Meta’s Ad Machine Is on Fire — And It’s Willing to Spend $135B on AI to Keep It That Way

Meta’s latest earnings report delivered exactly what investors were hoping to see: a core advertising business that continues to compound at a healthy pace, paired with clear evidence that the company’s massive AI spending push is translating into real performance and profit. While Meta once again raised eyebrows with aggressive expense and capital expenditure plans, management’s confidence that operating income will grow in 2026 helped frame the quarter as one where monetization is outpacing investment rather than being overwhelmed by it.

Q4 revenue came in at $59.89B vs. ~$58.35B expected, while EPS was $8.88 (ahead of consensus). The operating margin printed at 41% — down from last year, but still the kind of number that makes “spend panic” harder to sustain when the top line is growing 24% and operating income is massive.

The ad business did the heavy lifting, as usual, but the composition matters: impressions rose 18% and average price per ad rose 6%, which is the “volume + pricing” combo that tells you demand is healthy and performance is improving rather than just stuffing more ads into feeds. CFO Susan Li put it plainly: “The average price per ad increased 6% year-over-year, benefiting from increased advertiser demand, largely driven by improved ad performance.” That performance theme runs through the call — Meta keeps marrying better recommendation systems with better ad ranking, and then monetizing the incremental engagement without wrecking the user experience.

What’s different in this cycle is how explicitly Meta is tying AI investment to measurable ad monetization gains. Li walked through the mechanics: scaling ranking models (like GEM) and using that knowledge to improve lightweight runtime models used for inference. In Q4, Meta said GEM plus a new sequence learning architecture drove “a 3.5% lift in ad clicks on Facebook” and “more than 1% gain in conversions on Instagram,” while a new runtime model rollout helped drive a “3% increase in conversion rates” across key Instagram surfaces. That’s the “AI spend isn’t charity” argument investors want to hear — especially when it shows up as stronger conversion growth and pricing power.

On top of the ad flywheel, Meta is still quietly building a second monetization leg through messaging. Paid messaging on WhatsApp crossed a $2B annual run rate, and “click-to-message” ad revenue growth accelerated with the U.S. up more than 50% YoY. Meanwhile, Meta is expanding ad inventory in newer surfaces: Threads ads are rolling out more broadly, and Li said Meta expects to complete the rollout of ads on WhatsApp Status through the year, “with the level of ads remaining low in the near term.” The strategy is classic Meta: prove engagement, introduce light ads, optimize performance, then ramp.

Now to the part that should have spooked the market — but didn’t. Expenses and CapEx were both big, and guidance is even bigger. Q4 total expenses were $35.1B, up 40% YoY, driven by compensation (AI talent), legal, and infrastructure costs. For 2026, Meta guided total expenses to $162B–$169B, with expense growth led by infrastructure (cloud spend, depreciation, operating expenses) and then employee compensation as it continues to invest in technical talent. And the headline number: 2026 CapEx (including principal payments on finance leases) guided to $115B–$135B, both numbers were well above expectations.

So why was the market still in a good mood? Because Meta didn’t pair “spend up” with “profit down.” Li explicitly framed the year as reinvesting from strength and said, “Despite the meaningful step up in infrastructure investment, in 2026, we expect to deliver operating income that is above 2025 operating income.” That’s the tell: profitability is expected to outpace spend growth in absolute dollars, which signals management believes ROI is showing up fast enough to keep the model self-funding. In other words, investors heard “we’re spending like it’s 2021,” but also “we’re earning like it’s 2025.”

A second reason the tape rewarded it: management acknowledged constraints (which implies demand) and described active mitigation. Li said they “continue to be capacity constrained” because demand for compute has increased faster than supply, and expects they’ll “likely still be constrained through much of 2026” until more capacity comes online later in the year. In the current AI narrative, “constrained” is basically corporate-speak for “we could deploy more and still get paid for it.” Meta is also diversifying silicon supply and improving efficiency — including extending Andromeda to run across NVIDIA, AMD, and MTIA, and nearly tripling Andromeda’s compute efficiency.

The call also reinforced a culture-level productivity angle that investors increasingly treat as margin insurance. Meta’s CFO said “output per engineer climbed 30%, with most gains driven by agentic coding,” and Zuckerberg echoed the same theme in plain language: “We’re elevating individual contributors and flattening teams,” adding that “projects that used to require big teams now be accomplished by a single, very talented person.” If you’re asking markets to underwrite a CapEx super-cycle, pairing it with internal efficiency gains is the closest thing to a “don’t worry about it” button you can press without getting punished.

Guidance helped, too. Q1 revenue guide of $53.5B–$56.5B came in strong, with a noted FX tailwind. And the business-level color was constructive: Li said advertiser demand remained healthy, with online commerce the largest contributor to YoY growth (lapping election-driven political spend), followed by professional services and tech. That’s important because it implies broad-based demand rather than a single vertical propping up the quarter.

From a stock setup standpoint, the reaction makes sense in context. Meta’s shares (around $668 late Wednesday) reflected a market willing to pay up for a company that’s still growing the ad core at a 20%+ clip while building an AI platform on top. The move ripping through the 200-day moving average and nearly filling the post-Q3 gap is exactly what a “fundamentals + narrative + positioning” squeeze looks like when investors decide the spending surge is a feature, not a bug.

What to watch next is pretty straightforward (and inconveniently important): (1) whether ad pricing stays supported as compute scales into 2026 (i.e., do the performance gains keep translating into pricing power), (2) whether expense growth remains explainable (infrastructure + AI talent) rather than “miscellaneous bloat,” (3) evidence that WhatsApp/Threads ads can ramp without hurting engagement, and (4) any incremental disclosure on how MetaCompute financing/JV structures could smooth the cash flow optics of that $115B–$135B CapEx guide. Meta’s message is basically: we’re spending more, because we’re making more — and for now, the market is buying it.