structureOutsourcing rate and OPM sensitivity analysis

  Links the outsourced-cost-to-revenue ratio with operating-profit-margin improvement scope to assess the potential impact of AI insourcing on margins.

  The report notes that outsourcing costs are about 31% to 40% of revenue for primary contractors, and if AI-driven development reduces dependence on subcontractors, the current OPM range of 10% to 20% has roughly 5 to 10 percentage points of improvement potential.

• Technology adoption curveFrom pilots to scaled deployment

  Corporate AI competitive differentiation is shifting from whether firms adopt AI to whether they can execute organization-wide rollout of AI at scale with process and data foundations.

  The report cites that 88% of organizations frequently use AI, about two-thirds remain in pilot mode, only 1% have reached the mature stage, and only 19% can validate ROI, indicating that scaling deployment is still the main bottleneck.

• Business model transformationFrom man-month pricing to value pricing

  Whether efficiency gains convert into profit depends on whether providers can move away from time-and-materials billing and charge based on customer value.

  The report stresses that after AI boosts development productivity, if providers still bill by man-month, gains may be absorbed through client price cuts; therefore, value pricing, base fee plus usage fee, and similar models are key to margin realization.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Fujitsu (6702.T)

  Core case company

  Strengths

  Has the AI-Driven Software Development Platform centered on Takane, covering requirement definition to integration testing through multiple AI agents, and demonstrates about 100x productivity gains.

  Weaknesses

  Needs to convert technology efficiency into service value pricing, and must handle platform expansion, quality assurance, and client adoption challenges.

  Comparison

  Among listed companies, Fujitsu's case most clearly demonstrates end-to-end automation and transformation from man-month billing to value pricing.

  Risks

  Client price compression, upfront AI-platform investment, cross-industry scalability difficulty, and organizational role-transition risk if it falls short.

• NEC (6701.T)

  Core case company

  Strengths

  Proposes becoming an AI Native Company, pursues around 30 AI eXperience (AX) scenarios around BluStellar, and combines the cotomi Japanese-language LLM with over 100 AI Platform Service capabilities.

  Weaknesses

  Margin improvement depends on BluStellar revenue growth, commercialization of consulting-led scenarios, and realization of SG&A improvements.

  Comparison

  NEC places greater emphasis on enterprise AI service portfolios, scenario-driven solutions, and group-level synergy, not only development-process automation.

  Risks

  Speed of AI-scenario commercialization, execution of partner ecosystems, client budget constraints, and conflicts of interest in investment-bank disclosures.

• Nomura Research Institute (4307.T)

  Core case company

  Strengths

  Uses the AI-driven development model as a core production innovation and plans to extend financial-service experience to broader system integration businesses.

  Weaknesses

  The roadmap spans FY2025 to FY2030, and actual profit contribution requires a long period of validation.

  Comparison

  NRI's strength lies in financial industry knowledge and specialized LLM direction, with a more phased organization-level rollout path.

  Risks

  Pace of AI-agent full adoption, financial clients' demands for reliability and governance, and whether productivity gains can translate into profits.

• NTT DATA

  Core case company

  Strengths

  In 2025 practice, achieved about 70% reduction in web application development man-hours and about 3x productivity gains, and launched the LITRON Builder agent-based AI development platform.

  Weaknesses

  Long-term targets depend on continued increases in project coverage and platform adaptation in governance, security, and client environments.

  Comparison

  NTT DATA's roadmap is highly quantified, with clear phased productivity targets for FY2025, FY2027, and FY2030.

  Risks

  Speed of scaling generative-AI project deployment, attainment of LITRON business sales targets, client governance requirements, and platform competition.

Key data

• Outsourcing costs as a share of revenue31% to 40%The report believes outsourcing-cost levels at primary IT service providers are large enough to support 5 to 10 percentage points of potential OPM improvement.
• FY2024 information-services outsourcing rate33.5%Weighted-average basis, above the 27.3% labor-cost rate in the same year.
• Fujitsu productivity case4 hours to complete system modifications previously requiring about 3 person-months or roughly 480 hoursFY2024 Japan medical fee revision PoC showed roughly 100x productivity improvement.
• NTT DATA long-term targetReduce end-to-end development workflow by 70% by FY2030FY2025 plan to apply generative AI to 500 projects and a FY2027 target of 50% AI project share.
• NRI roadmapAI-driven development adoption of 100% by FY2028 and fully autonomous AI-agent deployment rate of 100% by FY2030The report states that in selected financial ASP cases, detailed design-to-unit-test-stage productivity improved by 10x to 30x.
• Japan enterprise scaled AI deployment rate8.9%The report says Japan has the lowest enterprise-scale deployment rate among five countries.

Impact & implications

The investment implication is that the impact of the AI theme on Japanese IT service providers is not only incremental AI project revenue, but also a reshaping of their own delivery process, cost structure, and pricing power. Leading SIers with proprietary LLMs, AI-agent platforms, quality-control mechanisms, industry know-how, and client-specific scenarios may benefit from outsourcing insourcing and value-based pricing. However, if clients demand sharing efficiency gains, or if organizations cannot reconfigure delivery processes, margin expansion may be lower than the space suggested by technology demonstrations.

Risks

• Efficiency gains may be absorbed by client price reductions, making margin improvement weaker than expected.
• Front-loaded investments in AI platforms, proprietary LLMs, quality control, and talent training may suppress near-term profits.
• Organizational inertia and difficulty in converting engineer roles could block full deployment of AI-native development.
• Data foundations, process redesign, AI talent, and governance reliability remain key bottlenecks to enterprise-scale AI deployment.
• Japanese legacy systems and a low-failure-tolerance culture may slow AI rollout speed on the client side.

What to watch

• Whether top SIers' outsourcing rates begin to decline and whether that decline translates into OPM expansion.
• Coverage rate, defect rate, and delivery-cycle improvement of Fujitsu, NEC, NRI, and NTT DATA AI development platforms in real-world projects.
• Whether pricing models move from man-month billing to value pricing, base-fee-plus-usage-fee, or other outcome-based models.
• Whether company-level AI deployment goals from FY2027 to FY2030 are executed on schedule.
• Whether clients demand sharing of efficiency gains and whether industry competition compresses AI service prices.