Why Is Meta Betting Big on AI Advertising?

Just a few years ago, Meta was betting its future on the metaverse. Now, it’s going all-in on artificial intelligence. The latest evidence came this week with the announcement of a new tool that allows advertisers to use AI to instantly create video ads from still product images.

The move is part of a much larger, high-stakes pivot for the owner of Facebook and Instagram. CEO Mark Zuckerberg has made AI a top priority, punctuated by a massive $14.3 billion investment in the data-labeling firm Scale AI and the high-profile hiring of its CEO, Alexandr Wang, to join a new “superintelligence” team. As Meta competes fiercely with OpenAI, Google, and Microsoft, its advertising business—the engine of its revenue—is becoming the primary battlefield for its AI ambitions.

What’s behind Meta’s AI push?

Meta’s aggressive move into AI is driven by both opportunity and necessity. Advertising accounts for about 98% of the company's annual revenue, and AI is proving to be a powerful tool for enhancing ad effectiveness. For the small businesses that form the bulk of Meta’s advertisers, AI tools that lower the cost and complexity of creating marketing materials mean they have more budget to spend on promoting those ads on Facebook and Instagram.

There’s also intense competitive pressure. The launch of ChatGPT in late 2022 sent a shockwave through the tech industry, and Meta has been racing to keep up. Zuckerberg was reportedly frustrated with Meta’s efforts in the AI development, and in particular with the company’s Llama 4 model launch in April 2025 , which was seen by the industry as underwhelming. That motivated him to become more personally involved, leading to the creation of the new superintelligence team and the expensive talent grab from Scale AI.

How is AI changing Meta's advertising?

Artificial intelligence isn't just a defensive play for Meta; it's a fundamental rewiring of its cash machine. On the most recent earnings call, Zuckerberg framed "improved advertising" as the first of his five major AI opportunities, envisioning a future where advertising is essentially an "AI agent that delivers measurable business results at scale."

The strategy attacks the problem on two fronts, creating a powerful business flywheel.

First, AI is democratizing ad creation. The goal, Zuckerberg said, is to allow any business to simply state its objective and have Meta's AI "do the rest," handling both audience targeting and the ad creative itself. This lowers the barrier to entry for the millions of small and medium-sized businesses that are Meta's core advertisers. The company is already seeing strong adoption, with CFO Susan Li noting a 30% increase in advertisers using its AI creative tools in the last quarter alone. New features, like automatically expanding video to fit full-screen formats and virtual try-ons for clothing, reduce the need for outside creative agencies. This frees up marketing budgets, leaving businesses more money to spend on what Meta actually sells: ad placement.

Second, and more importantly, AI is supercharging ad performance. The company’s new "GEM" (Generative Ads Recommendation) model, trained on Meta's largest-ever cluster of GPUs for ads, is already yielding a 5% increase in ad conversions in Reels. An even more powerful tool, the "Incremental Attribution" feature, delivered an average 46% lift in incremental conversions for advertisers in testing. When AI makes ads not just easier to create but demonstrably more effective, it gives businesses a clear justification to increase their spending on the platform, fueling the entire Meta machine.

What is Meta’s AI strategy?

Meta is pursuing a two-pronged strategy: developing powerful, open-weight large language models (LLMs) and integrating AI deeply into its products, especially advertising and hardware.

On the model front, its flagship effort is the Llama series. The company recently released Llama 4, a family of models that includes the small “Scout” version and the larger “Maverick.” While Meta claims they are competitive with models from Google and OpenAI, the release was met with some criticism. Critics pointed to a sly approach to benchmark rankings on the LLM leaderboard LMArena and noted that its open-source license still contains restrictions for very large companies.

To power these models, Meta is investing staggering sums in infrastructure. In May, the company raised its 2025 capital expenditure forecast to between $64 billion and $72 billion, up from a previous plan of $60 billion. CFO Susan Li partially blamed the increase on the uncertainty and higher costs associated with President Trump’s trade tariffs.

Meta is also competing in the emerging AI hardware space. Its Ray-Ban smart glasses, which feature a camera and AI assistant powered by Llama models, have been a surprise hit, selling over 2 million pairs since 2023. This puts Meta a step ahead of rivals in a category where others, like the Humane Ai Pin, have failed publicly.

How does Meta compare to the competition?

Meta's strategy of releasing open-weight models contrasts with the more proprietary approach of competitors like OpenAI. This has helped foster a community of developers building on its technology. However, the AI landscape is crowded and rapidly evolving.

In the AI hardware race, Meta's Ray-Ban glasses are already in the market, while Google is just now launching prototypes of its Android XR glasses with Warby Parker and Gentle Monster. The biggest future threat may come from OpenAI, which recently announced a $6.5 billion deal to acquire the device startup of legendary Apple designer Jony Ive, with the goal of creating a new family of AI-powered consumer hardware.

On the enterprise side, Meta is using AI to boost productivity and improve its core advertising products. This mirrors a broader trend, with analysts noting that big tech players like Meta and Amazon are the immediate beneficiaries of the AI boom, using the technology to directly monetize their existing businesses.

What are the risks to Meta's AI ambitions?

The biggest challenge is the intense war for talent. Top AI researchers are a scarce resource, and companies are offering compensation packages that run into the tens of millions to lure them away from competitors. Meta’s $14.3 billion deal for a 49% stake in Scale AI is widely seen as an expensive “acquihire” designed primarily to secure the leadership of Alexandr Wang.

Furthermore, not every AI bet pays off. The April launch of Llama 4 was considered a misfire by some in the industry, highlighting the difficulty of consistently delivering breakthroughs. Meta must also contend with the immense cost of compute. The AI arms race requires building and maintaining massive data centers, a financial commitment that pressures companies to find a clear path to profitability for their AI investments. While AI is boosting Meta’s ad business today, the long-term return on its massive AI spending remains a critical question as it battles for dominance in the next era of technology.

Reference Shelf:

Meta Testing New Auto-Generated Video Ads as Part of AI Push (Bloomberg)

Google Succeeds With LLMs While Meta and OpenAI Stumble (IEEE Spectrum)

Why is Meta Fueling the Frenzy for AI Talent and Acqui-Hires (ARPU)