Meta Platforms reports second-quarter results after the close today, and the market has already decided the story: the stock has clawed back roughly 10% from its 2026 low to sit near $600, lifted by chatter about a cloud computing pivot and a new AI model called Muse Spark 1.1. The problem is that neither shows up in the numbers. The Anthropic cloud talks are described as very preliminary and the model launched barely two weeks ago. What actually matters for the quarter is whether the advertising engine can keep paying for the most expensive infrastructure build-out in corporate history.
The ad machine slows
Analysts expect revenue of $60.21 billion, a year-over-year increase of just under 27%. That sits at the top end of Meta’s own $58 billion to $61 billion guide, but it represents a clear deceleration from the 33% growth posted in the first quarter, the fastest pace since 2021. Adjusted earnings per share are seen at $7.20, barely 1% higher than a year ago, and the Q1 beat of $3.77 was mostly an unusual tax benefit. The underlying levers tell the same story: ad impressions delivered rose 19% in Q1 and price per ad rose 12%, both the best rates in over a year. With revenue growth stepping down, a modest slip in both metrics would be the logical result.
The spending question
The number that moved the stock last time was capital expenditure. In April Meta lifted its full-year CapEx range to $125 billion to $145 billion, an 8% increase at the midpoint, and shares fell more than 8% in response. Another increase tonight would test the market’s patience again, but the context has shifted. Investors are now weighing that spend against a 40% upside consensus price target and the promise of new revenue lines. The discipline on display will matter more than the dollar figure: whether management frames the outlay as a bridge to cloud services or simply the cost of defending the ad moat.
Cloud talk is cheap
Management will almost certainly address the cloud rumors, it would be strange not to after weeks of reporting, but confirmation is not a business model. The options range from selling raw compute to offering managed AI infrastructure, and the economics of each are wildly different. Muse Spark 1.1 is similarly long-dated. For tonight, the only real-time verdict comes from the ad business and the CapEx guide. Everything else is option value, and the market has a habit of pricing that before it exists.