Believe it or not, Meta’s AI investments made a meaningful difference to its advertising business in Q2 — it’s just that those models aren’t the kind that’s got everyone, including the Social Network, plowing tens of billions of dollars a year into datacenters.
It’s going to be a few more years before the Zuckercorp’s rampant spending on infrastructure and AI visionaries starts to bear fruit, CFO Susan Li reiterated on Wednesday’s second quarter earnings call.
“We don’t expect that our genAI work is going to be a meaningful driver of revenue this year or next year,” she said.
For now, improvements in the more conventional machine learning models that power Meta’s recommender systems are paying the bills.
Broader deployment of the company’s new AI-powered recommendation models drove a roughly five percent increase in ad conversions across Instagram and a three percent gain across Facebook, CEO Mark Zuckerberg touted in his prepared remarks on Wednesday.
As their name suggests, these models are designed to boost engagement across the company’s various platforms by connecting users to friends, posts, and other content. And in the last quarter, Zuck boasted improvements in these systems led to a five to six percent increase in time spent doomscrolling Facebook and Instagram.
More importantly, these recommender models are how Meta makes money by sorting through tens or hundreds of thousands of potential ads to find the ones most likely to be relevant to individual users. This process is so important, in fact, that Meta has developed custom accelerators to ensure they run as quickly as possible.
Going forward, Li and Zuckerberg say genAI will eventually play a bigger role in its advertising business. The Facebook parent is already using genAI to help users create advertising collateral, for example.
“Nearly two million advertisers are now using our video generation features, image animation, and video expansion, and we’re seeing strong results from our text generation tools,” Li said.
Li also revealed that Meta has begun incorporating large language models (LLMs) into the recommender systems used by Meta’s X-competitor Threads. “LLMs are now driving a meaningful share of the ranking-related time spent gains on Threads,” she said.
Meta is toying with LLMs and other genAI tools in other areas as well. Zuck has previously teased an AI engineer designed to speed development of his Llama model herd.
On Wednesday’s call, he said teams had begun using Llama 4 to build autonomous AI agents to “improve the Facebook algorithm to increase quality and engagement.”
Speaking of the open weights models, Zuck said work on Llama 4.1, 4.2, and future models is ongoing, but stopped short of providing a concrete roadmap.
All of this is “happening in low volume right now, so, I’m not sure that result, by itself, was a major contributor to this quarter’s earnings… but I think the trajectory on this stuff is very optimistic,” Zuckerberg added.
- Zuck tries to justify AI splurge with talk of ‘superintelligence’ for all
- Europe’s AI crackdown starts this week and Big Tech isn’t happy
- New Google AI model maps world in 10-meter squares for machines to read
- Altman embraces inner Viking, raids Europe with 100K GPU supercluster in Norway
For now, Meta is focused on building out the infrastructure necessary to fuel the development of new generative AI models by investing in infrastructure and talent.
The company is currently in the process of building out a series of AI compute clusters, including a gigawatt scale facility called Prometheus set to come online in 2026 and another called Hyperion which Zuck boasts is will be the size of Manhattan and eventually scale to five gigawatts of capacity.
Meta is also spending lavishly on eight-plus figure salaries as it works to build out its AI superintelligence team to build new systems that not just match human intelligence but exceed it.
So, it’s a good thing that Meta’s old school ML is still paying the bills, as Li revealed that the compensation packages tied to these strategic hires will be the second largest driver in expense growth in 2026.
“We really believe 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,” Li said.
At least for the moment, Zuck’s AI spending and superintelligence pipe dream aren’t going to put Meta in the poor house just yet. In Q2, Meta’s profits jumped 36 percent year over year to $18.3 billion on revenues of $47.5 billion. ®
Believe it or not, Meta’s AI investments made a meaningful difference to its advertising business in Q2 — it’s just that those models aren’t the kind that’s got everyone, including the Social Network, plowing tens of billions of dollars a year into datacenters.
It’s going to be a few more years before the Zuckercorp’s rampant spending on infrastructure and AI visionaries starts to bear fruit, CFO Susan Li reiterated on Wednesday’s second quarter earnings call.
“We don’t expect that our genAI work is going to be a meaningful driver of revenue this year or next year,” she said.
For now, improvements in the more conventional machine learning models that power Meta’s recommender systems are paying the bills.
Broader deployment of the company’s new AI-powered recommendation models drove a roughly five percent increase in ad conversions across Instagram and a three percent gain across Facebook, CEO Mark Zuckerberg touted in his prepared remarks on Wednesday.
As their name suggests, these models are designed to boost engagement across the company’s various platforms by connecting users to friends, posts, and other content. And in the last quarter, Zuck boasted improvements in these systems led to a five to six percent increase in time spent doomscrolling Facebook and Instagram.
More importantly, these recommender models are how Meta makes money by sorting through tens or hundreds of thousands of potential ads to find the ones most likely to be relevant to individual users. This process is so important, in fact, that Meta has developed custom accelerators to ensure they run as quickly as possible.
Going forward, Li and Zuckerberg say genAI will eventually play a bigger role in its advertising business. The Facebook parent is already using genAI to help users create advertising collateral, for example.
“Nearly two million advertisers are now using our video generation features, image animation, and video expansion, and we’re seeing strong results from our text generation tools,” Li said.
Li also revealed that Meta has begun incorporating large language models (LLMs) into the recommender systems used by Meta’s X-competitor Threads. “LLMs are now driving a meaningful share of the ranking-related time spent gains on Threads,” she said.
Meta is toying with LLMs and other genAI tools in other areas as well. Zuck has previously teased an AI engineer designed to speed development of his Llama model herd.
On Wednesday’s call, he said teams had begun using Llama 4 to build autonomous AI agents to “improve the Facebook algorithm to increase quality and engagement.”
Speaking of the open weights models, Zuck said work on Llama 4.1, 4.2, and future models is ongoing, but stopped short of providing a concrete roadmap.
All of this is “happening in low volume right now, so, I’m not sure that result, by itself, was a major contributor to this quarter’s earnings… but I think the trajectory on this stuff is very optimistic,” Zuckerberg added.
- Zuck tries to justify AI splurge with talk of ‘superintelligence’ for all
- Europe’s AI crackdown starts this week and Big Tech isn’t happy
- New Google AI model maps world in 10-meter squares for machines to read
- Altman embraces inner Viking, raids Europe with 100K GPU supercluster in Norway
For now, Meta is focused on building out the infrastructure necessary to fuel the development of new generative AI models by investing in infrastructure and talent.
The company is currently in the process of building out a series of AI compute clusters, including a gigawatt scale facility called Prometheus set to come online in 2026 and another called Hyperion which Zuck boasts is will be the size of Manhattan and eventually scale to five gigawatts of capacity.
Meta is also spending lavishly on eight-plus figure salaries as it works to build out its AI superintelligence team to build new systems that not just match human intelligence but exceed it.
So, it’s a good thing that Meta’s old school ML is still paying the bills, as Li revealed that the compensation packages tied to these strategic hires will be the second largest driver in expense growth in 2026.
“We really believe 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,” Li said.
At least for the moment, Zuck’s AI spending and superintelligence pipe dream aren’t going to put Meta in the poor house just yet. In Q2, Meta’s profits jumped 36 percent year over year to $18.3 billion on revenues of $47.5 billion. ®
