New Accounting Rules Will Force Big Tech to Break Down AI Spending

Major technology companies are preparing for new accounting disclosures that will give investors unprecedented detail about their artificial intelligence spending, according to Bloomberg Law.

Alphabet and Meta Platforms have disclosed they are evaluating guidance from the Financial Accounting Standards Board that will require them to break down expenses currently hidden within broad income statement categories. The standard, known as the disaggregation of income statement expenses (DISE), takes effect for calendar-year companies in early 2028.

Google parent Alphabet reported $18.2 billion in research and development expenses in the second quarter of this year, while Meta's R&D spending reached $21.6 billion in the same period. Under current practice, these figures lump together salaries, asset depreciation, and other costs without further detail.

Why it matters

Investors have pressed tech companies for clarity on AI capital expenditures as spending balloons. PwC projects annual data center spending could climb from roughly $800 billion in 2026 to $1.8 trillion by 2050. The new disclosures won't solve every question—companies can still aggregate AI and non-AI costs within categories—but they represent the first mandatory step toward transparency in an area where voluntary disclosure has been limited.

What companies must reveal

The FASB standard requires detailed breakouts of expenses such as employee compensation and depreciation within commonly presented categories like cost of sales and research and development. These details will appear in footnotes to financial statements filed with the Securities and Exchange Commission.

"If you are someone who really wants to narrow down on a company's expense profile related to their AI investments, you're going to have a much better chance of doing that successfully after DISE is implemented than you do today," Ohio State University assistant professor Brian Monsen told Bloomberg Law.

For hyperscalers—large cloud service providers—the disclosures could reveal data center-related depreciation within cost of revenue. Depreciation spreads the cost of assets like buildings and equipment across their useful lives, providing insight into how corporate infrastructure is being utilized.

Employee compensation disclosures may also shed light on AI hiring trends. Meta has offered packages exceeding $200 million to attract talent for its superintelligence team, Bloomberg News previously reported. However, the standard does not require companies to separate AI-specific salaries from other roles within broader categories.

Implementation challenges

Accounting advisors are urging companies to begin preparation immediately despite the 2028 deadline. "If I were giving a message to the companies who are working on this, it's 'start yesterday,'" Chris Bolash, a partner in EY's financial accounting advisory services practice, said.

The compliance process will be data-intensive, particularly for multinational companies with employee compensation information scattered across systems in multiple jurisdictions. Timothy Phelps, a partner in KPMG's department of professional practice, noted that compiling this information may prove complex.

FASB spokesperson Christine Klimek emphasized that the standard is "intended to improve transparency across public companies, including those with significant technology investments," though it is not AI-specific.

Alphabet disclosed in its most recent quarterly report that it is evaluating how to adopt the guidance. Meta declined to comment beyond its annual report reference to the accounting standard. Alphabet did not respond to a request for comment.

Details were first reported by Bloomberg Law.