When should AI development costs be capitalised under IAS 38?

Under IAS 38, AI development costs can be capitalised only if all six recognition criteria in IAS 38.57 are met: (1) the project is in the development phase (not research), (2) technical feasibility can be demonstrated, (3) the company intends to complete and use or sell the asset, (4) probable future economic benefits are demonstrable, (5) adequate resources are available, and (6) expenditure can be reliably measured. Once all criteria are met, capitalise all subsequent development costs as an intangible asset.

What is the difference between IAS 38 and ASU 2025-06 for AI capitalisation?

IAS 38 (IFRS) requires six recognition criteria to be met before capitalisation. ASU 2025-06 (US GAAP, effective 2026) simplifies this for internal-use software: capitalise costs incurred during the application development stage (between preliminary project stage and post-implementation); capitalise costs include design, purchasing, installation, and testing. ASU 2025-06 removes some of the subjective IAS 38 criteria, making capitalisation more straightforward for US reporters under GAAP.

What happens to capitalised AI development costs after the asset is complete?

Capitalised AI development costs are recorded as an intangible asset on the balance sheet and amortised over the asset's useful life (typically 3–7 years for AI systems, depending on technology obsolescence and competitive advantage duration). The company must disclose in financial statements the gross amount, accumulated amortisation, and remaining useful life, plus any impairment indicators (model degradation, data obsolescence). Intangible assets with indefinite useful lives are tested for impairment annually.

Accounting Guidance

Should You Capitalise Your AI Investment?

Companies are spending billions on AI development but most are expensing everything—hiding enormous intangible asset value on the balance sheet. This interactive decision tree guides CFOs, auditors, and finance leaders through the IAS 38 and ASU 2025-06 recognition criteria to determine when AI costs should be capitalised.

The Accounting Challenge: The Capitalisation Gap

IAS 38 Intangible Assets and ASU 2025-06 (US GAAP Software Capitalisation) permit companies to capitalise AI development costs when specific recognition criteria are met. Yet fewer than 10% of companies capitalise AI investments—the majority expense everything to the P&L, treating AI as operating cost rather than capital asset.

This creates a systemic understatement of intangible asset value. A company that invests £100 million in proprietary AI over three years, expensing everything, understates its balance sheet intangible assets by £100m+ and overstates operating expenses by the same amount. For investors, auditors, and acquirers, the number becomes invisible.

Conversely, misapplication of capitalisation rules—capitalising costs that fail to meet the recognition criteria—creates audit risk, restatement exposure, and potential regulatory challenge. The SEC has increasingly focused on the classification boundary between research (always expensed) and development (potentially capitalisable).

★ Key Takeaway

The capitalisation decision hinges on six specific recognition criteria defined in IAS 38.57 or the application development stage boundary in ASU 2025-06. This decision tree walks you through each criterion to determine the correct accounting treatment with precision and audit defensibility.

IAS 38 vs ASU 2025-06: The Key Differences

Both standards permit capitalisation of software development costs, including AI, but they differ significantly in structure and subjectivity:

IAS 38 (IFRS)

Six recognition criteria must all be met: development phase (not research), technical feasibility, intent to complete, probable future benefits, adequate resources, reliable measurement. Requires significant judgement and documentation.

ASU 2025-06 (US GAAP)

Capitalise during application development stage: between preliminary project stage (planning, vendor selection) and post-implementation stage. Application development costs include design, purchasing, installation, testing. Clearer stage boundaries, less subjective.

6 recognition criteria under IAS 38.57

£176B Big Tech AI depreciation exposure (Burry estimate)

<10% of AI investments appearing on balance sheets

Interactive AI Capitalisation Decision Tree

Your path: No decisions yet

Question 1 of 7

14%

Is the AI development project in the development phase (not research)?

Research is exploratory work to discover new knowledge or techniques. Development begins when you move to practical application and building a specific AI solution. Once the research phase ends and development commences, the path to capitalisation opens.

Aspect	IAS 38 (IFRS)	ASU 2025-06 (US GAAP)
Capitalisation Trigger	All 6 criteria in IAS 38.57 must be met	Application development stage (post-preliminary, pre-post-implementation)
Research vs Development	Must distinguish; research always expensed (IAS 38.54)	Preliminary project stage treated as research; costs expensed
Technical Feasibility	Must be demonstrated before capitalisation	Assumed at application development stage start
Future Economic Benefits	Must be proven probable and demonstrable	Inherent in application development assumption
Amortisation Period	3–7 years typical for AI (based on useful life assessment)	3–7 years typical for software/AI (based on technology obsolescence)
Documentation Burden	High—must document all 6 criteria with rigour for audit defence	Moderate—stage boundary documentation; fewer subjective criteria

Related Resources

AI Capitalisation Guide

Comprehensive guide to IAS 38 and ASU 2025-06, with worked examples, audit checklist, and disclosure templates for your financial statements.

Read the Guide →

AI ROI Framework

Measure what AI investments actually create with a step-by-step framework and worked examples for boards.

Learn the Framework →

Value Assessment for AI Assets

Frameworks for valuing capitalised AI intangible assets using relief-from-royalty, excess earnings, and other methods for M&A and financial reporting.

Explore Methods →

Subscribe to our newsletter

Get the latest insights on intangible asset growth and productivity delivered to your inbox.