Mark Zuckerberg's Meta may be spending hundreds of millions of dollars a month on Anthropic'sartificial intelligence tools, even as the two companies compete to develop increasingly powerful models.
Despite cutting "some" of its spending on Anthropic over the past few months, Meta is "still spending hundreds of millions of dollars a month on the start-up’s tools," The New York Times reports, citing people familiar with the matter. In fact, the company has projected it could spend billions more.
"At one point this year, Meta internally projected that it could spend as much as $10 billion annually on Anthropic’s A.I. models," according to the NYT report. "That would have formed a major chunk of Anthropic’s yearly revenue, which the start-up estimated in July would pass $65 billion."
The potential spending would also underscore the extent to which Meta is relying on outside AI companies even as it pours billions of dollars into developing its own models.
That would also be a significant chunk of Meta's known AI tool spend. The company committed over $14 billion in Scale AI and reportedly agreed to a five-year deal for Scale’s AI and data service at least $450 million per year.
Meta is both competing with Anthropic on models while potentially supplying Anthropic with the computing infrastructure needed to compete. Of note, Reuters reported in July that Meta and Anthropic were in talks for a potential $10 billion compute lease deal over two years.
For its part, Meta's Superintelligence Labs is trying to build a vertically integrated AI ecosystem. It is developing the Muse family, with Muse Spark 1.2 currently its most advanced model.
Meta has forecasted up to $145 billion in capex on AI - however, that's primarily based on infrastructure such as data centers, servers and networking equipment.