Meta is spending over $140 billion on AI this year.
$META reported second-quarter 2026 numbers after the market closed on Wednesday. Revenue of $60.8 billion, up 28 percent year-over-year. Earnings per share of $6.18, down 13 percent, and well below analyst estimates of over seven dollars. Operating margin fell from 43 to 31 percent. The market reacted with a decline, and the stock fell toward its annual low around $520. But the most interesting thing about the entire report isn’t the number itself. It’s what’s behind it.
$MSFT reported on Tuesday and beat estimates across all key metrics. Meta reported yesterday and slumped. Tonight it’s $AMZN’s turn. So this week we’re getting a complete picture of how the three giants are spending and earning from AI at the same time, and that picture shows the market is judging each company differently, even though they’re all doing essentially the same thing.
Meta has already burned $70 billion on a bet that didn’t pan out
This isn’t the first time Meta has told investors “trust us, this investment will pay off” and gotten hammered for it. In 2021, it renamed the entire company from Facebook to Meta and bet the future on the metaverse. It created the Reality Labs division and started pouring billions into it.
The result was brutal. Since 2021, Reality Labs has accumulated over $70 billion in operating losses, at a pace that worsened every year, from ten billion in 2021 to current annual losses of around $18 to 19 billion. Meanwhile, the division’s revenue is in the hundreds of millions, not billions. Meta’s stock lost nearly two-thirds of its value in 2022, and much of the blame fell on the metaverse that nobody wanted.
This year Meta finally admitted defeat. It plans to cut Reality Labs’ budget by up to 30 percent and shift the freed-up money to where it’s already seeing returns, which is AI. The market rewarded this news with a several-percent stock rally, because it finally saw discipline where it had been missing for years.
That’s the key difference from today’s sell-off. The metaverse was a bet without evidence, years without a single signal that it was paying off. Today’s AI capex is already showing results, but no one is looking because a headline about a 13 percent profit drop is easier to write.
Where the profit actually went
Meta booked $2.4 billion in second-quarter costs for ongoing litigation and $1.18 billion in severance following the layoff of roughly 8,000 people this May. Combined, that’s more than $3.5 billion in one-off charges that have nothing to do with AI infrastructure.
CFO Susan Li said that without these one-time items, operating profit would have risen 9 percent year-over-year, instead of falling 8 percent as the official number shows. So the core business isn’t slowing. It’s being slowed by one quarter’s accounting, not the performance of the ad machine.
And that machine, meanwhile, is performing better and better. Meta reports that thanks to AI model deployment, ad clicks rose 8.3 percent and Facebook conversions rose 15.7 percent. Over a million businesses a week are now using the new AI business assistants. Those are exactly the kind of numbers Reality Labs never managed to produce in four years.
CapEx that changes the game
Meta raised its full-year capex estimate to $130 to $145 billion this year, another increase from the earlier $125 to $145 billion. It spent $72.2 billion last year, so that’s nearly double. In the second quarter alone, $31.1 billion went to servers and data centers, more than double the $17 billion a year earlier. Part of that includes a more-than-ten-billion-dollar partnership with BlackRock to build a data center in El Paso, Texas.
The number only makes sense in the context of the whole industry. Microsoft plans around $175 billion in capex for its next fiscal year. Amazon has roughly $200 billion slated for this year. Google raised its estimate to $195 to $205 billion, and the market punished it with a 6 percent drop after its last report, even though revenue and earnings themselves beat estimates. So Meta isn’t an outlier that lost its mind. It’s one of four companies that decided no one can afford second place in this infrastructure war.
Why I see this sell-off as an opportunity, not a warning
The metaverse taught me to be cautious of Meta’s promises without numbers. That’s exactly why today’s report doesn’t make me nervous—it reassures me. This time the numbers are there, and they’re here now, not in five years under a best-case scenario.
A company with 3.6 billion daily users and a cash reserve of over $90 billion can afford to build the infrastructure it will need for the next decade. The margin compression doesn’t worry me because I know exactly where it comes from, and more than half of that pain is one-time legal and personnel costs, not a failing business.
What reassures me most is the combination of both stories at once. Meta is cutting where money has disappeared for years without results, while pouring billions into areas where it already sees a direct revenue impact. That’s exactly the discipline I had been criticizing it for lacking with the metaverse. I’m bullish on $META, and today’s numbers only reinforce that.