The surge in artificial intelligence start-ups has generated remarkable valuations, but concerns are growing among investors about the reliability of reported revenue figures. Increasingly, companies in the AI sector are being scrutinised for employing what critics describe as “creative accounting” techniques to inflate sales and present a rosier financial outlook than might be justified.

Common practices under question include counting projected or unsigned deals—referred to as “contracted annualised recurring revenue” (CARR) or “pre-contracted revenue”—as actual recurring revenue. Other issues involve incorporating one-off set-up fees into recurring revenue, overlooking discounts, recognising government grants as sales, misreporting customer churn, and treating short pilot agreements as binding contracts despite customers retaining the option to withdraw.

Highlighting these concerns, Ada Ventures, a UK-based investor, recently cautioned its portfolio companies against misleading revenue representations, urging transparency and accuracy in financial disclosures. Similarly, Y Combinator, a notable start-up accelerator, has advised founders to be “truthful and precise” regarding revenue reporting.

Insiders suggest these accounting irregularities are often less about deliberate fraud and more about founders feeling intense pressure to hit specific annual recurring revenue (ARR) benchmarks ahead of fundraising rounds. A handful of AI start-ups are indeed experiencing rapid growth, setting high expectations that may prompt others to present their figures more optimistically to remain competitive. Investors themselves might inadvertently exacerbate this by encouraging companies to showcase strength in favourable market conditions.

Even established AI firms are embroiled in disputes over accounting methods. For instance, OpenAI has accused rival Anthropic of overstating its revenue by billions, citing differences in reporting gross customer revenue versus net revenue after partner payouts. Anthropic maintains it complies with generally accepted accounting principles (GAAP) and provides consistent, accurate figures.

These tensions echo previous tech industry episodes, such as the controversy surrounding Mike Lynch’s software firm Autonomy, which was found to have inflated revenues amid a market boom. Veterans from that era recall such practices were widespread, underscoring a culture of competitive pressure and ambiguous accounting standards that continue to challenge the sector. Experts note that accounting in the technology space often involves complex judgments and interpretations, particularly for emerging AI businesses with less stable sales patterns and nascent business models.

Traditionally, software companies command high valuations due to steady, long-term contracts and predictable recurring revenue. However, AI firms frequently operate with usage-based pricing models, where revenue and margins fluctuate, complicating valuation efforts. Combined with accounting ambiguities, this creates significant difficulties for investors seeking to assess the true financial health of AI start-ups.

The current environment reflects heightened scrutiny reminiscent of previous tech bubbles, where regulatory bodies imposed substantial penalties on audit firms for oversight failures, such as Deloitte’s £15 million fine related to Autonomy. As concerns mount about a potential AI market correction, the focus on accounting practices underscores the challenge of distinguishing between genuinely high-performing companies and those benefiting from inflated valuations.