Meta jumps 15% this week: Wall Street backs its AI spending

Meta rolled out Meta Compute this week, opening up its AI compute and models to outside customers and, at least on paper, giving investors a fresh revenue angle beyond advertising. The catch is that Meta still hasn’t actually sold the service, and it would be walking into a market where Amazon Web Services, Microsoft Azure, and Google Cloud already have deep roots.

Investors liked the pitch anyway. Shares rose 6% Friday and 15% for the week in July, Meta’s best stretch since early 2024. Part of the reaction was simple: the market finally had a clearer revenue narrative for Meta’s planned $125 billion to $145 billion in 2026 spending on data centers, chips, and models, after a year when the stock trailed the Nasdaq-100’s 18% gain. Ads still account for 98% of revenue. Meta said Q1 revenue came in at $56.3 billion, up 33%. But the rally was really about cloud compute, the Meta Model API, Muse Spark 1.1, Muse Image, and Muse Video, all priced aggressively.

If you watch Meta as more than an ads business, this update is worth a look. Nothing has been sold yet. Meta has never operated a public cloud at scale. Amazon Web Services, Microsoft Azure, and Google Cloud are firmly entrenched. And if the company builds too much capacity, that could end up looking like overbuilding.

Still, there’s plenty here for the market to latch onto. Meta’s Training and Inference Accelerator chips are meant to reduce its dependence on Nvidia. Iris is set to enter mass production in September 2026. Analysts think data-center costs in 2027 could come down by as much as 35%. Wolfe says each gigawatt at a $25 billion annualized run rate could raise EPS by 20%. Options volume ran more than 3x average, with 78% of the $1.8 billion premium going into calls, and price targets sit between $720 and $869.

The tools themselves will be available through Meta Compute and the Meta Model API.