Q3 revenue $64.6B vs $63.3B expected
Q4 revenue guide $73–76B vs $73.7B expected
Q3 capex $34.9B; free cash flow turned negative
2026 capex $135–145B (floor raised from $130B)
$825B in future spending commitments
New nonvoting Class C shares through a 2-for-1 split
Meta beat on revenue, beat on guidance, and the stock is still down 7% and that’s a troubling sign for everything touching AI. It’s also trading below where it did before the rollout of the impressive Muse app.
Shares are down intraday to $651.61 and were down 7.3% around midday. Meta was off about 3% premarket at $677.50, and the selling has intensified as the session wears on.
w
The ad business isn’t the problem. Revenue rose 26% to $64.6 billion, comfortably ahead of estimates. The Q4 range of $73–76 billion has a $74.5 billion midpoint, above the $73.7 billion consensus, and implies roughly 24% growth. Strip out the lawyers and earnings were even better. GAAP EPS was $3.83 vs $3.56 expected and would have been $7.40 without the legal charge.
If you showed me that revenue line a year ago, I’d have said the stock goes up 10%.
The cash flow line
The problem sits below revenue. Q3 capex hit $34.9 billion, up from $31 billion in Q2 and $19.4 billion a year ago. That pushed free cash flow to negative $2.1 billion, the first negative quarter since 2012.
Total expenses rose 68% y/y to $51.6 billion on infrastructure, third-party AI token costs and a $9.9 billion charge tied to an August agreement with several state attorneys general.
Then there’s the number that will stick. A regulatory filing disclosed about $825 billion in future spending commitments for AI data centers, energy and compute, up from nearly $700 billion in July. That includes $356 billion in leases that haven’t started yet and $455 billion in non-cancelable contracts, with $121 billion due in 2027 alone. The two big buckets went from about $139B to $810B in a year, nearly 6x
Class C is the tell
Financing is increasingly a problem throughout the industry. Meta is creating a nonvoting Class C stock, effective January 19, 2027, and distributing it to all holders as a dividend. It works as a two-for-one split.
Meta says the purpose is to give the company the ability to finance its growth and stay ahead of fierce competition at home and abroad. The market is seeing the new class as a currency that will dilute the company.
The obvious comparison is Alphabet’s 2014 recapitalisation. Class C gives Meta stock to use for employee pay, acquisitions and raising capital. With free cash flow negative, buybacks paused and the AI build-out escalating, it lets management issue stock without diluting Zuckerberg’s voting control.
The bears had already done the arithmetic. Oppenheimer projects about $135 billion in AI- and Muse-driven free cash flow losses through 2028, more than Meta’s roughly $90 billion in cash and securities, and it has flagged the risk of an equity raise and now Meta built the tool to do exactly that. The years of buybacks are going to be unwound quickly.
The bull case hasn’t gone away
The sell side isn’t capitulating. Barclays kept its Overweight rating and raised its target to $880 from $780.
Muse adoption is the main argument. Zuckerberg said more than 10 million people in the US and Canada set up Muse personal agents in the first six weeks, and about a quarter use them daily. He used the word non-stop in the call. Meta also said Reality Labs losses should shrink to around $15 billion in 2027 and move toward profitability after that.
The spending is contractual and front-loaded. Revenue from agents is promising but early.
It’s not just Meta
This is bleeding into the broader AI trade. Tech is the weakest sector, led by Nvidia and AMD, and memory makers Micron and SK Hynix are lower too. Lam Research fell as much as 10% after management flagged supply chain constraints into 2027. The Nasdaq is down 2.3%, and the thinking is that there just isn’t enough capital out there to build all the AI that hyperscalers want and they’re either going to pay more or scale back plans.
That last point matters more than it gets credit for. A company funding an $825 billion build-out is starting to look at equity markets while the risk-free rate is 5%. Every hurdle rate goes up. The patience for “trust us, the ROI is coming” goes down.
The takeaway is simple. Meta’s core business is as strong as it has ever been. But for the first time in more than a decade, it doesn’t generate enough cash to pay for its ambitions, and management just told you how it plans to fill the gap. Until there are hard returns on Muse, the market will price Meta as a capital-hungry AI infrastructure company rather than a cash-printing ad machine. That’s a lower multiple.
At around $1.65 trillion and 22x trailing earnings, Meta looks cheap on earnings. The market is valuing it on cash flow instead.
This article was written by Adam Button at investinglive.com.
