META's choppy Muse debut: an overvaluation warning or a long-term buying signal?

Meta's second-quarter print on July 29 was brutal on cash flow and nearly painless everywhere else. Revenue rose 28% to $60.8 billion, and ad sales hit $59.4 billion, up 27%. Free cash flow collapsed 91% to $784 million, because quarterly capital spending devoured roughly 98% of operating cash flow. Shares fell about 10% in after-hours trading and have since recovered to sit roughly 22% below the 52-week high. The market is pricing Meta as if the spending is the business.

The launch-day noise around Meta's new AI work should not decide the answer. When Meta debuted Muse Spark, its first major AI model in over a year, built around the $14.3 billion Scale AI tie-up and led by chief AI officer Alexandr Wang, the stock popped 6.5% in a day. A strong update, Muse Spark 1.1, popped it again in July, priced at roughly a quarter of what OpenAI and Anthropic charge. It has chopped around anyway, because the effort's revenue is still unproven. A launch-day pop, like a post-earnings drop, tells you about sentiment, not about whether money can be generated here.

The same goes for the other two signals. Zuckerberg has sold stock for years through pre-scheduled 10b5-1 plans, and a director's small sale this summer was disclosed as part of one of those plans. Insider sales scheduled in advance are not a message about valuation. And the analyst targets show only how wide the disagreement is: after the Q2 report, they ran from $640 to $930, with Citi at $800, Goldman at $725, and Rosenblatt holding the top at $886. Even so, no firm downgraded Meta to a sell. Cutting the ceiling while keeping the rating is a market saying it wants proof before it pays up.

The one number that sets the multiple

That proof is free cash flow, and it is worth looking at as arithmetic rather than as drama. Free cash flow is operating cash flow minus what the company spends on plants and equipment. For the trailing year, Meta generated about $130 billion of operating cash flow, spent roughly $92 billion on capex, and kept about $38.5 billion of free cash flow — down 23% from a year earlier even as revenue grew 28%. On that basis Meta's stock costs about 40 times trailing free cash flow. That is the expensive-looking number, and it exists only because spending has jumped ahead of what the ad engine is converting into cash.

Get the spending to stop growing faster than cash flow, and that multiple screws down quickly. The 2026 capex guide of $130 billion to $145 billion, with the lower end just raised by $5 billion on higher component costs, says the pressure is not over. The honest warning from the sell side is that it may not be over next year either: JPMorgan's model projects 2027 capex could reach $243 billion, up 70%. If that is even close, the free cash flow crush extends well beyond the next 12 months, and the 40-times multiple stays high with no earnings escape valve. That is the specific condition that turns the overvaluation case from a fear into a fact.

The counterweight is real and measurable: the ad business is not merely intact, it is accelerating. Ad revenue up 27%, propped up by 14% more impressions and 12% higher prices, at an 82% gross margin. Even the quarter's falling operating margin — 31% versus 43% a year ago — is masked by $2.4 billion of legal charges and $1.18 billion of severance; strip those out and operating income still grew 9%. Meta is in the unusual position of having its core cash engine revving while its own spending temporarily eats the cash before it reaches the net line. That is the expectations-reset contrast: the market is still pricing the old risk profile of an AI spender, while the ad engine underneath keeps getting cleaner.

The bridge, and the break

This is not yet a clean long-term buy, and it is not yet a clean signal to avoid. It is a thesis with two defined outcomes, and the deciding variable is whether free cash flow re-stabilizes as the capex wave crests. Watch three proof points. First, does capex peak near the top of the $145 billion guide, or does it keep climbing toward the $243 billion projection? Second, does ad-led operating cash flow — still growing with revenue — re-widen the gap above capex over the coming quarters, so free cash flow stabilizes and the 40-times multiple starts to compress? Third, does Muse turn per-token pricing into actual dollars, or does the cloud/compute capacity-selling Meta now talks about become material revenue rather than a hedge?

The case for the entry dies on a specific condition: if capex keeps rising while free cash flow stays depressed and Muse revenue stays immaterial, then the overvaluation view wins, because the market is paying a high multiple for an engine that keeps transferring its output to spending. The case for the entry is proven by the opposite — ad-led growth re-stabilizing free cash flow even before Muse contributes anything.

I can be wrong again; the 2027 capex projection is exactly the kind of number that humbles a thesis. But that is the point of laying it out this way. The choppy Muse debut and the $640-to-$930 target spread are the market's confusion made into noise. The number that decides Meta's forward multiple is free cash flow re-stabilizing through the capex peak. Until a quarter proves that, the disciplined move is not to buy the dip by reflex and not to call it overvalued by reflex — it is to hold one question in place and wait for the cash-flow answer.