A newly unredacted court filing has revealed that Meta engages in revenue-sharing agreements with companies hosting its Llama AI models. This disclosure appears to contrast with previous statements made by Meta CEO Mark Zuckerberg, who had asserted that “selling access” to Llama models was not part of Meta’s business model.
The court filing, submitted in the ongoing copyright lawsuit Kadrey v. Meta, where the company is accused of training its Llama models using vast amounts of pirated e-books, highlights that Meta “shares a percentage of the revenue” generated by Llama model hosts. However, it does not specify which hosts are involved in these agreements.
The Llama host partnerships of Meta include companies such as AWS, Nvidia, Databricks, Groq, Dell, Azure, Google Cloud, and Snowflake, which the company has publicly identified. Developers opt to use hosting solutions because these platforms provide extra functionality in addition to simplified deployment despite their ability to work independently with Llama models.
This revelation aligns with comments Zuckerberg made during an earnings call last April, where he hinted at potential licensing arrangements. He stated, “If you’re someone like Microsoft or Amazon or Google and you’re going to basically be reselling these services, that’s something that we think we should get some portion of the revenue for. So those are the deals that we intend to be making, and we’ve started doing that a little bit.”
Zuckerberg has also emphasized that Meta benefits from Llama’s open distribution primarily through AI research advancements, which enhance the company’s products. He reiterated this stance during Meta’s Q3 2024 earnings call, saying, “I think it’s good business for us to do this in an open way. [I]t makes our products better rather than if we were just on an island building a model that no one was kind of standardizing around in the industry.”
According to the lawsuit, Meta violated copyright law in two ways: by training Llama with stolen content and leaking these works through unlawful torrenting techniques. Plaintiffs maintain Meta exploited dishonest torrenting techniques to obtain e-books, leading in the accidental transfer of those books via peer-to-peer networks.
Meta intends to increase its capital expenditures by 100% in 2025 by investments ranging from $60 billion to $80 billion, according to corporate predictions. The extra cash will mostly be used to establish more data centers and improve the company’s AI development skills. Meta intends to launch a subscription model for its AI assistant, Meta AI, in order to give enhanced functionality while covering growing AI costs.
