Meta Classifies AI Data Centers as Pilot Models for Research Tax Credits

Meta classifies its AI data centers as “pilot models” to claim federal research tax credits, the New York Times disclosed in a September 30 investigation. The approach treats chips bought from suppliers including Nvidia as supplies used in experiments. Meta’s 2025 annual filing records $3.912 billion in total research and development tax credits for the year ended December 31, 2025. It does not break out credits from data centers or Nvidia chips.

The figure covers the company’s research credit as a whole, so it cannot establish how much the reported data-center strategy saved or what share of the total it accounts for. The account of Meta’s internal classification rests on interviews with people familiar with its operations; public filings do not disclose the pilot-model designation. Whether those purchases qualify could affect how much of the claimed benefit Meta could retain if the IRS were to challenge the strategy.

Chip classification

Meta reportedly began using the approach in 2024, labeling chips bound for AI facilities differently for tax purposes from those sent to standard data centers. The company sought advice from several law firms before proceeding, amid questions within its finance department about whether the treatment would withstand scrutiny.

The relevant criterion is whether the claimed supplies are used to resolve technical uncertainty in qualified research. Calling a facility a pilot model does not by itself establish that its equipment meets that criterion.

The experimental-use argument

Tax lawyer Jeffrey Moeller, of Ivins, Phillips and Barker, said commercially available, proven products could qualify as research supplies if needed to resolve technical uncertainty in a project. He declined to discuss Meta specifically.

Meta is testing server-rack layouts and ways to connect thousands of chips for AI training. The Nvidia chips themselves are proven technology. In its March 12, 2024 engineering post, Meta described two clusters of 24,576 graphics processors each, connected by different types of networks. It used the two designs to assess how well each network could support AI training and how it would perform in larger clusters.

The IRS research-credit audit guide says supplies generally must be nondepreciable tangible property used directly in qualified research. Property subject to depreciation falls outside that definition. The available account does not establish Meta’s exact equipment treatment or an IRS determination on this strategy.

Shawn Marchant, who runs Tanner’s credit and incentives practice, said he would be “skeptical” of claiming the credit for every chip in every data center. Andre Shevchuck, a BPM partner specializing in the credit, called characterizing AI data centers as experimental “kind of wild and out there.”

Meta’s response

Meta spokesman Andy Stone defended the company’s use of incentives Congress enacted to encourage domestic investment. He said Meta had invested $200 billion in research and development over the five years preceding the investigation, including $57 billion in 2025.

A separate figure in Meta’s June 30, 2026 quarterly filing covers a much broader set of tax positions: $18.74 billion in gross unrecognized tax benefits. Stone described unrecognized tax benefits as “simply a mandated accounting measure of uncertainty,” covering many unresolved issues.

The filing says those uncertainties primarily concern research credits and transfer pricing with foreign subsidiaries, including intellectual-property licensing and services. It does not isolate exposure from the data-center strategy. The gross amount is an accounting measure for uncertain tax positions, not cash taxes owed or an estimate of losses from this approach. Recording a credit also does not establish that the IRS has approved it.

Meta declined to explain what made its AI facilities experimental or why their chips and computing equipment qualified.

Mistral Raises Record €3 Billion Led by Samsung to Fund Data Centers

Mistral announced a €3 billion, about $3.5 billion Series D on September 8 to expand its data center buildout and computing-capacity business beyond model development. Samsung Electronics led the roun

The Implicator

Abbott and Shapiro Restrict Data Centers After Courting $60 Billion

In November, Texas Gov. Greg Abbott sat beside Google chief executive Sundar Pichai at a data center in Midlothian as Google announced plans to invest $40 billion in data centers across the state. Abb

The Implicator

Qualcomm Targets Over $15 Billion in Data Center Revenue, Names Meta CPU Customer

Qualcomm told investors Wednesday that it expects more than $15 billion in data center revenue by fiscal 2029. In a same-day announcement, Meta agreed to use Qualcomm's Dragonfly C1000 CPUs in servers

The Implicator

Marcus Schuler

San Francisco

Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: editor@implicator.ai