Tech debt statistics 2026: the numbers that matter
Key statistics with named sources, methodology notes, and per-developer dollar conversions. The condensed version. For full deep dives by industry and team size, see technicaldebtcost.com/statistics.
Headline Strata
$1.52T
Annual US tech debt cost
CISQ 2022
of $2.41T total poor-quality cost
33%
Engineering time on debt
Stripe 2018
Harris Poll, 1,000 devs across 5 countries
20-40%
Tech-estate value at risk
McKinsey 2020-23
of the technology estate's value
21-40%
IT spending on debt
Deloitte 2026
Global Technology Leadership Study
61bn days
Global tech-debt repair time
CAST 2025
Coding in the Red; 10B+ LOC across 17 countries
15-25%
Annual compound rate (modeled)
Modeling convention
interest-rate analogy (Cunningham, McConnell)
Reference Card 02 / Per-developer
Annual cost per engineer at 30% drag
Median industry drag is 30%. These are the figures you cite when somebody asks “how much per head?”
Reference Card 03 / By team size
Annual debt cost by team size
$130K fully-loaded engineer at 30% drag (industry midpoint). Multiply by team size, watch the number grow.
Source Layer / The studies
Where these numbers come from
CISQ Cost of Poor Software Quality
2022Finding: $2.41T total US cost of poor software quality; the accumulated technical debt principal is ~$1.52T.
Method: Aggregated macroeconomic data, public industry research, expert estimation. Macro-level.
Stripe Developer Coefficient
2018Finding: Developers spend 17.3 of 41 hours a week on maintenance (42%), including 13.5 hours on technical debt (33%).
Method: Harris Poll survey, 1,000 developers across 5 countries. Dated but widely cited.
McKinsey: Tech Debt, Reclaiming Tech Equity
2020-23Finding: Tech debt is 20-40% of the technology estate's value; 10-20% of the new-product budget is diverted to servicing it.
Method: McKinsey Digital CIO survey + client benchmarks (2020, reaffirmed 2023).
CAST: Coding in the Red
2025Finding: Global technical debt totals roughly 61 billion days of repair time; the analysis spans more than 10 billion lines of code.
Method: Static analysis of production code across 17 countries representing 51% of world GDP. Code-level forensic data.
Modeling convention (interest-rate analogy)
1992-Finding: Tech-debt cost models commonly assume 15-25% annual compounding on untouched debt; a healthy ratio sits under ~10% of dev cost.
Method: Not a single study: the debt-as-interest metaphor from Ward Cunningham's 1992 WyCash report, extended by Steve McConnell and others. The calculator uses this band for accrual projections.
Deloitte Global Technology Leadership
2026Finding: 21-40% of IT spending on tech debt; older organisations skew higher.
Method: Global CIO survey, 660+ technology leaders worldwide (Dec 2025-Feb 2026). Budget-allocation lens.
Want the deep research? technicaldebtcost.com/statistics has the full deep dive. For by-industry and by-team-size breakdowns, see technicaldebtcost.com/by-industry and technicaldebtcost.com/by-team-size.