The Productivity Mirage: AI Spending Without the Returns

The earnings season script has become grimly predictable. Tech executives spend forty minutes describing their artificial intelligence roadmaps, board members nod sagely, and the market prices in salvation. SPY closed down 0.27% last week anyway. Broad equity weakness. The narrative still dominates—AI will solve everything, they promise. The data suggests otherwise.

Corporate capex is surging. Tech firms are pouring billions into GPU clusters, data centers, and algorithmic infrastructure on the assumption that productivity gains will materialize at scale. This is not inherently irrational. Productivity gains would solve their actual problem: margin compression in a synchronized global slowdown.

But here's the difficulty. Wage growth remains elevated across the OECD despite labor markets cooling and growth signals softening. In the U.S., average hourly earnings growth is running above trend. In Germany, labor negotiations are occurring against the backdrop of stagnation. In the UK, wage growth remains sticky even as services inflation cools. Companies are caught between two price points: what they must pay workers and what customers will pay for goods and services. That second number is shrinking.

This is the late-cycle bind. Historical pattern suggests what happens next. When corporate capex surges without corresponding demand growth, margin compression follows. The investment has to be justified against revenue that isn't materializing. Productivity gains from AI adoption are still theoretical. Actual productivity data from the past two quarters remains pedestrian. We are not seeing the inflection point.

What we are seeing is companies making a bet. They're spending on the assumption that AI will eventually unlock margin expansion when labor hours decline and output scales. It's a rational gamble in a vacuum. But vacuums don't exist in macroeconomics. Synchronized slowdown is not a vacuum. It's the environment in which capex cycles historically deteriorate into write-downs.

The equity market knows this on some level. Hence the 0.27% decline despite AI enthusiasm dominating earnings commentary. Investors can hear the internal contradiction: we're investing massively in productivity tools while our ability to price goods is evaporating and our wage bills remain stuck.

Consider the timeline. Companies are making multiyear capex commitments now, in Q4 of a cycle where growth is visibly moderating. Demand in developed markets is soft. China's property sector remains a weight. The synchronized global slowdown is not speculation—it's observable in PMI data, consumer confidence indices, and capex intentions themselves. And yet capex is accelerating anyway, driven by the AI narrative.

When the productivity gains don't materialize on the timeline promised, we'll see the familiar reversal. Not immediately. First, guidance will soften. Then, capex guidance will be reduced. Then, the write-downs on underutilized infrastructure will begin. None of this is forecasting. It's pattern recognition.

The mirage is not artificial intelligence itself. AI will probably do useful things eventually. The mirage is the assumption that capex cycles can occur divorced from demand cycles, and that margin expansion is algorithmic rather than economic. It isn't. Margins expand when demand exceeds supply. Right now, we're seeing the opposite: supply-side investment racing ahead of demand growth, driven by narrative rather than fundamentals.

Wages will stay sticky because labor markets, despite softening, remain relatively tight in skilled sectors. Pricing power will remain constrained because customers are becoming more price-sensitive in a slowdown. And capex will continue because executives have staked their credibility on the AI thesis. That's a recipe for margin compression, not expansion, and equity markets are pricing that risk at the margins. Hence the 0.27%.

The productivity mirage won't disappear until either demand accelerates or capex stops. Neither is likely in the next two quarters.