Meta saved billions in US taxes by calling AI data centers experiments

Meta has used a US federal research and development tax credit to cut its 2025 tax bill by $3.9 billion, classifying its AI data centers as "pilot models" and Nvidia chips as experimental materials, according to a New York Times report.

The savings have grown quickly: $700 million in 2023, $2 billion the following year and $3.9 billion in 2025. That makes Meta the largest beneficiary of the credit among publicly traded companies.

A pilot model, or core infrastructure?

Meta's own public statements are hard to reconcile with the label. In July 2025, Mark Zuckerberg said the company would "invest hundreds of billions of dollars into compute to build superintelligence," anchored by several multi-gigawatt clusters. The first, Prometheus, is already partly online, while Hyperion is supposed to scale to 5 GW over several years.

In January 2025, Zuckerberg said the data centers would "drive our core products and business" and announced a facility with more than 2 GW of capacity. By June 2026, Meta was openly detailing its compute infrastructure, including partnerships with Nvidia, AMD, AWS, Arm and Broadcom and its own custom MTIA chips. None of that sounds like an experiment.

An old credit, a new scale

The credit dates back to a 1981 law. James Shannon, the congressman who introduced it, told the Times it was meant for "people power, knowledge, information," and said Meta's use has "gone way, way beyond what anybody could have imagined."

Defense, reserves and the auditor

Meta defends the practice by pointing to $200 billion spent on R&D over the past five years. Even the company's own accountants treat it as legally risky: in SEC filings Meta warns the savings could be challenged, and its reserves for uncertain tax positions rose 45 percent to $18.74 billion.

Even if the IRS claws the money back, Meta likely still comes out ahead, because the capital was put to work in the meantime and helped lift its stock price, the Times reports. EY, Meta's auditor, approved the strategy, helped set up the scheme and is now pitching the same approach to other companies looking to offset their AI chip purchases.