Meta Claims Billions in Tax Credits by Classifying AI Data Centers as Experimental

Meta is using an aggressive tax strategy to claim billions of dollars in research and experimentation tax credits by classifying its AI data centers as experimental pilot models, according to people familiar with the company’s operations. The approach allows Meta to receive taxpayer subsidies for expensive computer chips from companies like Nvidia, even though these components are commercially proven technology widely used across the industry.

The tax break, created during the Reagan administration in the 1980s, was designed to encourage American companies to take risks on uncertain technological innovations. Companies must meet a complex four-part test proving they are conducting genuine experiments rather than standard business operations. While thousands of companies benefit from this credit, they typically claim it for researcher and engineer salaries, not for physical infrastructure and equipment.

Meta began claiming the credit for data centers two years ago, and its tax savings from the program have surged dramatically. The company saved approximately $2 billion in 2024 and $3.9 billion in 2025, up from $700 million in 2023, making Meta the largest beneficiary of the research tax credit among publicly traded companies. However, the company’s own securities filings warn investors that billions in tax savings could be overturned by the IRS due to uncertainties surrounding these research tax credits.

Tax experts describe Meta’s interpretation as unusually aggressive. The IRS has previously rejected similar claims for commercially available equipment, and Meta is already engaged in a separate dispute with the IRS over a $355 million tax benefit related to CEO Mark Zuckerberg’s stock options. Despite internal concerns within Meta’s finance department about whether the strategy would withstand IRS scrutiny, the company proceeded after consulting with multiple law firms.