Meta has reportedly classified massive AI data centers as ‘pilot models’ to claim US research tax credits, according to a New York Times investigation.
The strategy cut Meta’s tax bill by about $700 million in 2023, before the AI build-out strategy began, rising to $2 billion in 2024 and a reported $3 billion-plus more recently, per filings cited by the NYT.
Tax experts, including Andre Shevchuck of BPM and Shawn Marchant of Tanner, question whether buying commercially available AI chips qualifies as genuine ‘research,’ raising potential IRS scrutiny risk.
A tax credit designed in 1981 to reward genuine scientific experimentation is now, reportedly, helping Mark Zuckerberg’s company shave billions off its bill for building some of the world’s most expensive AI infrastructure. According to a New York Times investigation based on Meta’s securities filings and interviews with four people familiar with operations, the research tax credit reduced Meta’s tax bill by roughly $700 million in 2023 — before its AI data-center strategy ramped up — before jumping to $2 billion in 2024 and reportedly surpassing $3 billion more recently. These figures come from company filings as reported by the NYT, not an independent audit, and the IRS has not publicly commented.
Pilot Projects, or Just Pricey Data Centers?
The central question, per the NYT’s sourcing, is whether buying commercially available computer chips for large-scale AI infrastructure counts as a qualifying research expense, which requires resolving genuine technical uncertainty through experimentation. Some Meta finance-department employees reportedly questioned internally whether the ‘pilot model’ framing would hold up. Tax expert Andre Shevchuck called characterizing the data centers as experimental ‘kind of wild and out there,’ while Shawn Marchant said he would be skeptical of applying the credit across all chips used in Meta’s data centers.
What This Means for Your Portfolio and Wallet
For investors in Meta or AI-infrastructure-heavy peers, this is a reminder that today’s tax-advantaged earnings could face retroactive IRS clawbacks tomorrow — a liability risk not always priced into valuations. Broader tax-policy scrutiny of AI capex could also ripple into how other hyperscalers report earnings.
Strategic Positioning & Defense Ideas
Educationally, this underscores the value of not over-concentrating portfolios in single mega-cap tech names and instead spreading exposure across diversified tech or broad-market funds, while keeping some cash on hand for headline-driven volatility. Disclaimer: This analysis is for educational and informational purposes only and should not be construed as financial or investment advice.
What to Watch Next
Watch for any formal IRS response or audit disclosure from Meta, as well as how other AI-heavy firms address similar tax treatment in upcoming filings. Full reporting is available via LiveMint, citing The New York Times.
