Meta Missed by a Dollar and Microsoft Crushed It — and Both Say the Same Thing About AI

Meta Missed by a Dollar and Microsoft Crushed It — and Both Say the Same Thing About AI

Let me tell you something that kept me up past midnight last night.

July 29, 2026, is going to go down as the single most revealing day in the AI infrastructure story so far. Not because of any one number — but because of what the numbers together tell us about where we're headed. Meta reported Q2 earnings and missed EPS by almost a full dollar. Microsoft reported Q4 and crushed expectations across the board. Same day. Same AI capex firehose. Two completely different verdicts from the market. And somehow, both of them confirm what I've been saying for two weeks straight: this AI buildout is entering a new phase, and most people aren't ready for it.

Let me walk you through what happened, what it means, and what you need to do about it if you're running an independent hosting business.

Meta Missed, Microsoft Crushed — Same Market, Two Verdicts

Menlo Park, CA & Redmond, WA — July 29, 2026 — Meta Platforms reported Q2 2026 earnings after the bell Wednesday, posting revenue of $60.80 billion — a 28% year-over-year increase that actually beat the $60.22 billion consensus. Sounds like a win, right? Except net income fell 13% to $15.85 billion, and earnings per share came in at $6.18 — a full $0.95 below the $7.13 analysts were expecting. The stock dropped nearly 8% in after-hours trading.

Over in Redmond, Microsoft was telling a completely different story. The company posted $90.0 billion in revenue, up 18% year-over-year, beating on both the top and bottom lines. Azure crossed $100 billion in annualized revenue for the first time. The stock — which had been trading near a one-year low ahead of earnings — held firm.

Same day. Same market headwinds. Same AI capex questions. Completely different outcomes.

The Meta Story — Revenue Is Growing but Earnings Are Shrinking

Let me start with Meta because the numbers tell a story that should scare anybody who's been buying the "AI capex is always good" narrative.

Meta's revenue grew 28% year-over-year to $60.80 billion. That's real growth. The family of apps now reaches 3.60 billion daily active people. Ad impressions are up 14%. Average price per ad is up 12%. On the surface, the machine is working.

But net income dropped 13% to $15.85 billion. That's a $2.5 billion gap between what they earned last year and what they earned this quarter, on $13 billion more revenue. The difference? AI infrastructure spending. Meta is on track to spend $135 to $145 billion on capex this year, most of it on AI data centers, GPUs, and the power to run them. And that spending is compressing margins faster than revenue can keep up.

Mark Zuckerberg said all the right things on the call — "AI is accelerating our core business," "the results are already showing," "optimistic about the potential ahead." And I believe him. I really do. But $6.18 versus $7.13 is not a rounding error. That's a 13% earnings decline in a quarter where revenue grew 28%. The market noticed.

Meta narrowed its capex guidance range, which the bulls are spinning as "discipline" and the bears are reading as "we're not sure we can keep this up." I think it's somewhere in between — but the direction matters more than the range.

The Microsoft Story — Azure Crossed $100 Billion and Nobody Cheered

Now let me tell you about the other side of the coin. Microsoft's Q4 was genuinely impressive by any traditional measure. $90 billion in revenue, 18% growth, Azure hitting $100 billion annualized for the first time. Satya Nadella has turned Microsoft into a machine that prints enterprise money.

But here's the thing that kept me up — the stock was trading near a one-year low BEFORE earnings. One year. A company growing 18% with $90 billion quarterly revenue and Azure crossing $100 billion. And the stock was at a 52-week low.

Why? Because the market is already pricing in the capex risk. Microsoft's AI infrastructure spending is embedded in every Azure data center, every GPU cluster, every power purchase agreement. And the market is starting to ask the question nobody wanted to ask six months ago: "What happens when the growth in AI revenue doesn't keep up with the growth in AI spending?"

Microsoft beat this quarter. Azure is still a monster. But the stock didn't pop the way it would have a year ago. The market's reaction function has fundamentally changed. Beats are now met with skepticism. Capex raises are met with selloffs. The honeymoon phase of the AI trade is over.

I watched the Bloomberg live reaction stream — the analysts on the call were asking about "returns on invested capital" and "payback periods" in ways they simply weren't six months ago. The tone has shifted, and it's not going back.

The Counter-Argument — Both Companies Are Still Investing Like There's No Tomorrow

Before I get too bearish, let me acknowledge the other side because it matters. Combined, Meta and Microsoft are on track to spend roughly $375 to $425 billion on capex between them this year. That's not a typo. Somewhere between $375 billion and $425 billion. In one year. On infrastructure.

The bulls will tell you this is rational because AI compute demand is still outstripping supply by a wide margin. Nvidia's Blackwell, AMD's MI450, all of it is still allocation-constrained. Data center lead times are 24 to 36 months in most markets. If you're not building now, you're losing market share for the next three years.

I get that argument. I've made it myself in at least six articles over the past two weeks. But the problem is that both things can be true at the same time: AI demand is real AND the pace of spending is creating a balance-sheet vulnerability that will show up in the next downturn.

Meta's $6.18 EPS miss is not a one-quarter phenomenon. It's the first signal that the margin compression from infrastructure spending is hitting the P&L in ways that revenue growth can't fully offset. When your revenue grows 28% and your net income drops 13%, you have a cost structure problem — not a revenue problem.

The Bigger Pattern — Today Was the Market's Reckoning With AI Capex

Here's what keeps me up at night. Google reported on July 22 with $119.8 billion in revenue and 82% Cloud growth. The stock dropped 4% because they raised capex to $195-205 billion. Meta reports a $60.80 billion quarter with 28% revenue growth. Stock drops 8% because EPS missed on AI spending. Microsoft reports $90 billion with Azure crossing $100 billion. Stock barely moves because everyone's already priced in the capex risk.

Three companies. Three quarters that any of them would have been proud of five years ago. And the market punished or ignored all three. The only thing these earnings calls are now about is one question: "When does the AI spending start generating returns that justify the spending?" And nobody has a good answer yet.

The Economist ran a piece two days ago titled "AI revenues are growing fast, but not fast enough." That's the whole story in one headline. AI is real. AI is growing. AI is transforming businesses. But it's not growing fast enough to justify $725 billion in combined Big Tech capex this year.

This is the structural tension that defines the entire AI infrastructure story right now. The demand is real. The buildout is real. But the financial math doesn't work at this scale without something giving — either revenue acceleration that hasn't materialized yet, or a capex slowdown that nobody wants to admit is coming.

What This Actually Means for Independent Hosting Providers

Now let me tell you what this means for people running actual hosting businesses, because the market's drama in Menlo Park and Redmond has real consequences for your pricing, your supply chain, and your competitive positioning.

First — watch Meta's capex guidance as a leading indicator. Meta narrowed its range this quarter. If they cut it next quarter, that's the first domino. Microsoft, Google, and Amazon will follow. A capex slowdown at the hyperscaler level means hardware oversupply in 6-12 months as pre-ordered GPUs and servers hit the secondary market. Be ready to buy when that happens, not after.

Second — the "AI premium" on colocation pricing has a ceiling. If hyperscalers start pulling back on expansion plans, the supply-demand imbalance in the colocation market will ease. Don't sign multi-year deals at peak pricing right now. Negotiate shorter terms with rate review clauses. The market dynamics are shifting under your feet.

Third — Microsoft and Meta's divergent results tell you the market is now discriminating between AI winners and spenders. Microsoft has Azure generating real AI revenue. Meta has ad targeting improvements from AI but no direct AI revenue stream to point to. The market punished the one without a clear monetization path. If you're building your hosting business around AI workloads, make sure you can show your customers a clear ROI — because your bank and your investors will ask the same question.

Fourth — this earnings season is going to shake loose hardware. When Meta and Microsoft pre-ordered $375+ billion worth of infrastructure, they locked up GPU supply for years. If they slow down even 10%, that's tens of billions of dollars worth of hardware that needs a home. Independent hosting providers who have relationships with hardware distributors are going to find deals in Q4 2026 and Q1 2027 that didn't exist in Q1 2026.

The Structural Reality — The Buildout Continues, But the Terms Are Changing

I want to be clear about something because I've seen people misinterpret my takes on this. I am not saying the AI buildout is over. I am not saying data centers are going to stop getting built. I am saying the terms of the buildout are changing, and if you're running a business that depends on understanding those terms, you need to pay attention to what yesterday's earnings calls actually revealed.

What they revealed is that the market has reached peak tolerance for AI spending without corresponding AI revenue. From here on out, every dollar of capex is going to be scrutinized. Every earnings miss is going to be blamed on AI infrastructure costs. Every guidance cut is going to signal a capex slowdown.

That doesn't mean the buildout stops. It means the buildout becomes more rational. More measured. More focused on projects with clear ROI. And for independent hosting providers, that shift from "spend everything" to "spend wisely" is actually an opportunity — because you've been running lean and efficient your whole career.

This is where the capital-light model wins. When the hyperscalers are fighting over $200 billion capex plans and getting their stocks punished for it, the hosting provider who can deliver the same service at a fraction of the infrastructure cost starts looking a lot more attractive to customers who are also getting squeezed by AI pricing.

The Bottom Line

Meta missed by a dollar. Microsoft crushed it. Both reported on the same day, and both told the same story: AI capex is real, AI demand is real, but the financial math is getting tighter by the quarter. The market's patience is not infinite, and last night proved that the honeymoon phase of the AI trade is officially over.

For independent hosting providers, this shift is your window. The next six months are going to decide which way the AI infrastructure story goes — acceleration or correction. And unlike the hyperscalers, you don't need a $200 billion capex plan to win. You just need to be paying attention, be ready to move when the signals shift, and never forget that in this business, the ones who survive are the ones who see the turn before everyone else does.

I saw the turn last night. I hope you did too.

— Allan Ali, Founder