Everyone's asking the same question heading into Meta's Q1 print: can this company actually make real money from AI outside its consumer ad business? Analysts fret that without proven enterprise sales, API licensing, or breakout products from Llama, the massive infrastructure spend looks reckless next to hyperscalers with cloud revenue to offset it. You hear the worry—$115-135 billion in 2026 capex on top of a business that's supposedly one-trick. Markets price the risk of dilution and margin pressure if the AI bets don't spawn new lines fast.
That's the consensus view, and it's missing the punchline. Meta generated $200.97 billion in revenue for full-year 2025, up 22% year-over-year, with Q4 alone hitting $59.89 billion, up 24%. Net income for the year landed at roughly $60.5 billion. Nearly all of it came from ads. This isn't a speculative pivot—it's a pure-play ad efficiency engine that's scaling at unprecedented levels while rivals chase diversified AI services with slower, less measurable ROI.
Look at the flywheel in action. Q4 2025 ad impressions rose 18% while average price per ad climbed 6%, driving the bulk of that revenue acceleration. Advantage+ campaigns, powered by AI, now handle over $60 billion in annualized ad spend and have delivered CPA reductions in the 9% range for users, with broader AI tools showing 22% higher return on ad spend versus manual setups in testing. These aren't hypothetical gains—they're showing up in the numbers that matter: higher engagement, better conversion, and advertisers pouring more budget into the platform because the math works better than anywhere else.
The video generation tools inside the suite reached a $10 billion annualized run rate by late 2025, growing nearly three times faster than overall ad revenue. That's not a side project; it's supercharging creative output for millions of advertisers, especially smaller ones who couldn't previously compete on production quality. Meta AI itself passed 700 million monthly actives, and Llama hit over 1 billion downloads. Open-source distribution builds a developer moat that loops back into platform stickiness and data advantages, lifting engagement indirectly without needing direct licensing revenue in 2026. This is AI monetization happening in the place Meta dominates: the attention economy it already owns.
Now tie it to the spend everyone panics over. 2025 capex was $72.22 billion. Guidance for 2026 jumps to $115-135 billion, funding custom MTIA chips, data centers, and even nuclear power deals. Total expenses are eyed at $162-169 billion. Yet management explicitly guided for 2026 operating income to exceed 2025 levels, holding 41% operating margins in Q4 despite the ramp. Q1 2026 revenue guidance of $53.5-56.5 billion already sits well above Street expectations around $51 billion, signaling the ad engine is absorbing the investment and accelerating. Meta doesn't need a cloud business by design—its AI is laser-focused on making the $200 billion ad machine print even more cash at scale.
The deadpan fact that reframes everything: Meta printed roughly $60 billion in net income in 2025 almost entirely from ads while running at 41% operating margins. That $115-135 billion capex is an efficiency tax on a proven cash compounder, not a Hail Mary into unproven markets. Hyperscalers spread their bets thinner; Meta doubles down where the returns are fastest and most measurable. The market is lazy here, treating the spend as pure risk instead of the logical next gear for a business that's already overtaking Google in projected ad revenue by 2026.
You don't need new revenue lines to justify this if the core keeps compounding. The AI investments are paying off in ad yield today—higher impressions, better pricing power, lower effective costs for advertisers. That's the variant perception: the capex collision isn't a threat to margins; it's the fuel keeping the flywheel spinning faster than competitors can match.