Meta is leveraging a federal research tax credit to reduce its US tax liability by billions of dollars, a strategy that involves classifying its massive AI data centers as "pilot models" and its Nvidia chips as experimental materials.
What Happened
According to reporting by The New York Times, Meta saved $3.9 billion in taxes in 2025 through this credit, a significant increase from $2 billion in 2024 and $700 million in 2023. This figure makes Meta the biggest beneficiary of this specific credit among all publicly traded companies. The classification allows the company to treat infrastructure typically viewed as operational capital as research and development.
This tax strategy contrasts with public statements from Meta CEO Mark Zuckerberg. In July 2025, Zuckerberg announced plans to "invest hundreds of billions of dollars into compute to build superintelligence," citing multi-gigawatt clusters like Prometheus, which is already partly online, and Hyperion, which is expected to scale to 5 GW. Additionally, in January 2025, Zuckerberg stated that these data centers would "drive our core products and business," a description that implies established utility rather than experimental status.
Why It Matters
The tax credit in question dates back to a 1981 law. James Shannon, the congressman who originally introduced the legislation, told The New York Times that it was intended for "people power, knowledge, information," and described Meta's application of it as having "gone way, way beyond what anybody could have imagined." Despite the controversial classification, Meta defends the practice by pointing to $200 billion spent on R&D over the past five years.
However, the strategy carries legal risks acknowledged by the company itself. In SEC filings, Meta warns that these savings could be challenged, and its reserves for uncertain tax positions have jumped 45 percent to $18.74 billion. Even if the IRS were to claw back the money, Meta likely remains financially advantaged because the capital was deployed in the interim, potentially boosting its stock price. The accounting firm EY, which approved the strategy and helped establish the scheme, is reportedly pitching the same approach to other companies seeking to offset AI chip purchases.
The Bottom Line
Meta's use of the federal research tax credit highlights the tension between aggressive AI infrastructure scaling and traditional tax definitions of R&D. While the company has secured billions in savings, it faces increasing regulatory scrutiny and rising internal reserves for potential tax disputes, signaling that this financial tactic may become a broader industry trend for major AI developers.