Listening to Mark Zuckerberg speak during Meta’s Q1 earnings call on Wednesday, you could almost forget that Meta makes the majority of its money from advertising.
He spent his opening remarks on “personal superintelligence,” custom silicon, Ray-Ban and Oakley AI smart glasses and swarms of business agents that can work “day and night” to help people achieve their goals.
“We are living through a historic technological transformation,” he declared. “We are among the few companies positioned to shape the future.”
But all of that takes a lot of money.
Meta now expects to spend between $125 billion and $145 billion this year, mostly to fund the compute needed to train its new generation of AI systems. That’s up from Meta’s prior range of between $115 billion and $135 billion.
When CFO Susan Li took over from Zuckerberg a few minutes later, she explained how Meta’s existing ad machine is bankrolling this push, with plenty of AI enhancements baked in, of course.
AI, AI, AI …
But it apparently wasn’t what investors wanted to hear.
Although Meta beat expectations, with $55 billion in ad revenue for the quarter, up 33% year over year, its stock was down more than 6% in after-hours trading.
Not that there appears to be any weakness in Meta’s core business. The ad machine is purring, and much of the growth is already being driven by AI under the hood.
For example, Meta is weaving more AI into its ranking and recommendations to keep people engaged. It’s training on longer, more detailed interaction histories so its systems have a clearer, richer picture of what people click on and watch. Meta is also pulling new posts into the system faster so that fresh content rises into people’s feeds and recommended Reels more quickly.
Li said those ranking changes drove a 10% lift in time spent with Reels on Instagram and a more than 7% increase in total video engagement time on Facebook. The mix of recommendations is getting fresher, too, with same‑day videos now making up more than 30% of recommended Reels on both Facebook and Instagram, which is roughly double what it was a year ago.
That same playbook – more AI, more signals, more surfaces – is also at work on the monetization side. In the first quarter, Meta expanded ads on Threads into more markets and continued rolling out ads to WhatsApp Status, helping it squeeze more value out of all that extra engagement.
These moves helped increase ad impression growth – the total number of ad impressions served across all of Meta’s services increased 19% in Q1 – while the average price per ad climbed 12% year over year, although that was offset a bit by lower-monetizing regions.
… and more AI, AI, AI
But beyond just creating more ad slots, Meta is also upgrading its models that decide which ads to show and when.
Meta made changes to Lattice, its AI architecture for predicting and improving ad performance across its systems, and to GEM, its model for ad ranking, which together helped drive a more than 6% increase in conversion rates for landing page ads, Li said.
On top of that, Meta is also starting to plug AI directly into the way advertisers manage their campaigns. During the call, Li announced the open beta of Meta Ads AI connectors, which lets advertisers and agencies link their Meta ad accounts to the AI tools they already use without the need for coding or an API setup. This way they can handle tasks like querying, creating and adjusting campaigns using natural language.
Cool, cool.
But for all the time spent on AI agents, higher-performing ads, upgraded models and everything else, one topic was mostly an afterthought: the lawsuits accusing Meta of creating products that harm young users.
Juries in New Mexico and California have already handed down guilty verdicts holding Meta liable for endangering children and designing addictive features. If that pattern holds, Meta may be facing not just a legal nuisance, but a pipeline of costly losses.
Li, in passing, mentioned “headwinds in the EU and the US” and made a fleeting reference to “scrutiny on youth-related issues” that “may ultimately result in a material loss.”
No one on the call, however, asked a single follow-up question.
