Your company just approved a £3 million AI investment. The board expects it on the balance sheet. The auditor says it should be expensed. The CEO wants to know why the income statement just took a £3 million hit in a quarter when AI was supposed to improve margins. Welcome to the CFO's AI reporting challenge.
AI investments cut across nearly every category in the chart of accounts — capital expenditure, operating expense, research, development, licensing, training, and infrastructure — creating a classification puzzle that current accounting standards were not designed to solve. This guide provides practical answers.
AI spending falls into multiple accounting categories, and the correct treatment depends on the nature of each component — not on the project label.
| AI spending component | Typical treatment | Accounting standard | Notes |
|---|---|---|---|
| AI research and exploration | Expense as incurred | IAS 38 / ASC 730 | Cannot be capitalised |
| AI development (post-feasibility) | Potentially capitalise | IAS 38 (six criteria) | See capitalisation guide |
| Third-party AI licences | Expense or capitalise | IFRS 16 / IAS 38 | Depends on licence structure |
| Cloud compute for AI training | Expense as incurred | IAS 1 | Typically operating expense |
| GPU/hardware purchases | Capitalise | IAS 16 | Tangible asset, depreciate |
| Training data acquisition | Expense or capitalise | IAS 38 | Depends on data asset recognition |
| Employee training for AI tools | Expense as incurred | IAS 38.69 | Cannot be capitalised |
| Integration and customisation | Expense or capitalise | IFRIC / SaaS guidance | Depends on who controls the asset |
There is no single "AI cost" category. AI investment must be decomposed into its components, and each component treated according to the applicable standard. The classification decision has material impact on reported earnings, balance sheet strength, and financial ratios that investors and lenders use.
Under IAS 38, development costs can be capitalised when six criteria are simultaneously met: technical feasibility, intention to complete, ability to use or sell, probable economic benefits, adequate resources, and reliable cost measurement. Most AI spending falls in the research phase and must be expensed.
The narrow window for capitalisation is the development phase — after a model has been validated as technically feasible and economically beneficial, but before it enters production. This window is typically small relative to total AI spending.
Practical advice: Do not fight to capitalise AI costs that genuinely belong in the research phase. Aggressive capitalisation invites auditor pushback and creates impairment risk. Instead, ensure the board understands why AI investment hits the income statement and how to evaluate AI returns on a cash-flow basis rather than an earnings basis.
Many AI capabilities are delivered via SaaS subscriptions — OpenAI, Anthropic, Google AI, industry-specific AI tools. The IFRS Interpretations Committee has clarified that SaaS configuration and customisation costs should generally be expensed unless the customer controls the underlying software.
For most SaaS AI tools, the customer does not control the underlying model. Therefore:
AI depends on data, and data can be enormously valuable. But current accounting standards provide no clear framework for recognising data assets on the balance sheet. The SNA 2025 revision recognised data as productive capital at the macroeconomic level, but this has not yet translated to corporate accounting standards.
A retail company has 10 years of transaction-level customer data worth millions in AI training value. Under current standards, this data does not appear on the balance sheet because it was generated internally through normal business operations and there is no reliable measurement basis. The economic value is real; the accounting recognition is zero. CFOs should disclose data asset value in management commentary even when balance sheet recognition is not possible.
Disclosure requirements for AI investments are evolving. Currently, IFRS and US GAAP require disclosure of:
The EU AI Act introduces additional disclosure obligations from August 2026 for high-risk AI systems, including transparency reports and conformity assessments.
CFOs need a management reporting framework that captures what IFRS cannot. The AI investment dashboard should track:
Break down AI spending into research, development, licensing, infrastructure, training, and data preparation. This transparency helps the board understand where AI money is going and why most appears as expense rather than asset.
Track cost savings, revenue improvements, and efficiency gains attributable to AI using the four-dimension ROI framework. Present alongside the investment figures so the board can evaluate returns regardless of accounting treatment.
Even when AI investments cannot be capitalised under IFRS, track the intangible assets being built: proprietary models, training datasets, AI-enhanced processes, and organisational AI capability. This gives a more complete picture of value creation.
The gap between accounting treatment and economic reality is not a problem to solve — it is a communication challenge to manage. The best CFOs educate their boards about why AI investment looks different on financial statements than traditional capital expenditure, and provide supplementary metrics that capture AI value creation beyond what IFRS shows.
Capitalised AI assets must be tested for impairment when indicators suggest the asset's carrying amount may exceed its recoverable amount. For AI assets, impairment triggers include:
Given the rapid pace of AI technology change, impairment testing for capitalised AI assets should be conducted at least annually — more frequently than the typical cycle for traditional intangible assets.
The Opagio Growth Platform helps CFOs track AI intangible asset value alongside financial reporting metrics, providing the supplementary data boards need to evaluate AI investment decisions.
AI investments create a persistent gap between economic value creation and financial statement presentation. Most AI spending will — correctly — appear as expense, not asset. The CFO's role is not to fight the accounting treatment but to ensure the board has the supplementary metrics and context to evaluate AI investment on its merits. Decompose spending by component, track returns across all four dimensions, and communicate the intangible value being built even when IFRS cannot recognise it.
David Stroll is Co-Founder and Chief Scientist at Opagio. His research bridges the gap between economic value creation and financial reporting, with particular focus on intangible capital measurement. Learn more about the Opagio team.
IAS 38 sets six criteria for capitalising development costs, but applying them to AI projects is fraught with ambiguity. This article maps each criterion to AI development realities and provides practical guidance for CFOs and accountants navigating the recognition question.
Read more →
Under current accounting standards, most AI development costs are expensed immediately — treated as period costs rather than capital investments. Yet new FASB guidance (ASU 2025-06) and SNA 2025 recognised data as productive capital. This creates a widening gap between economic reality and financial reporting. Here is how to navigate it.
Read more →Get the latest insights on intangible asset growth and productivity delivered to your inbox.
Book a free consultation to see how the Opagio Growth Platform can help your business.