The AI Spending Death Spiral Nobody Wants to Admit Exists

Or: How Wall Street Became a Cargo Cult

---

Let's talk about what actually happened this week in the markets, stripped of the bullshit.

Microsoft reported Azure cloud sales up 39% year-over-year. Quarterly revenue beat by a comfortable margin. Adjusted EPS came in at $4.14 against expectations of $3.97. By every traditional measure, this is what "beating earnings" looks like. The stock immediately fell 5% in after-hours trading.

Why? Because Microsoft disclosed that capital expenditure hit a record $37.5 billion in the quarter. Let that number sit for a second. Thirty-seven and a half billion dollars. In one quarter. On AI data centers that don't yet have a clear path to profitability.

Meta, meanwhile, actually managed to exceed expectations—$8.15 EPS versus $8.08 forecast, $58.4 billion in revenue with 20.7% year-over-year growth. The stock spiked 10% at one point because investors were relieved the company didn't accidentally crater its margins. Relieved. At beating what they were supposed to beat.

On Thursday, Apple reported what should have been the earnings report of the week: $143.756 billion in revenue, up 16% YoY. iPhone sales of $85.269 billion—the best-ever quarter for iPhones, per Tim Cook himself. Services revenue broke $30 billion, a new record. EPS of $2.84 beat estimates by $0.17. And yet here we are, with Apple still taking it on the chin.

The Fed paused rates at 3.5-3.75%. Completely expected. Powell delivered carefully hedged remarks about being "well positioned" and letting "the data speak." Translation: We're not cutting, we're not hiking, we're just sitting here while the most consequential economic policy question of our time—whether we can fund a $470+ billion AI infrastructure buildout without blowing the whole joint up—remains entirely unresolved.

---

Here's what's actually happening, and I want to be precise about this because the financial media sure isn't.

We are living through a capital allocation crisis disguised as an earnings beat. The hyperscalers—Microsoft, Meta, Alphabet, and Amazon—are expected to deploy $470+ billion in capex this year, up from $350 billion in 2025. That's 34% growth in spending on infrastructure that, by definition, must achieve ROI at scales we've never seen before. Not because these companies are greedy or irrational, but because that's what artificial intelligence requires.

The problem is accounting. When Microsoft spends $37.5 billion on servers and you show up as a shareholder expecting to see that reflected in profit margins, you're going to be disappointed. The moment a company starts capitalizing these costs versus expensing them, they buy themselves a temporary reprieve—but it's a game you can only play for so long before the amortization schedule catches up to you and strangles your operating margins.

Wall Street understands this at some level. This is why the market's response to "we beat earnings but we're spending more on AI" has evolved from euphoria to anxiety to something approaching dread. Meta was the first to get the memo last October and watched the stock crater. Yesterday and today, it was Microsoft's turn. The company literally did everything right—the cloud business is humming, enterprise customers are locking in and spending—and the response was still "sell."

This is the trap. These companies cannot stop spending. Stop spending on AI, and you become the dinosaur in five years that missed the inflection. Keep spending, and you're cannibalizing near-term profitability on what amounts to a bet that the monetization will materialize. Except the monetization models are still unclear. You know what's clear? The costs.

---

Meanwhile, the Fed is sitting in neutral, which is banker-speak for "we have no idea what we're doing and we're hoping something happens that clarifies the situation." Powell spoke to "tariff inflation topping out" in the middle quarters of 2026. He acknowledged that inflation remains elevated. He noted that the labor market is stabilizing but hiring is still weak.

What he didn't do was commit to anything. One interest rate cut projected for all of 2026—and that's not happening until June, probably, after someone else is chair.

This is significant because the Fed's implicit thesis—that the economy can sail forward at 3.5% growth with inflation above 2.5% without needing more stimulus—is built on some pretty heroic assumptions. Chief among them: that the private sector will continue deploying hundreds of billions in capex on pure faith.

The S&P 500 is down, the Nasdaq got slammed 1.3%, and the Russell 2000 actually outperformed because small-cap stocks are mostly domestic businesses that don't have to justify $37.5 billion quarterly AI spending to a skittish shareholder base. Southwest Airlines pops 16% after announcing it's charging people for bags and calling it a business model improvement. That's the market we're in now.

---

Here's the uncomfortable truth nobody wants to write down plainly: We're in a situation where the fundamental unit of analysis—the quarterly earnings report—has become almost useless for understanding what's actually happening in corporate America.

You can beat earnings, grow revenue 15-20% YoY, achieve record profitability in your legacy business, and still get sold because you dared to spend like the future is actually coming. This is the opposite of how markets are supposed to work. Investors are supposed to care about long-term value creation. Except right now, they're running from it.

Meta understood this faster than everyone else. The company basically told Wall Street: "We're going to massively increase capex and we don't care if it makes you nervous." The stock tanked in October. Then it just... stabilized. Then yesterday, when it beat and guided to more spending, it rallied 10% because the market had already priced in the disaster and moved on.

Microsoft is going through the exact same arc right now, compressed into a single after-hours sell-off. Within 24 hours, you're going to have three calls from Goldman and two from Morgan Stanley explaining why the stock is "attractively valued." The narrative will shift from "AI spending is a problem" to "this is a strategic investment in future competitive positioning."

The stock will recover. It always does. Because the alternative—that these companies are making a colossal mistake—is too uncomfortable for asset managers who've built careers on the premise that you should own mega-cap tech.

---

The real story this week wasn't the earnings. It wasn't the Fed pause, though that's consequential. It was the moment when the market had to confront the fact that growth and profitability might be diverging for the first time since the cloud computing boom, and that the cloud companies themselves are the ones engineering that divergence.

Investors discovered they have two choices: Trust that AI ROI materializes as promised, or sit on the sidelines while everyone else builds the future. For most of them, there is no third option. The alternative to buying the Magnificent Seven is... what, exactly? Small caps that don't have capex discipline because they can't afford to? Bonds paying 4.15% in a world where inflation is 2.8% and structural forces are everywhere?

Gold hit $5,463.70 and silver rallied 3.2%. You know what that means? Some very smart people are starting to hedge. Not panicking, not selling everything. Just quietly asking whether there's a version of this story where we get inflation without growth, where AI capex becomes a sunk cost, where Powell's successor has to choose between supporting the dollar and supporting tech stocks.

The earnings didn't fail. The earnings were fine. It's the underlying story that's becoming impossible to ignore: We're betting the future on things that don't make money yet, and we're doing it with borrowed time and the assumption that the Fed will stay supportive.

What happens when the market stops believing that?

That's the question nobody asked in the press conference Wednesday. Powell, mercifully, wasn't forced to answer it.

---

Next week: Alphabet and Amazon report. Brace yourself.