Series: AI Accounting Insights: Navigating the Financial Landscape of Emerging Technologies
Impairment Considerations for AI Assets
Introduction
The rapid advancement of artificial intelligence (AI) technologies has revolutionized various industries, leading to significant investments in AI assets such as large language models (LLMs) and the extensive data assets that support them. As companies allocate substantial resources to develop and maintain these assets, it becomes crucial for financial leaders to understand the accounting implications, particularly regarding impairment considerations for AI assets. Accurate impairment assessments ensure that financial statements reflect the actual economic value of AI assets, thereby maintaining transparency and compliance with accounting standards.
In this article, part of our ongoing AI Accounting Insights series discussing accounting considerations for companies developing generative AI technologies, we discuss the impairment considerations for (1) LLM development assets and (2) data assets useful lives, providing insights into the factors influencing impairment assessments and offering guidance on how to navigate these complex challenges effectively.
AI Assets
The AI industry has witnessed exponential growth, with companies investing heavily in AI assets such as LLMs and developing generative AI applications and the extensive data required to train them. These investments represent significant capital expenditures and are often recognized as intangible assets on the balance sheet.
See our prior articles discussing some of the accounting considerations around the development of AI technologies:
Capitalized Development Costs
Developing an LLM and generative AI applications can be a resource-intensive process that involves substantial costs across various stages of development. These costs include salaries and benefits for AI researchers, data scientists and software engineers who design, code and test the models. Additionally, the training of LLMs requires significant computational power, often necessitating investments in high-performance hardware or cloud-based services. Other direct costs may include software licenses and expenses for specialized tools used in the development process.
Under ASC 350-40, Internal-Use Software, entities can capitalize certain costs incurred during the application development stage of software intended for internal use, assuming there is meaningful future economic benefit. This includes costs directly attributable to the development of the technology after the preliminary project stage is complete and management commits to funding the project. Capitalizing these costs recognizes the future economic benefits the AI technology is expected to generate over its useful life, aligning expenses with the periods in which the revenue is realized.
Capitalized Data Assets
Data is the lifeblood of LLMs, enabling them to learn and improve their language understanding and generation capabilities. Data is also important to developing and fine-tuning the generative AI applications that run off LLMs. Entities often incur significant expenses acquiring data through purchases or licensing agreements. These costs can include fees for access to proprietary datasets, expenses for data collection initiatives, and costs associated with converting non-digital data into machine-readable formats. Additionally, substantial resources are invested in data preparation activities such as cleaning, labeling, and organizing the data to ensure it is suitable for training purposes.
Under ASC 350, Intangibles—Goodwill and Other, data acquisition costs can be capitalized as intangible assets if they meet certain criteria. Specifically, if the data provides probable future economic benefits and the costs can be reliably measured, they should be capitalized and amortized over the data's useful life, unless the useful life is indefinite. This treatment reflects the enduring value of high-quality data assets, which can be utilized across multiple models and projects, contributing to economic benefit over an extended period.
Impairment Considerations
While capitalizing LLM development costs and data assets aligns the recognition of expenses with the realization of benefits, it also necessitates careful consideration of impairment. The AI industry is characterized by rapid technological advancements, intense competition and shifting market dynamics. Technological obsolescence can occur swiftly as new models with superior capabilities emerge.
These factors can lead to situations where the expected future cash flows from capitalized AI assets diminish significantly, indicating that their carrying amounts may not be recoverable. In such cases, companies must assess these assets for impairment following relevant accounting standards and recognize impairment losses if appropriate.
Under U.S. GAAP, the accounting for the impairment of long-lived assets is primarily governed by ASC 360, Property, Plant, and Equipment, which addresses the impairment and disposal of long-lived assets. Although ASC 360 is traditionally associated with tangible assets, its guidance extends to intangible assets subject to amortization.
ASC 360 requires that long-lived assets be tested for recoverability whenever events or changes in circumstances indicate that their carrying amounts may not be recoverable. In the case of AI assets, the primary drivers that may trigger an impairment assessment are as follows:
Technological Obsolescence. In the AI industry, rapid technological advancements can swiftly render existing AI assets obsolete. Entities continually acquire additional data and leverage increased computational power to develop more complex models. Innovations in AI algorithms, model architectures and training techniques can lead to significant performance improvements, making prior models less competitive. For instance, developing new neural network architectures or advancements in natural language understanding can quickly outpace existing LLMs.
Technological obsolescence is a critical impairment indicator, especially for capitalized LLM development costs. When a company's AI assets become outdated due to new technological developments, the expected future cash flows from these assets may decline. Finance and accounting leaders must monitor technological trends and assess whether their AI assets can continue to generate sufficient economic benefits. If not, an impairment test may be necessary to determine whether the asset's carrying amount exceeds its recoverable amount.
Market Demand. Shifts in market trends and customer preferences can substantially impact the utility and profitability of AI assets. The AI market is characterized by rapidly evolving consumer demands, with clients often seeking the latest and most advanced AI solutions. A shift towards newer technologies or alternative solutions can reduce the demand for products or services based on existing AI assets.
For example, if customers begin favoring AI models that offer enhanced features, better performance, or improved ethical considerations, such as greater transparency or reduced bias, the demand for older LLMs may wane. Additionally, regulatory changes or heightened data privacy and security awareness can influence market demand. A decline in market demand is an impairment indicator, prompting a review of whether the AI asset's carrying amount remains recoverable.
Competitive Landscape. The competitive landscape significantly affects the economic viability of AI assets. The emergence of new competitors and the rapid development of more advanced models by existing players can erode a company's market position. The open-source AI community, in particular, plays a pivotal role in intensifying competition. Open-source LLMs, developed collaboratively and often freely available, can rival or surpass proprietary models in performance and capabilities.
Data assets deemed to have an indefinite useful life should not be amortized until it's determined that their useful life is finite. This assessment is revisited each reporting period to ensure that the indefinite classification remains appropriate. These assets are subject to more rigorous impairment testing under ASC 350. This means they must undergo an annual impairment test — and more frequently if circumstances change — by comparing their fair value to their book value at the time of the impairment test.
As open-source models narrow the technological lead of proprietary AI assets, companies may be compelled to invest heavily in updating or replacing their models and applications to maintain a competitive edge. This scenario can shorten the useful life of existing AI assets and may lead to impairment if the assets cannot generate expected future cash flows. Finance and accounting leaders must consider the impact of open-source competition on their AI assets and assess whether impairment indicators are present.
Impairment Testing Process
When impairment indicators are identified for intangible assets being amortized, companies must perform an impairment test following ASC 360 guidance. The steps include:
- Assessing Recoverability: Estimate the future undiscounted cash flows expected from the use and eventual disposition of the AI asset. This assessment should consider factors such as projected revenue, cost savings and market conditions.
- Comparing Carrying Amount to Cash Flows: If the carrying amount of the AI asset exceeds the estimated undiscounted cash flows, the asset is not recoverable, and impairment exists.
- Measuring Impairment Loss: Determine the fair value of the AI asset, often using discounted cash flow models or market-based valuation techniques. The impairment loss is the amount by which the carrying amount exceeds the fair value.
- Recognizing Impairment Loss: Record the impairment loss in the financial statements, reducing the carrying amount of the AI asset to its fair value.
Key Considerations for Financial Leaders
- Regular Monitoring: Establish processes to routinely monitor for impairment indicators, including technological advancements, market dynamics and competitive pressures.
- Collaboration with Technical Teams: Work closely with AI engineers and data scientists to understand technological trends and the potential impact on AI assets.
- Transparent Assumptions: Document the assumptions and methodologies used in impairment testing to ensure transparency and support during audits.
- Strategic Asset Management: Consider strategies to extend the useful life of AI assets, such as updating models, investing in complementary technologies, or repositioning assets in new markets.
Conclusion
Impairment considerations for AI assets are a critical aspect of financial management in the AI industry. The rapid pace of technological innovation, shifting market demands, and intensified competition, particularly from the open-source community, present significant challenges in maintaining the economic value of AI assets. Financial leaders must remain vigilant, regularly assessing impairment indicators and conducting thorough impairment tests when necessary.
By proactively addressing impairment considerations, companies can ensure that their financial statements accurately reflect the value of their AI assets, maintain compliance with accounting standards, and make informed strategic decisions. Navigating these complexities requires a collaborative approach, integrating financial expertise with technical insights to effectively manage the risks and opportunities inherent in AI asset management.
Connect with us to learn more.
About CBIZ ARC
At CBIZ ARC, we specialize in providing top-tier technical accounting and financial consulting services to growth-oriented companies, including those leading in AI innovation. Our expert teams bring a unique combination of deep accounting knowledge and a clear understanding of emerging technologies, helping companies navigate complex financial, systems, and data management challenges. From addressing technical accounting issues and preparing for IPOs to unlocking deeper insights with AI-driven tools, we deliver customized solutions that ensure compliance, financial transparency, and performance optimization. Renowned for our commitment to seamless execution and responsive client service, CBIZ ARC provides the strategic support to drive your company's sustained growth and long-term success.