According to recently unsealed court documents, Meta expects its generative AI products to pull in as much as $3 billion in 2025 alone, and, by 2035, possibly rake in $1.4 trillion.
These figures result from a legal case accusing the tech giant of building its AI empire on pirated content.
The court filings don’t fully spell out what Meta defines as “generative AI products,” but we already know the company is investing heavily in tools like its Llama models and the Meta AI assistant.
It’s building APIs for developers, forming partnerships with major cloud platforms, and reportedly planning to monetise features like ads and premium subscriptions.
In 2024, Meta poured over $900 million into its generative AI groups. That number could top $1 billion this year, and that’s not even counting the infrastructure costs.
Meta has already forecast capital expenditures of $64 to $72 billion in 2025, mostly to support its growing network of data centres needed to power these AI systems.
But behind the massive financial drive lies a major ethical and legal question.
Meta allegedly considered paying for data to train its models, including books. One internal discussion from 2023 referenced a potential $200 million budget, with half of that reserved for book licences. That deal never happened.
Instead, according to the lawsuit, Meta turned to pirated material from sites like LibGen, a shadowy online repository known for hosting millions of illegal ebook copies.
The documents point to internal discussions that reveal some employees were uneasy with this approach. Their concerns were reportedly overruled. Meta, for its part, isn’t denying the use of copyrighted content, but insists it’s protected.
In a statement sent to TechCrunch, the company said: “Meta has developed transformational [open] AI models that are powering incredible innovation, productivity, and creativity for individuals and companies. Fair use of copyrighted materials is vital to this. We disagree with [the authors’] assertions, and the full record tells a different story. We will continue to vigorously defend ourselves and to protect the development of generative AI for the benefit of all.”
Meta’s CEO, Mark Zuckerberg, previously claimed the company wasn’t looking to profit from access to its AI models. That doesn’t quite align with the recent revelations. Behind the scenes, Meta has entered revenue-sharing deals with heavyweights like AWS, Google Cloud, Databricks, and Nvidia — all hosting its Llama models and feeding the ecosystem.
This lawsuit might be one of many. The AI gold rush has turned books, code, and images into training fodder. But should tech giants be allowed to mine creative work without permission or pay?