The news
Meta Platforms (META) has used the federal research tax credit to cut its taxes by billions of dollars by describing its AI data centers as experiments, The New York Times reported on September 30, 2026. The Meta research tax credit claims reduced its taxes by $2 billion in 2024 and $3.9 billion in 2025, after $700 million in 2023, according to the Times as summarized by The Next Web.
For tax purposes, Meta classifies its AI data centers as pilot models and treats chips bought from suppliers including Nvidia (NVDA) as supplies used in experiments, according to the report. The research and experimentation tax credit, created in 1981, gives companies a credit for qualifying research spending. Gizmodo said the Times based its account on four sources.
Meta's auditor, EY, signed off on the approach and has since pitched it to other companies, the Times reported. The Next Web cited the Times as saying the Joint Committee on Taxation projected the credit would cost the federal government $32.1 billion in 2025, with Meta accounting for more than one-tenth of that.
The approach carries risk. According to the Times, as reported by The Next Web, Meta holds a reserve of $18.74 billion for uncertain tax positions, up 45% in two years, and lists uncertainty over research tax credits as the main factor. Gizmodo, summarizing the Times, said Meta's own accountants see the classification as a gray area that the Internal Revenue Service could overturn.
Meta spokesman Andy Stone told the Times that, like other companies investing at this scale, Meta uses tax incentives Congress established decades ago. The Next Web also reported that Meta is separately fighting the IRS over $355 million in savings from treating Mark Zuckerberg's 2013 stock options as a research expense.
The numbers
- Research credit tax savings, 2025
- $3.9 billion (NYT)
- Research credit tax savings, 2024
- $2 billion (NYT)
- Research credit tax savings, 2023
- $700 million (NYT)
- Projected federal cost of the credit, 2025
- $32.1 billion (Joint Committee on Taxation, via NYT)
- Meta reserve for uncertain tax positions
- $18.74 billion (via NYT)
Why CEOs should care
For CFOs and tax leaders, the report signals that aggressive treatment of AI capital spending will draw scrutiny. Companies claiming research credits on infrastructure should review how their positions are documented, what share of reserves covers them, and how a reversal would hit cash. If an approach is being marketed by advisers as a repeatable product, as the Times reported EY did, regulators tend to notice.
For investors and boards, tax treatment is part of the economics of the AI buildout. A credit worth billions a year lowers the effective cost of data centers. If the IRS or Congress narrows it, the after-tax cost of AI infrastructure rises for Meta and any company copying the approach. Audit committees should ask whether reported tax rates depend on positions the company itself describes as uncertain.
For executives in other industries, the story is also about public perception. Data centers already face local pushback over power, water and tax breaks. A national report that taxpayers are subsidizing them through a research credit adds political risk to projects and incentive negotiations.
The bigger picture
Big tech companies are spending at record levels on AI infrastructure and looking for ways to ease the cost, from sale-leasebacks to joint ventures. James Shannon, a sponsor of the 1981 law that created the credit, told the Times that its use has gone well beyond what anyone imagined, according to The Next Web. Whether the IRS challenges the classification will shape how far other companies push similar claims.
What’s next
Watch for IRS action on Meta's research credit claims, any response from lawmakers on the tax-writing committees, changes to Meta's tax reserve in its next quarterly filing, and whether other AI spenders disclose similar credits.
What “Fact-checked” means
Fact-checking means testing a story’s facts against the evidence before it is published. This story went through at least two separate checks before this version was published.
- What we checked
- Its names, figures, dates, job titles, quotes and who said what were checked against the story’s sources, including its main source where it could be opened. The headline was checked for accuracy and overstatement.
- How
- A first check reviewed the whole story. If it passed, a second, skeptical check went back to the sources to look for mistakes in the most important facts. If a check flagged the story, it was edited to fix the problems found, and a separate re-check then reviewed the whole story again.
- Who
- The checks are made by our newsroom, as steps kept separate from the writing, under rules set by our editor, Hussein Mukhtar. A story the checks still flag is held for the editor, who decides whether it is fixed, published or dropped.
- If something is wrong
- “Fact-checked” does not mean error-free. If a material error is found after publication, we correct the story and add a note saying what changed. Report an error
Companies in this story
