Meta has publicly stated that monetizing access to its Llama AI models isn’t its primary goal—but newly unsealed legal documents suggest the company still sees financial benefit from them. In a July blog post, CEO Mark Zuckerberg emphasized that “selling access” wasn’t part of the business model. Yet in court filings related to the Kadrey v. Meta copyright case, plaintiffs revealed that Meta collects a share of revenue from companies hosting and offering Llama-based services.
The lawsuit alleges Meta trained Llama using vast amounts of copyrighted e-books obtained through torrenting, and in doing so, may have redistributed those files. The documents claim Meta not only downloaded pirated works but also inadvertently shared them due to the nature of torrenting software.
Despite Zuckerberg’s comments, Llama is generating income. While it’s unclear which hosts are paying Meta, several major players—like AWS, Google Cloud, Nvidia, Snowflake, Azure, Dell, Groq, and Databricks—have been publicly named as partners in Meta’s blog posts. These companies offer infrastructure and support tools that make it easier for developers to deploy Llama models, though the models can be freely downloaded and run independently.
During an earnings call in April 2024, Zuckerberg floated the idea of monetizing Llama in other ways—through licensing, business messaging, and even ad-driven AI experiences—though he didn’t provide specific plans. He noted that if companies such as Microsoft or Amazon are reselling access to services powered by Llama, Meta should receive a portion of that value. According to him, those types of partnerships were already in motion.
The company maintains that the real advantage of Llama lies in its open model, which attracts research contributions and improvements from the wider AI community. These enhancements, Meta says, benefit its own products—like the Meta AI assistant—more than direct monetization ever could.
Zuckerberg reiterated this belief in the Q3 2024 earnings call, explaining that openness leads to better products through collaboration rather than isolation. Meta’s broader strategy, however, points to an aggressive investment in AI infrastructure. In January, the company announced plans to double its capital spending to between $60 and $80 billion in 2025, with much of that allocated to AI research, development teams, and data centers.
To help recoup some of those expenses, Meta is reportedly weighing the launch of a premium version of Meta AI, featuring added tools or exclusive functionality—signaling that while open access may be the messaging, monetization is still very much on the table.