The data every decision-maker needs to build and defend a legacy modernization business case — ROI benchmarks, failure rates, AI tools, and industry breakdowns.
The average global enterprise wastes more than $370 million annually on legacy inefficiency — yet the modernization failure rate sits at 70–88%, almost always due to organizational issues, not technology. Organizations that execute well report 228–362% ROI within three years and 30–50% operational cost reductions. AI-powered tooling has fundamentally changed the cost equation, with real deployments showing 50–80% reductions in timelines. The window to modernize on favorable terms is narrowing: technical debt compounds at ~20% annually, and 92% of COBOL developers will have retired by 2030.
Most organizations dramatically underestimate what their legacy systems cost. A Deloitte Banking Survey (2024) found that actual legacy TCO runs 3.4× higher than initial estimates — a mid-sized European bank budgeting €2 million per year for core system costs discovered the true figure was €6.8 million once compliance overhead, integration friction, and innovation drag were included.
At the macro level, enterprises allocate 60–80% of IT budgets to maintaining existing systems, leaving only 20–40% for innovation. Gartner predicted that by 2025, companies would spend 40% of IT budgets on technical debt alone. The U.S. federal government remains the starkest example: approximately 80% of its $100 billion IT/cybersecurity budget goes to operating and maintaining legacy infrastructure.
| Industry | % of IT Budget on Legacy Maintenance |
|---|---|
| Healthcare | Up to 75% |
| Financial Services | 70–78% |
| U.S. Federal Government | ~80% |
| Average Enterprise | 60–80% |
The Pega/Savanta study of 500+ IT decision-makers (October 2025) quantified enterprise waste with striking granularity:
At the national level, CISQ estimates $2.41 trillion annuallyin U.S. costs from poor software quality, with accumulated technical debt principal reaching $1.52 trillion. Oliver Wyman's analysis shows global technical debt roughly doubled from 2012 to 2023, growing by approximately $6 trillion.
The application modernization services market was valued at $15–22 billion in 2024 and is growing at a 15–20% CAGR, with projections reaching $40–55 billion by 2030. Cloud migration services represent an even larger market at roughly $300 billion in 2025, projected to exceed $1 trillion by 2030.
ROI data from multiple authoritative sources paints a consistently positive picture for organizations that execute well:
| Source | Approach | Reported ROI |
|---|---|---|
| Kyndryl (2026, 500 senior leaders) | Modernize on mainframe | 288% |
| Kyndryl (2026) | Integrate with cloud | 297% |
| Kyndryl (2026) | Move off mainframe | 362% |
| Forrester TEI / Microsoft | Azure PaaS (3-year) | 228% |
| Nucleus Research | Cloud migration (per dollar) | $3.86 return |
The Forrester Total Economic Impact study commissioned by Microsoft documented $43.17 million in benefits versus $13.15 million in costs over three years. Average modernization cost dropped from $9.1 million in 2024 to $7.2 million in 2025 due to AI-powered automation.
Payback periods cluster around 6–24 months depending on scope and approach:
89% of large companies have digital/AI transformation underway, but have captured only 31% of expected revenue lift and 25% of expected cost savings. The gap between potential and realized value is substantial — and top performers dramatically outperform the average.
The cost of not modernizing extends well beyond maintenance spending into security, compliance, talent, and competitive positioning.
Outdated systems have 3× more security vulnerabilities than modern counterparts. U.S. breach costs hit $10.22M in 2025. Healthcare breaches average $9.77M; financial services $6.08M. Companies on legacy systems are 40% more likely to experience compliance failures.
Over 90% of mid-size and large enterprises report hourly downtime costs exceeding $300,000 (ITIC 2024), with 41% putting costs at $1–5M per hour. IDC reports enterprises on legacy spend 42% more on operational overhead and experience 4× more downtime.
The avg COBOL developer is 58.3 years old. Fewer than 2,000 graduated worldwide in 2024; 47% of organizations already cannot fill COBOL roles. Salaries rising 25% per year. 42% of critical business knowledge sits with 1–2 people who understand legacy systems.
Nearly 60% of AI leaders view legacy-system integration as the primary barrier to agentic AI adoption. Companies with fragmented legacy systems are 30% more likely to experience AI implementation delays.
The most effective modernization business cases are built around a three-horizon framework rather than a single ROI number projected 3–5 years out. PwC research confirms that programs with ROI horizons stretching beyond three years without interim proof points lose executive patience and get cancelled.
Any credible business case must acknowledge the J-curve. McKinsey found organizations overspend by approximately 14% annually during migration transitions due to dual-run costs. A projection showing a clean downward cost curve from day one will destroy credibility with finance audiences.
The business case should quantify three ROI categories separately:
The CapEx-to-OpEx shift from cloud modernization is strategically attractive — converting large, unpredictable capital expenditures into manageable operating expense streams.
PE sponsors increasingly view technology as the primary value creation lever. Simon-Kucher's 2025 study found 33% of deal teams now rank operational improvements as their primary equity story driver — nearly double buy-and-build strategies. Harvard Business School research (2024) confirmed that PE-backed companies significantly increase digital investments post-acquisition, with those increases correlating to stronger sales growth and higher productivity.
Companies actively investing in digital transformation are 3× more likely to achieve premium valuationsin M&A processes. The median PE hold period now sits at 6 years, giving more runway for modernization payback — but also creating pressure to demonstrate results within the investment horizon.
The emergence of AI-powered modernization tools represents the most significant shift in the modernization landscape since cloud computing. Gartner predicts that by 2027, GenAI tools will reduce legacy modernization costs by 70% by explaining legacy applications and creating replacements. Real-world results are already validating this trajectory.
IBM watsonx Code Assistant for Z uses a 20-billion-parameter LLM trained on COBOL-Java pairs and was named a Leader in the 2025–2026 IDC MarketScape for AI Coding Assistants. COBOL-to-Java automated conversion reliability has reached 70–85% in 2025, up from 40% in 2020.
With 220 billion lines of COBOL still in active use globally, processing $3+ trillion in daily commerce, and over 85% of universities having dropped COBOL from their curricula, the workforce to maintain these systems is literally disappearing. 80% of banks now plan to modernize existing COBOL code through AI-assisted refactoring — not because it's the preferred approach, but because manual alternatives will soon be unavailable. AI-enabled mainframe workloads are expected to rise from 3% in 2024 to 25% by 2026 — an 8× increase in two years.
The failure statistics are sobering. Bain & Company's 2024 study of 24,000+ transformation initiatives found 88% fail to achieve their original ambitions. BCG reports only 35% of digital transformations meet value targets. McKinsey's study with Oxford of 5,400+ IT projects found that large projects (>$15 million) run 45% over budget and deliver 56% less value than predicted, with 17% going so badly they threaten the company's existence.
McKinsey consistently identifies organizational culture as the dominant obstacle — organizations investing heavily in culture change see 5.3× higher success rates than technology-only approaches. Bain identified three critical mistakes:
Only 16% of executives feel comfortable with available tech talent, and 83% of organizations lack employees with necessary change management skills.
The most effective de-risking strategy is the Strangler Fig pattern: gradually replacing legacy functionality with modern services while maintaining full business continuity. This converts high-risk big-bang events into manageable incremental delivery, contains the blast radius of any failure, and delivers business value in weeks rather than years.
Gartner explicitly recommends continuous modernization over rip-and-replace. McKinsey's research on P&C insurers found core replacement projects routinely run 50%+ over budget.
COBOL powers 43% of all banking systems and 95% of ATM transactions. Banks spend 70–78% of IT budgets on legacy maintenance. Regulatory mandates (PSD3, DORA, FedNow) require real-time APIs that legacy cannot provide. JPMorgan Chase committed $17B in 2024 — moving 80% of apps off legacy data centers.
Only 10% of large insurance providers have modernized more than half their systems. BCG projects insurers will spend ~$17B across EMEA and North America on core IT modernization from 2024–2026.
The VA EHR modernization ballooned from $10B to $37B lifecycle cost, with only 6 of 171 medical centers deployed after ~$5B in spending. The positive case: modern systems deliver 50% fewer errors, 40% faster processing, and 45% better care outcomes.
68% of U.S. manufacturers rely on applications over 15 years old. A Midwest automotive supplier lost $10M in 2024 from a single legacy system failure during just-in-time delivery. ERP market reached $23B in 2025, growing at 8% CAGR.
The business case for legacy modernization in 2025–2026 is defined by converging pressures that make inaction increasingly untenable.
The strategic posture the data supports: modernize incrementally with AI acceleration, present the business case through a three-horizon framework with quantified cost-of-inaction scenarios, and invest at least as heavily in organizational change as in technology.
Sphere helps engineering and finance leaders build defensible modernization business cases — grounded in real benchmarks, J-curve realism, and the cost-of-inaction data boards require.
Talk to Sphere →