Meta used a controversial technique to get a tax credit, according to a new report from the New York Times. If the report is true, taxpayers could have subsidized Meta’s AI data center expansion over the past two years.
Meta has spent billions of dollars on the company’s artificial intelligence turnaround efforts, a sizable investment that includes operating 28 data centers across the country. Zuckerberg said during the company’s latest earnings conference call that this investment in AI is already “paying off.”
But, according to four people with knowledge of the company’s operations who spoke to The New York Times, Meta sets a different tone at the Internal Revenue Service.
When filing taxes, Meta describes AI data centers as “pilot models” in a precarious experiment. This allows the company to take advantage of the Research and Experimentation Tax Credit, which provides discounts to businesses for supplies used in experimental efforts rather than for standard business operations.
Using this method, the New York Times reports, Meta has claimed billions of dollars in tax credits over the past two years, becoming the largest publicly traded beneficiary of the tax credit. But the Times also notes that Meta’s own accountants are reportedly concerned that the classification is in a very gray area and that the IRS could decide to reverse the savings. The uncertainties linked to the research tax credit are even listed as a risk in the financial information.
In this case, those “supplies,” for which Meta allegedly received billions in rebates, are the chips it buys for use specifically in AI data centers. Meta is buying these chips from Nvidia, with which the tech giant signed a huge multi-year chip deal earlier this year. According to Bloomberg estimates, Meta was the second-largest buyer of Nvidia products in the last fiscal year, even before committing to buy millions of additional chips in the latest deal announced in February. The New York Times claims that Meta’s auditor, EY, has since introduced this tactic to other AI companies.
Even with the so-called tax credit, Meta’s investments in AI have put pressure on the company’s finances. In the latest earnings report, executives reported that free cash flow for the quarter fell to just $784 million, down about $8 billion from the same period last year. The company is one of the hyperscalers, or large technology companies that operate AI data centers, including Amazon, Microsoft and Google. Hyperscalers are also the biggest customers of the chip giant and the world’s most valuable company, Nvidia, which sits at the center of the circular AI trading network. Many observers of the AI bubble say that if hyperscalers are not as financially healthy as expected and demand fails to catch up with investments, it could have a significant negative impact on Nvidia’s business, creating a domino effect that could bring down the entire system.
One such critic is investor Michael Burry, who said earlier this week that he expected an AI bubble burst scenario “to happen over the next year,” sooner than he had previously anticipated. The investor, who became famous for predicting the 2008 housing crash, now expects Nvidia shares to fall by next September, even though the company insists it is very confident in its financial future.
Gn bussni
