AI Token Costs and MD&A Disclosure

Key Points

• The shift from flat-fee AI subscriptions to usage-based, token pricing is creating a new category of operating cost that may be volatile, difficult to predict, and material to public companies’ financial results. 
• Public companies may need to disclose material AI token costs in their MD&As, including in discussions about period-over-period changes in expenses, known trends and Liquidity and Capital Resources. 
• Where management internally tracks AI usage metrics such as token consumption rates or per-employee AI spending, companies should evaluate whether those metrics constitute key performance indicators requiring disclosure and period-over-period analysis. 

Overview

The fast-changing economics of artificial intelligence (AI) use by public companies will require public companies’ disclosure obligations to change with them. AI model providers, including Anthropic, OpenAI, Microsoft, and Salesforce, are moving away from flat-fee, unlimited-access subscription models toward usage-based pricing measured in tokens. A “token” is the basic unit of AI computing. The result is a new category of operating cost that may be volatile, difficult to predict, and already producing material financial surprises for many public companies across a host of industries.

For public companies and their legal, finance, and accounting teams, the implications extend well beyond the information technology department. The shift to token-based AI pricing may create direct disclosure obligations under the Management’s Discussion and Analysis (MD&A) requirements of Item 303 of Regulation S-K (Item 303) promulgated under the Securities Exchange Act of 1934, as amended, as they apply to public companies’ periodic reports on Forms 10-Q and 10-K, as well as long-form registration statements on Form S-1.

AI Token Costs Rise

Recent reporting makes clear that AI token costs are no longer a theoretical concern. Major financial media publications report that companies are now exceeding their annual AI budgets in as few as three months, with token usage exploding as employees and autonomous “agentic” AI systems consume computing resources at unprecedented rates. A KPMG survey cited recently by The Wall Street Journal found that only 26% of companies report having a comprehensive view of their AI costs, while 22% report having no visibility at all, or visibility only after receiving a bill. KPMG is also reported to have indicated it is working with clients whose token usage has increased sixfold this year.

“Tokenmaxxing,” which is the practice of employees using as much AI computing as possible to appear AI-forward, has added further unpredictability to company AI budgets, with workers at some companies reportedly burning hundreds of thousands of dollars per month in tokens, sometimes using expensive premium-tier AI models to answer simple questions. For this reason, many companies are now instead beginning to take steps to monitor, ration, or even curtail employee AI usage in response to spiraling costs.

MD&A Impacts

The MD&A provisions of Item 303 require public companies to provide investors with a management’s eye view of the company’s financial condition and results of operations. The objective of MD&A is to provide material information relevant to an assessment of the financial condition and results of operations of such company, including an evaluation of the amounts and certainty of cash flows from operations and from outside sources. This includes descriptions and amounts of matters that have had a material impact on reported operations, as well as matters that are reasonably likely to have a material impact on future operations. Companies are directed to focus specifically on material events and uncertainties known to management that are reasonably likely to cause reported financial information to not necessarily be indicative of future operating results or of future financial condition. In short, MD&A is meant to be a narrative description of changes in the company’s financial statements from period to period.

Results of Operations — Increase in Expenses

The increased expenses incurred with AI token costs may have a material impact on a company’s income statement. Different companies may account for AI token costs differently. For example, some companies may account for token costs in costs of revenue, while others may account for them as general and administrative (G&A) costs. For other companies, AI token costs might be accounted for research and development (R&D) expenses. For example, AI has started to significantly affect biopharmaceutical and biotechnology companies by rapidly transforming the drug development process, enhancing and speeding target identification, molecular design, clinical trials optimization, and regulatory processes — these companies are likely to record AI token costs as R&D expenses.

In any case, the increased costs are likely to impact many companies’ income statements, and management will be expected to discuss the reasons for these changes from period to period in quantitative and qualitative terms.

The MD&A rules also require public companies to describe any known trends or uncertainties that have had or are reasonably likely to have a material favorable or unfavorable impact on net sales, revenues, or income from continuing operations. This also requires disclosure of known events that are reasonably likely to cause a material change in the relationship between costs and revenues. This could include known or reasonably likely future increases in costs of materials or services, even if those changes have not yet fully materialized.

The “reasonably likely” threshold requires a two-part assessment: (1) whether the trend, demand, event, or uncertainty is reasonably likely to occur; and (2) if it were to occur, whether it would be reasonably likely to have a material effect on future results. If management concludes that the trend, demand, event, or uncertainty is not reasonably likely to occur, no disclosure is required. If, however, it cannot make such a determination, disclosure would be required unless management determines that a material effect is not reasonably likely to occur. For many public companies, the shift from flat-fee AI subscriptions to usage-based AI token pricing, if combined with accelerating growth in AI adoption across operations, may satisfy both prongs of this standard. Where AI costs have already impacted operations (as may already be the case for many companies), companies must address those impacts in the period-over-period discussion of results. Where the full impact is still unfolding, the MD&A requirement to disclose known trends and uncertainties will compel prospective disclosure of the anticipated effects.

Liquidity and Capital Resources — Less Available Cash

Public companies must also identify any known trends, demands, commitments, events, or uncertainties that are reasonably likely to result in the company’s liquidity increasing or decreasing in any material way. Companies that have made, or are contemplating making, significant AI investments, or that face significant open-ended AI token consumption by their workforce, should assess whether those commitments warrant discussion in the Liquidity and Capital Resources section of their MD&A, as these costs could result in materially less cash on hand as of the end of each relevant balance sheet date.

Key Performance Indicators and Metrics

The SEC’s interpretive guidance has long emphasized that MD&A is not merely a restatement of financial statement data; it must include analysis of the underlying causes and implications of material changes. Where management is internally tracking AI usage metrics (such as AI token consumption rates, per-employee AI spending, or AI return-on-investment metrics), companies should evaluate whether those metrics constitute key performance indicators (KPIs) that are used to manage the business and that investors would find material. If so, disclosure and period-over-period analysis of those metrics may be required.

Practical Implications for Form 10-K and Form 10-Q Filers

Against this backdrop, public companies should consider taking the following steps as they prepare their next quarterly and annual reports:

1. Assess materiality of AI costs now. Finance and legal teams should work with IT and operations to understand the company’s current AI spending, the trajectory of AI token consumption, and whether those costs have had or are reasonably likely to have a material impact on results of operations. Companies that have already seen unexpected cost spikes or budget overruns should treat this as a disclosure-relevant event. 
2. Evaluate the cost structure shift. The move from subscription-based to usage-based AI pricing is already significantly altering the cost structure of AI adoption and use. Even if costs to date have been modest, companies should assess whether the foreseeable growth in AI usage on a metered-pricing basis is reasonably likely to produce material future cost increases requiring current or prospective disclosure. 
3. Review internal AI governance and monitoring practices. A company’s ability to prepare meaningful and accurate disclosure about AI costs depends on having internal visibility into those costs. As noted above, according to the widely reported KPMG survey, only 26% of companies currently report comprehensive visibility into AI spending. If these costs are or become material, then it is imperative that companies put adequate disclosure controls and procedures in place to measure, understand and prepare disclosure regarding AI expenses. 
4. Coordinate across legal, finance, and IT. MD&A disclosure about AI costs is likely not a question that can be answered by any single department. Legal, finance, accounting, and technology teams must collaborate to ensure that material AI cost developments are identified, assessed, and reflected in periodic filings on a timely basis, both as part of effective disclosure controls and procedures and to ensure that investors are informed of material events and trends affecting the financial results and financial condition of the company. 

Conclusion

The era of unlimited, low-cost AI experimentation is giving way to a more disciplined environment in which AI token consumption is becoming a significant and often unpredictable operating cost. For public companies, this shift carries direct implications for MD&A disclosure in Forms 10-Q and 10-K periodic reports, as companies must discuss material changes in operating results from period to period, including changes in expenses, and companies must disclose known trends and uncertainties that are reasonably likely to materially affect results of operations or the relationship between costs and revenues. To the extent that the shift to usage-based AI pricing leads to simultaneously larger and more unpredictable material expenses, disclosure under either, or both, of these frameworks may be needed. Public companies should actively monitor their AI token spending and, should these costs become material on an aggregate basis or in relation to any particular expense category, be prepared to craft appropriately tailored disclosures that satisfy their disclosure obligations and to manage investor expectations.