To The Moon

AI Capex Showdown: Why Microsoft Got a Standing Ovation and Meta Took a Hit

I've been watching these tech earnings for years, and last night's reports from Microsoft and Meta delivered one of the cleanest case studies in how Wall Street actually prices AI bets. Both companies dropped monster revenue numbers. Both are pouring tens of billions into the same infrastructure gold rush. Yet the market sent MSFT shares up sharply after hours while hammering META down nearly as hard in the other direction.

Let's break it down without the hype.

The Numbers Side by Side

Microsoft's fiscal Q4 (ended June 30, 2026) came in at $90 billion revenue, up 18% year-over-year and ahead of expectations. Adjusted EPS hit $4.74, crushing the street. Azure cloud growth accelerated to 43%, pushing the annual cloud run rate past $100 billion for the first time. Microsoft Cloud overall did $59.3 billion for the quarter. Copilot adoption is scaling fast with millions of paid seats. Backlog exploded. Free cash flow took a hit from capex but still beat lowered fears.

Meta posted Q2 revenue of $60.8 billion, up a robust 28%. Ad impressions and pricing both rose nicely, user metrics held steady. But EPS came in at $6.18, missing estimates partly due to one-time legal and severance charges. Operating margins compressed hard. And free cash flow? It cratered from $8.55 billion last year to just $784 million. Capex guidance got a floor raise—now $130-145 billion for the year.

On paper, Meta's top-line growth looks sexier. So why the divergent reactions?

The Real Story: Monetization Paths

This isn't really about who spent more on GPUs. It's about whether investors can see the dollars flowing back in real time.

Microsoft sells the picks and shovels. Azure is the hyperscale platform where enterprises pay premium rates for AI workloads today. You can point to signed contracts, a massive remaining performance obligation north of $678 billion, and Copilot seats turning into predictable subscription revenue. Satya's team isn't just building capacity—they're renting it out profitably right now. Capex looks like smart capacity planning for known demand.

Meta is mostly eating its own cooking. Yes, AI is improving ad targeting and recommendations, which helped drive that 27-28% ad revenue growth. But the massive infrastructure build is still largely internal. The returns show up indirectly in better engagement metrics rather than direct high-margin AI service revenue. Investors watched free cash flow evaporate and heard another big capex commitment without the same visible payback timeline. Zuckerberg is betting big on future AI products and efficiency gains, but the market wanted clearer proof this quarter.

I've seen this movie before. When spending outpaces visible returns, patience wears thin—especially after years of elevated valuations across Big Tech.

Context Matters

Microsoft had some helpful tailwinds: efficiency gains in its own models, custom silicon progress, and a diversified business mix that includes steady Office and Windows revenue. Meta's ad business remains incredibly strong fundamentally, but it's also more exposed to any cyclical softness in advertising spend. Legal noise and severance costs didn't help the optics either.

Both stocks had been under pressure YTD coming into the print. The market was looking for reasons to differentiate winners in the AI arms race. Microsoft gave them a clean narrative: "Our AI investments are already driving accelerating growth." Meta's was more "Trust us, the payoff is coming—and we're doubling down."

What I'm Watching Next

As a long-time investor in this space, I'm not bailing on either story. Meta's core business throws off enormous cash when capex normalizes, and their AI talent and data moat are real. But near-term, the bar for "AI spend forgiveness" just got higher. Companies need to show either strong current monetization or ironclad evidence of future ROI.

Microsoft reminded everyone why enterprise cloud leaders have structural advantages in this cycle. The reaction wasn't random—it was the market doing its job: rewarding proof over promise.

We'll see how the rest of the hyperscalers handle this new scrutiny. For now, the message seems clear. Build it, sure. Just make sure customers are paying you enough to justify the bill.