Can AI-developed software be capitalised under IAS 38?
Short Answer
Yes, AI-developed software can be capitalised under IAS 38 once it meets the six development-phase recognition criteria, including demonstrable technical feasibility and the ability to measure costs reliably.
Full Explanation
IAS 38 distinguishes between research and development phases. During the research phase — exploring AI approaches, testing architectures, and evaluating feasibility — all costs must be expensed. Once a project enters the development phase, costs can be capitalised if all six criteria are met: (1) technical feasibility of completing the asset, (2) intention to complete and use or sell it, (3) ability to use or sell it, (4) probable future economic benefits, (5) availability of adequate resources, and (6) ability to reliably measure development expenditure. For AI projects, the critical challenge is determining when technical feasibility is established. Unlike traditional software where a working prototype is a clear milestone, AI models may produce useful outputs during training while still being in the research phase. Auditors typically look for evidence such as successful model validation on representative data, achievement of target accuracy thresholds, and a clear path to production deployment. Under US GAAP (ASC 350-40), internal-use software development costs follow similar capitalisation rules during the application development stage; ASU 2025-06 (issued September 2025) replaces those stages from 2028 with a committed-funding and probable-completion test, and is not AI-specific.
Related Glossary Terms
Related Questions
AI amplifies Total Factor Productivity by enabling firms to extract more output from the same inputs, with the productivity gains increasingly attribu...
AI assets are typically valued using the Cost Approach (replacement cost of training data and model development), the Income Approach (MPEEM or With-a...
AI creates intangible value by generating proprietary models, training data, algorithms, and automated workflows that improve efficiency, reduce costs...
Further Reading
AI and Competitive Moats: Valuing Proprietary AI Systems
Not all AI creates competitive advantage. This article examines the four types of AI moats — data, model, integration, and network — and provides a valuation framework for assessing the defensibility and durability of proprietary AI systems.
Read more →Want to see these concepts in action?
Take the free intangible asset assessment to see how these concepts apply to your business.