The number that framed Meta's Q2 for me wasn't the EPS miss but the free cash flow number, $784m, down 91% yoy. Capex hit $31.1b in the quarter and fy guidance sits at $130-145b. For a company that used to be a cash printing machine, that shift in spending is what scared most retail and drove the sell off.
Steelmanning the bears because they won the day with Meta dropping 9%. Meta is the only one of the mega cap AI spenders without a cloud business. Microsoft reported Azure up 43% and crossed $100b in annual Azure revenue, Amazon posted its best AWS growth in years. Both got 8-10% pops the same week and every dollar of their capex has a visible revenue meter attached. Meta's doesn't and it's funding one of the largest infrastructure builds in corporate history against a single ad business, it raised capex guidance twice this year, and it guided Q3 revenue below consensus. Two brokers cut targets on it this week and if you think AI capex needs a monetisation receipt, Meta is the name with the weakest one.
The other side was the miss being somewhat mis leading, EPS was dragged by $2.4b in legal charges and $1.18b of severance from an 8,000 person cut which were both largely one off, and they compressed operating margin to 31% from 43%. Revenue grew 28% to $60.8b and actually beat but no one bothered to really appreciate it. Strip the noise and the ad engine didn't crack with pricing holding, DAP hit 3.60b. The stock now trades near $557 on roughly 17x forward earnings, versus a 3 to 5 year average around 23x. On the call Zuckerberg's argument was that Meta plans to sell intelligence at a durably higher margin than renting raw compute and it's already fielding offers for compute at a premium to what it paid.
The debate everyone's having is ad growth vs capex which actually resolves on a third variable most people aren't tracking which is the compute optionality. Zuckerberg said on the call Meta's fielding offers to buy its compute at a premium to what it paid, and that selling intelligence carries a structurally higher margin than renting raw compute. That's a second revenue line that sits in zero analyst models today. So the real test by the Q3 print isn't the ad number, it's whether any enterprise/compute revenue gets disclosed or guided. If Q3 comes and goes with nothing on that front and ad growth also decelerates below 20%, the bears were right and it's a value trap dressed as a growth story.
I've been mapping the capex to monetisation gap across all the hyperscalers for a while now and who's actually earning their spend back and who's just spending, and Meta screens differently than the tape suggests once you separate the one off charges from the run rate.
What's the actual bear case here that isn't just the capex number is scary because I've looked and I can't find one that survives the ad growth. Drop your thoughts.