Blog→Technical Debt
Technical Debt9 min read

The 4 Types of Technical Debt (And Why They Compound Differently)

Code debt, architecture debt, infrastructure debt, and process debt each have different interest rates.

By Richard Ewing·
Share:

Not All Debt Is Equal

Code Debt compounds slowly - messy code slows development incrementally. Architecture Debt compounds fast - wrong patterns multiply across the system. Infrastructure Debt can be catastrophic - outdated infrastructure fails suddenly. Process Debt is invisible - bad processes accumulate friction silently.

Each type requires different remediation strategies and has different ROI timelines. Code debt: 3-6 month payback. Architecture debt: 6-18 months. Infrastructure debt: immediate to 3 months. Process debt: 1-3 months.

Measure your debt →

Like this analysis?

Get the weekly engineering economics briefing - one email, every Monday.

Subscribe Free →

More in Technical Debt

The Negative-Carry Code Crisis

Because AI makes generating code free, we are seeing a massive inflation in the volume of code pushed to repositories. But negative-carry code carries hidden debt.

8 min read

The Negative-Carry Code Crisis

Because AI makes generating code free, we are seeing a massive inflation in the volume of code pushed to repositories. But negative-carry code carries hidden debt.

8 min read

The Innovation Tax Audit: Is Your R&D Actually Just OpEx?

Are your software assets still generating value, or are they just burning resources? A guide to auditing your portfolio and killing zombie features.

10 min

Related Canonical Concepts

AI Volatility Tax

The compounding gross margin penalty incurred when variable LLM inference query costs scale faster than subscription revenue, shifting server hosting into variable Cost of Goods Sold (COGS).

Read Concept →

The Negative-Carry Code Crisis

The systemic financial risk created when high-velocity AI code generation produces massive volumes of un-audited, low-trust technical debt that inflates ongoing maintenance OpEx beyond marginal value creation.

Read Concept →

Vibe Coding Debt

The engineering debt accumulated when developers accept AI-generated code based on superficial execution ("vibes") without understanding underlying architectural assumptions or edge cases.

Read Concept →

Model Collapse

A degenerative process where AI models experience severe performance degradation after being iteratively trained on synthetic data generated by other models.

Read Concept →

Inference Economics

The financial discipline of managing, projecting, and optimizing the per query token costs associated with running large language models in production.

Read Concept →

Technical Insolvency

The critical threshold where the operational cost of maintaining a codebase and resolving technical debt exceeds the engineering capacity available for new feature development.

Read Concept →

The AI Margin Squeeze

The systemic erosion of traditional SaaS gross margins caused by the integration of generative AI features, as variable compute and API costs scale linearly or exponentially with user engagement, fundamentally altering software unit economics.

Read Concept →

Semantic Caching

The architectural pattern of storing and reusing similar LLM query results using vector embeddings to bypass redundant frontier model API execution and eliminate variable COGS.

Read Concept →

Zombie Code & The Sunset Protocol

Zombie Code refers to deprecated or unused features that continue to run in production, consuming maintenance budget, compute resources, and engineering focus. The Sunset Protocol is the structured mechanism for financial remediation through systematic deletion.

Read Concept →

SLM Repatriation

The strategic shift of migrating high-volume inference tasks from commercial Frontier APIs (OpenAI, Anthropic) to local Small Language Models (SLMs) to achieve financial breakeven on variable COGS.

Read Concept →

DORA Metrics Financial Translation

The analytical process of converting standard engineering performance metrics (Deployment Frequency, Lead Time, MTTR, Change Failure Rate) into direct financial liabilities and capitalization impacts on the P&L statement.

Read Concept →

Canonical Frameworks

The Software Phase Transition

The Software Phase Transition models the structural breakdown of traditional product management as the marginal cost of writing software approaches zero. In the pre-AI era, developer bandwidth was scarce and expensive. Organizations operated in the Solid state: managing 2-week sprints, grooming backlogs, and writing exhaustive PRDs to ration engineering hours. As tooling improved, organizations transitioned into the Liquid state of adaptive teams with fluid prototyping. With generative AI and autonomous agent pipelines, code generation costs collapse toward zero, propelling organizations into the Gas state. In the Gas state, developer capacity is no longer the rate-limiting constraint. Unbounded code generation creates exponential organizational complexity, coordination tax, and margin collapse. This forces a fundamental leadership evolution: product leaders must transition from managing feature velocity to becoming Product Economists who govern capital, system architecture efficiency, and uncertainty.

Read Definition →

Cost of Predictivity

The Cost of Predictivity measures the variable cost of AI accuracy. Unlike traditional software with near-zero marginal costs, AI features have significant variable costs that scale with both usage AND accuracy requirements. As AI correctness increases, cost scales exponentially - not linearly. This is the fundamental economic challenge of AI products. Traditional software follows a simple cost model: high fixed development cost, near-zero marginal cost per user. Build the feature once, serve it to millions for pennies. AI products break this model entirely. Every AI query costs compute. Every inference requires GPU cycles. Every improvement in accuracy requires either more sophisticated prompts (more tokens = more cost), retrieval-augmented generation (vector DB queries + embedding generation), or fine-tuned models (massive training costs amortized over queries). The cost structure looks more like a manufacturing business than a software business. The exponential curve is the killer. Moving from 80% accuracy to 90% accuracy might cost 2x. Moving from 90% to 95% might cost 5x. Moving from 95% to 99% often costs 10-20x. This is because the easy cases are solved by the base model, and each additional percentage point of accuracy requires increasingly sophisticated (and expensive) techniques to handle edge cases. This creates what Richard Ewing calls the AI Margin Collapse Point: the usage volume at which AI feature costs exceed the revenue they generate. Many AI features that work beautifully in prototype (low volume, don't need high accuracy) become economically devastating in production (high volume, users demand high accuracy). The AI Unit Economics Benchmark (AUEB) calculator at richardewing.io/tools/aueb helps companies calculate their Cost of Predictivity and identify their specific margin collapse point before it hits their P&L.

Read Definition →
📊

Richard Ewing

The AI Economist - Quantifying engineering economics for technology leaders, PE firms, and boards.

Book Advisory →Curriculum →Free Tools →

Keep exploring

blog

Engineering Culture Economics

comparison

PDI vs. CodeClimate

comparison

Svelte vs. Nuxt

← Back to Blog
⚡

Want to apply this to your organization?

Run a free diagnostic first. If the numbers concern you, book a session to build a remediation plan.

Run Free Diagnostic (Free)View Advisory Options

Richard Ewing: AI Economist & Capital Auditor