What is the useful life of an AI model for amortisation purposes?
Short Answer
AI models typically have useful lives of 2-5 years for amortisation, reflecting rapid technological advancement, though the underlying training data and architecture may retain value longer.
Full Explanation
Determining the useful life of an AI model is one of the more challenging judgements in intangible asset accounting. Unlike a patent with a defined legal life or a customer contract with a stated term, AI models depreciate based on technological and economic factors. Key considerations include: the pace of advancement in the specific AI domain (NLP models may obsolete faster than industrial ML models), the availability and quality of retraining data, competitive dynamics, and regulatory changes. Most auditors and valuers assign useful lives of 2-5 years for AI models, with 3 years being a common benchmark for general-purpose models. However, several components may have different useful lives: the curated training dataset may retain value for 5-7 years, the model architecture for 3-5 years, and the specific trained weights for 1-3 years. Under IAS 38, if the useful life cannot be determined, the asset is treated as having an indefinite life and tested annually for impairment rather than amortised. For AI assets, regulators generally expect a finite life to be determinable given the pace of technological change.
Related Glossary Terms
Related Questions
AI infrastructure investments — including GPU clusters, MLOps platforms, and data pipelines — are accounted for as either tangible assets (hardware), ...
AI washing — overstating AI capabilities or AI-driven value — can inflate intangible asset valuations, mislead investors, and create impairment risk w...
AI ROI is measured by comparing the total cost of AI development and deployment against quantifiable benefits including revenue uplift, cost reduction...
Further Reading
AI Cost Reduction: How to Measure Real Operational Savings
AI vendors promise dramatic cost reductions, but actual savings depend on total cost of ownership, implementation quality, and measurement rigour. This article provides a practical methodology for measuring real AI-driven operational savings — net of all hidden costs.
Read more →Want to see these concepts in action?
Take the free intangible asset assessment to see how these concepts apply to your business.