Meta's AI growth driven by old ML, not new GenAI... yet
Meta’s recent financial growth, particularly in its advertising segment, is primarily being driven by advancements in its established machine learning (ML) models rather than its substantial investments in cutting-edge generative AI (genAI). While the company is pouring billions into developing advanced AI, its Chief Financial Officer, Susan Li, stated that genAI is not expected to be a significant revenue contributor this year or next.
Instead, the company’s conventional machine learning systems, which power its content recommendation algorithms, are currently fueling its profitability. These systems are designed to connect users with relevant content, friends, and posts, thereby boosting engagement across Meta’s platforms. During the second quarter, improvements in these AI-powered recommendation models led to a notable approximately five percent increase in ad conversions on Instagram and a three percent gain on Facebook. CEO Mark Zuckerberg also highlighted that enhancements to these systems resulted in a five to six percent increase in time spent on Facebook and Instagram.
Crucially, these recommender models are central to Meta’s revenue generation. They efficiently sift through vast numbers of potential advertisements—tens or hundreds of thousands—to identify those most likely to resonate with individual users. Recognizing their importance, Meta has developed custom accelerators to ensure these processes operate with maximum speed and efficiency.
Looking ahead, both Li and Zuckerberg anticipate that generative AI will eventually play a more substantial role in Meta’s advertising business. The company is already leveraging genAI to assist advertisers in creating promotional materials. Li noted that nearly two million advertisers are now utilizing Meta’s video generation, image animation, and video expansion features, with strong results also observed from its text generation tools.
Furthermore, Meta has begun integrating large language models (LLMs) into the recommender systems for Threads, its competitor to X. Li reported that LLMs are now contributing significantly to the “ranking-related time spent gains” on Threads. Internally, Zuckerberg revealed that teams have started using Llama 4, Meta’s open-weights AI model, to build autonomous AI agents aimed at enhancing the Facebook algorithm for improved quality and engagement. While work continues on Llama 4.1, 4.2, and future iterations, a concrete roadmap has not yet been provided. Zuckerberg acknowledged that these advanced AI applications are currently operating at a low volume and are not major contributors to recent earnings, but he expressed optimism about their future trajectory.
Meta’s long-term strategy involves substantial investment in infrastructure and talent to support the development of these new generative AI models. The company is actively constructing a series of large-scale AI compute clusters, including Prometheus, a gigawatt-scale facility slated for 2026, and Hyperion, which Zuckerberg described as eventually reaching five gigawatts of capacity. Simultaneously, Meta is investing heavily in top-tier talent, offering eight-figure-plus salaries to build its “AI superintelligence team,” with the ambitious goal of developing systems that not only match but exceed human intelligence. Li indicated that the compensation packages for these strategic hires will be the second-largest driver of expense growth in 2026.
Despite these significant expenditures on future AI capabilities, Meta’s current financial health remains robust. In the second quarter, the company reported a 36 percent year-over-year increase in profits, reaching $18.3 billion, on revenues of $47.5 billion. This strong performance, largely underpinned by the effectiveness of its existing machine learning systems, provides the financial foundation for Meta’s ambitious push into the future of AI. Li affirmed the company’s belief that “this is the time for us to really make investments in the future of AI as I think it will open up new opportunities for us in addition to strengthening our core business.”