{{item.title}}
{{item.text}}
{{item.text}}
Companies across a wide range of industries are racing to transform their products, services, and operations with artificial intelligence (“AI”). The competitive urgency is driving significant investment and, as a result, compute costs have emerged as one of the largest and fastest-growing components of AI spending. As organizations accelerate their adoption of AI, finance and accounting leaders face a growing challenge: determining how AI-related investments should be planned, measured, and accounted for. Rising compute costs, expanding use of third-party AI platforms, increasing internal development efforts, and evolving accounting guidance require organizations to evaluate both the economics and accounting treatment of AI investments. Regarding the economics, organizations are becoming laser focused on their return on investment (“ROI”).
AI compute costs are becoming a structural cost of doing business rather than a discretionary technology investment. A thoughtful framework can help companies align AI strategy, capital allocation, and financial reporting objectives while maintaining compliance with U.S. GAAP.
Leading companies are increasingly discussing AI infrastructure investments in their quarterly earnings communications, and this trend extends well beyond the technology sector. Even some of the largest non-technology companies have discussed AI in their earnings releases, signaling that AI spending is now a strategic investor relations topic, not just a technology issue. Organizations that wait until AI compute costs become material to margins will risk responding to questions their peers have already anticipated through stronger governance, cost tracking, and financial planning. As AI workloads scale from hundreds to thousands of use cases, the aggregate compute bill could become a meaningful margin variable.
While AI investment is accelerating at a rapid pace, the related US GAAP accounting guidance was developed in the 1980s and 1990s. Recent market commentary has cited estimates of roughly $4 trillion to $8 trillion in AI-related capital investment over the next five years, underscoring the scale of spending now under review by finance teams. With this magnitude of spend, a common discussion topic is whether these costs are required to be capitalized and amortized over time or if they are instead required to be immediately expensed.
Before determining the accounting treatment of AI-related expenditures, organizations should consider performing an accounting assessment of their AI strategy and investment portfolio to assess its future state accounting and financial reporting requirements. Accounting fit gap analyses are intended to help expose any issues that the company may have with accounting-based problems (e.g. accounting for its AI investments) that could lead to a potential lack of data available and/or a lack of system calculation capabilities (resulting in inefficient, error-prone manual defaults).
Depending on the nature of the activity, AI investments may fall under the scope of different accounting models. Companies generally begin by asking whether the software is for internal use, delivered through a hosted or SaaS environment, or sold or licensed externally. That scoping decision matters because software within ASC 350-40 (internal use) follows a different capitalization model than software within ASC 985-20 (software to be sold).
For many AI initiatives, the answer is not obvious. For example, a company may license a third-party model, use purchased data, build application-layer functionality, and deliver the result to a customer through an existing SaaS platform. Below are the potential existing accounting models to evaluate:
Given the variety of deployment models available in today's AI ecosystem, similar business objectives may result in significantly different accounting outcomes.
As compute costs scale, accounting judgments also become more consequential.
Compute costs have become one of the most visible pressure points in the AI investment cycle. As companies scale model training, inference, and enterprise deployment, spend on GPUs, cloud capacity, and third-party AI infrastructure is quickly moving from a technology concern to a finance and accounting priority. A key question is whether this is a service arrangement or a capitalizable cost. Companies are investing in foundation models, training data, prompt engineering, software integration, and retraining, with finance teams applying existing accounting frameworks to fact patterns that are technical, iterative, and fast-changing. Generally, such compute costs would be expensed unless directly tied to qualifying development and the costs are traceable.
AI tokens have also become a common unit of service consumption for AI model usage that needs to be evaluated. Companies can purchase and consume AI tokens in various ways, but the key from a financial statement impact perspective is the nature of the service purchased and whether the resulting cost can be attributed directly to a qualifying software development activity.
With compute now driving much of the cost conversation, the accounting analysis should start by identifying what the spend is for and which model applies. Key judgments often turn on the arrangement’s scope, capitalization timing, whether costs are direct and incremental to a specific project, and whether future spend represents enhancement or routine maintenance.
This initial scoping decision is especially important because compute costs can arise in different forms: cloud-based access to model capacity, vendor-hosted AI platforms, internal development environments, or AI-enabled products delivered to customers. The accounting conclusion may change depending on whether the spend supports a hosted service, an internal-use software project, software to be sold or licensed, research activity, or ongoing maintenance. As compute becomes a larger share of AI spend, finance teams need to connect the economics of compute consumption to the applicable accounting model early, before costs scale and reporting judgments potentially become more challenging to support.
Taken together, these scoping considerations establish the accounting foundation for AI-related spending. The next step is to translate that foundation into the financial reporting decisions that will affect earnings, asset balances, forecasts, and the controls needed as investment scales.
Key financial reporting implications: AI accounting outcomes depend on five key decisions: how the investment is classified, when capitalization begins, which costs qualify, how subsequent spending is treated, and whether governance and reporting processes support the conclusions. These decisions affect earnings, capitalized assets, amortization, cash-flow presentation, forecasts, covenants, disclosures, and internal control over financial reporting. The executive framework below organizes each decision into three fields: Impact, Trigger, and Action.
CFO lens: Impact = where results or forecasts may change; Trigger = the judgment requiring alignment; Action = the policy, control, or governance response required before spending scales.
| Key decision | Impact | Trigger | Action |
1. Scope Initial accounting |
Sets the accounting model, expense pattern, balance-sheet classification, and disclosure requirements. | Is the investment internal-use software, customer-facing software, a hosted service, research, acquired technology, or a combination? | Require accounting review of the use case, contract structure, deployment model, and unit of account before material spending begins. |
2. Capitalization Initial accounting |
Affects near-term earnings, asset balances, future amortization, forecasts, and period-over-period comparability. | Has management committed funding, and are completion and intended use probable? | Define approval gates and retain evidence supporting funding, intended use, project milestones, the probability evaluation and the capitalization start date. |
3. Cost eligibility Initial accounting |
Influences EBITDA, operating income, capital expenditure budgets, asset values, and potentially cash-flow presentation. | Are payroll, vendors, data, compute, configuration, integration, and testing costs direct, incremental, and traceable to qualifying development? | Implement project codes, time tracking, allocation rules, and finance review to separate development from research, training, and overhead. |
4. Subsequent accounting Ongoing reporting |
Changes the timing of expense, prepaid implementation costs, enhancements, maintenance, amortization, and impairment. | Does later spending add functionality or maintain performance, and do useful-life, impairment, or abandonment indicators require reassessment? | Set enhancement-versus-maintenance criteria and perform recurring useful-life, impairment, and abandonment reviews. |
5. Governance Control readiness |
Affects forecast reliability, covenants, audit evidence, disclosures, stakeholder communications, and ICFR. | Can systems trace costs and judgments consistently, and are policies, ownership, and review controls operating before spending scales? | Assign accountable owners and establish policies, controls, recurring reviews, and reporting metrics across finance, technology, legal, procurement, and tax. |
Refer to our In-depth for a deeper discussion on the accounting and financial reporting considerations.
Even after scoping is resolved, companies still need to determine when capitalization begins. Under the current internal-use software guidance, capitalization starts only after the preliminary project stage is complete, management has committed funding, and completion is probable. In AI projects, that assessment may be more complex because development uncertainty may persist longer than it does in traditional software projects.
Because AI projects often involve experimentation and rapid tuning, companies may need clearer milestones and stronger evidence that the project is sufficiently defined and probable of completion. Once capitalization begins, another key judgment is the amortization period, considering the nature and pace of AI development.
After deployment, the key question is whether retraining or tuning creates new functionality or maintains existing performance. That distinction often determines whether follow-on spend is capitalized or expensed.
Where retraining expands a model’s capabilities in a meaningful way, the costs may be more consistent with an enhancement. Minor tuning to improve accuracy or maintain performance is often more akin to maintenance.
Because AI development is iterative and technical, the analysis may be difficult to support without robust policies, procedures, and documentation. Finance teams may need evidence of intended use, project authorization, probable completion, and the basis for classifying activities as development, research, maintenance, or training. As noted above, the guidance in this area is evolving with regards to the probability of completion determination. After adoption of ASU 2025-06, capitalization begins when management authorizes and commits to funding the project and it is probable the project will be completed and the software will be used to perform its intended function, including resolution of significant development uncertainty which is more prescriptive than existing guidance. Refer to our insight on the new guidance for additional information.
Companies should also consider whether their systems can accumulate costs at a sufficiently granular level to support the intended accounting treatment and related controls. As these costs become more material, the related systems may also come into scope for internal control over financial reporting (ICFR) under Sarbanes-Oxley. This may result in changes to the systems and processes used to capture, classify, capitalize, and report those costs.
The rapid adoption of AI has transformed AI spending from a technology budget item into a strategic capital allocation decision. Organizations are investing significant resources to:
At the same time, rising compute costs and growing consumption of cloud-based AI services are creating new pressures on operating budgets and financial forecasts. Both public and private companies are reassessing their AI investment strategies to determine how best to deploy capital while balancing expected returns, profitability objectives, and long-term growth plans.
The accounting treatment of AI investments can have a meaningful impact on financial results, including:
As a result, finance leaders should evaluate accounting implications early in the AI investment lifecycle rather than after implementation decisions have already been made. Organizations should ensure that accounting conclusions are aligned with annual operating plans, long-range strategic forecasts, and broader capital allocation objectives.
As AI investment and compute costs grow, finance teams may increasingly need to be prepared to discuss these topics in earnings releases, earnings calls, and other investor communications.
AI investment is a strategic consideration across industries, not solely within the technology sector. As spending on AI capabilities and supporting infrastructure grows, finance teams should be prepared to explain how these investments align with business priorities and affect operating budgets, financial forecasts, and expected returns.
Both technology and non-technology companies have discussed AI in their earnings releases, demonstrating that AI has become a strategic priority across industries, not just within the technology sector.
Companies that establish clear frameworks for tracking, reporting, and communicating AI-related costs may be better positioned to address questions regarding AI investments consistently across financial statements, earnings releases, and investor presentations.
Finance and accounting teams have a narrow window to get ahead of AI accounting. The judgments involved are complex, and the governance, documentation, and cost tracking to support them can take time to stand up. Waiting until implementation decisions have already been made is likely too late.
The first step is building a repeatable accounting framework. Updating capitalization policies, defining AI use cases, and ensuring cost tracking is aligned across cross-functional teams is critical. Companies that tackle these questions now will be in a much stronger position to apply the guidance consistently and explain their reasoning to auditors, regulators, and stakeholders.
Broader impacts for deals, tax and controls
As deal activity rebounds, AI capitalization practices will face greater scrutiny. The US Deals 2026 midyear outlook expects larger AI- and private equity-driven transactions, increasing focus on acquisition targets. Capitalization decisions affect earnings, budgets, cash flow, and covenants, while early tax involvement may identify R&D credit opportunities.
AI investments are attracting increasing attention from investors and dealmakers. As compute costs and AI-related spending continue to rise, stakeholders are scrutinizing whether those investments are generating sustainable returns and creating long-term enterprise value.
We are seeing leading companies move on several fronts at once. They are assessing AI investments and the related accounting implications, putting AI-specific controls and governance in place, upgrading project tracking systems, and bringing finance and engineering together around documentation standards. Many are also revisiting useful life assumptions and closing identified gaps in data and systems. The earlier these steps occur in the investment lifecycle, the easier it is to connect accounting conclusions to operating plans, strategic forecasts, and capital allocation decisions.
The accounting treatment of AI investments can also create opportunities to improve tax planning and compliance. As organizations implement more robust project governance, cost tracking, and documentation to support financial reporting, the same information can often be leveraged to support tax reporting, identify available incentives, and enhance forecasting. Early coordination between finance, tax, and technology teams can help organizations evaluate the broader tax implications of AI investments, including:
As AI investments grow, finance teams may need practical support translating accounting guidance into policies, controls, and documentation that fit how their organization develops and deploys technology.
PwC can help organizations assess scoping and capitalization questions, evaluate the business implications of AI-related software costs, and design approaches that support consistent application across projects and teams.
Our teams bring experience across accounting, tax, operations, people, and technology to help clients implement accounting policies, strengthen cost tracking, and prepare for stakeholder questions as AI programs scale.
We would like to thank Taylor Testa and Peter Feigenbaum for contributing to this article.
{{item.text}}
{{item.text}}