Meta looks for an AI reset

This is a column about AI. My boyfriend works at Anthropic. See my full ethics disclosure here.

Over the weekend, Bloomberg reported that Meta is considering a large investment in data infrastructure provider Scale AI. Today, fresh reports in Bloomberg and the New York Times revealed what Meta is getting for its roughly $14 billion. In addition to acquiring 49 percent of Scale, 28-year-old Scale founder Alexandr Wang will join Meta as an executive, leading a new team of extremely expensive researchers in an effort to build superhuman AI.

The deal could be announced as soon as Wednesday.

Meta and Scale declined to comment. But I spent the past day talking to people close to the deal, and can independently confirm many of the details in those Bloomberg and Times reports. Today, let's talk about what Meta is up to, and explore whether this is the AI reset that the company has been looking for since the disappointment of Llama 4.

I.

First, the details. Founded in 2016, Scale's primary business is in labeling data for use in the training of AI systems. For example, it might hire subcontractors to identify and categorize content that violates Meta's community standards to help train machine-learning classifiers. Scale, like Meta, has been sued by former content moderators over its labor practices.

Meta is one of its largest customers, but it also provides services to Meta's rivals, including Google DeepMind and OpenAI. One person I spoke with said Scale expects those companies will stop using Scale's services, for fear of Meta using information about their usage to gain a competitive advantage. As part of the deal, Meta has agreed to dramatically increase its spending with Scale.