Meta turns to Google’s Gemini to power its ad business despite $600B AI Bet

Meta has poured billions into artificial intelligence, but it may soon rely on Google to run one of its most critical businesses. According to The Information, Meta staffers have held talks with Google Cloud about using its Gemini models to sharpen the ad targeting that drives Facebook, Instagram, and WhatsApp.

“Meta Platforms staffers have had discussions with Google Cloud about the possibility of using Google’s artificial intelligence models to improve Meta’s ad business, The Information reported, citing people with knowledge of the matter.

The discussions are still early and might not produce a deal, but the very fact that they’re happening shows how much pressure Meta is under. For all of Mark Zuckerberg’s spending—an estimated $600 billion committed to AI infrastructure through 2028—the company’s in-house models have struggled to keep pace with rivals.

Meta’s AI investments have been vast. Earlier this year, the company raised its 2025 capital spending outlook to as much as $72 billion, much of it earmarked for data centers stacked with Nvidia GPUs. In total, the company expects to have more than 1.3 million GPUs in play as it races toward artificial general intelligence. It also went on a hiring spree, offering $100 million bonuses to lure OpenAI researchers, adding four to its newly reorganized AI group.

But in the short term, Meta is still looking outside. According to the report, employees have floated the idea of fine-tuning Google’s Gemini and open-source Gemma models on Meta’s ad data. That would mean using the very technology of its direct competitor in online advertising, an arena where Meta and Google already battle for dominance.

This wouldn’t be the first time Meta explored outside partnerships. Last month, The Information reported the company had weighed deals with Google and OpenAI for conversational AI features inside Meta’s apps. The move to consider Gemini for advertising, however, signals that Meta’s internal progress has not been enough to meet the demands of its $134 billion ad empire.

The company’s Llama models, including Llama 4, have shown strong benchmarks in some areas but still lag behind in reasoning and math—shortcomings that may explain why Meta is willing to consider Google’s help. The partnership talks also come against the backdrop of a broader AI arms race that’s pulling the biggest players into expensive bets and aggressive talent wars.

DeepSeek, an open-source model that shocked the industry in January, only intensified that race. It’s part of what pushed Meta to overhaul its superintelligence team and bring in heavyweight researchers. With Wang leading Meta’s AGI efforts, the company is signaling it won’t back down from trying to build AI systems that can reason and learn at higher levels.

For now, though, the irony is clear: Meta may need to lean on Google’s Gemini to keep its ad machine humming, even as it prepares to spend hundreds of billions to build a future where it doesn’t have to.