Meta has revealed highly ambitious revenue forecasts for its generative artificial intelligence (AI) ventures, projecting that earnings could reach as much as $1.4 trillion by 2035. In the nearer term, the company expects its AI products to generate between $2 billion and $3 billion in 2025. These figures were disclosed in court documents recently unsealed amid a lawsuit filed by authors accusing Meta of using their copyrighted works without authorisation to train its AI models.
Although the legal filings do not provide a clear definition of what Meta considers a “generative AI product,” publicly available information indicates several key components underpin the company’s AI revenue strategy.
Meta has developed the Llama family of large language models, which it has open-sourced and licensed to third-party companies under revenue-sharing agreements. Additionally, Meta has launched an API that allows businesses and developers to customise and evaluate Llama models, creating new commercial opportunities. CEO Mark Zuckerberg has also mentioned that Meta AI, the company’s AI assistant, may soon incorporate advertising and offer subscription plans with premium features, further diversifying revenue streams.
The scale of Meta’s investment in AI is substantial. In 2024, the budget allocated to its generative AI product groups exceeded $900 million, with expectations that it will surpass $1 billion this year.
These figures do not include the considerable capital expenditure required to build and maintain the infrastructure necessary for AI development, such as data centres. Meta plans to invest between $60 billion and $80 billion in 2025, primarily to expand its data centre capacity, which is essential for training and deploying large-scale AI models.
The court documents also reveal Meta’s approach to acquiring training data. Initially, the company considered spending over $200 million to license datasets for its Llama models, with approximately $100 million earmarked specifically for books. However, the lawsuit alleges that Meta ultimately abandoned licensing negotiations and instead used unauthorised sources, including pirated ebooks, to compile its training data. This decision reportedly involved Mark Zuckerberg, who evaluated the legal risks under the fair use doctrine.
“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,” a Meta spokesperson told TechCrunch in response to the allegations.
The enormous revenue projections show the commercial scale of Meta’s AI ambitions and the legal complexities surrounding copyright and fair use. Legal analysts suggest that such significant profit expectations underscore the highly commercial nature of Meta’s activities and the potential market harm to content creators. This context is vital when considering whether Meta’s unauthorised use of copyrighted works can be justified under fair use, especially given the level of investment and anticipated financial returns.
Meta’s forecast is supported by extensive investment in AI research and infrastructure, innovative monetisation models ranging from open-source projects to consumer AI assistants, and ongoing legal disputes over data sourcing practices.