Remember when everyone said AI was a bubble? The “it’ll never make money” crowd had a pretty good run. But here’s the thing about bubbles — they tend to pop when actual revenue shows up.
And boy, has it shown up.
The latest data from Exponential View just dropped the mic on the bear case: the global AI economy is churning out $175 billion in annualized revenue. Not projected revenue. Not “if everything goes perfectly” revenue. Real, actual customer money flowing in right now.
Now, $175 billion sounds massive — and it is. But here’s where it gets interesting: it’s only 0.5% of U.S. GDP. The entire digital economy is 10% of GDP. So AI is still basically a rounding error in the grand scheme of things. Which means there’s a *lot* of runway left.
But the real story isn’t the size — it’s the speed. The AI economy is scaling 3x faster than the internet and mobile booms. In 2023, it took 180 days to add $1 billion in cumulative revenue. Today? Less than two days. That’s a 90x acceleration. If that doesn’t make your brain hurt a little, you’re not paying attention.
**The CapEx Math Actually Works Now**
The bears’ favorite argument was always the same: “Hyperscalers are spending $2 trillion on infrastructure, and there’s no way the economics pencil out.” Fair point, right? Except the math is starting to work.
For every dollar of AI infrastructure that depreciates, roughly $1.19 in revenue is coming in to cover it. A year ago, that ratio was below 1. Now it’s above it. We’re not talking about massive profits yet — but we’ve crossed the line from “this will never work” to “this is actually working.”
That’s a direction change. And in markets, direction matters more than destination.
**Why Cheaper AI Is Actually Bullish**
Here’s where the sophisticated bears make their move: token prices are collapsing. Blended pricing has fallen from $17 per million tokens to $2. Margins are going to zero, they say. The boom is over.
Except they’re confusing price with value. When prices fall on elastic demand, you don’t get less consumption — you get *more*. Cheaper tokens unlock new use cases. More apps, more agents, more inference, more everything. It’s the Jevons Paradox playing out in real time: efficiency improvements lead to increased total consumption.
The bears are worried about price compression. The bulls are focused on volume elasticity. The data says volume wins.
**The Real Play**
Here’s the thing: 70% of AI claims from S&P 500 companies are about cost savings and efficiency, not new revenue streams. That’s actually normal for platform shifts. The internet’s first decade was all about cost reduction. Revenue came later — and when it came, it was enormous.
AI is following the same playbook. Efficiency now. Revenue later. And if the efficiency wave alone is already generating $175 billion in demand, imagine what happens when the revenue wave actually hits.
The infrastructure stocks — chips, memory, networking, servers, power, cooling — are the picks and shovels play. The fundamentals are inflecting positively, and the market is handing you a discount.
That doesn’t happen often.
