Every bull case for this market runs through one trade. AI pays off.

So let’s look at the ledger.

The big five will spend roughly $600 billion on AI infrastructure this year, per their own guidance. Published run-rate estimates put AI revenue at about $60 billion. That is a $540 billion gap between one year’s spend and one year’s earn. Sequoia Capital has tracked this gap since 2024, when it was $125 billion. It has quadrupled.

Bain & Company did the forward math in their Global Technology Report. To fund this buildout profitably, AI needs about $2 trillion in annual revenue by 2030. That is 33 times today’s number. In four years.

Fine. Big numbers everywhere in this business. Here is the question nobody asks.

Who Writes the Check?

$2 trillion a year has to come from somewhere. There are only three wallets on the planet that could plausibly pay it. Let’s open each one.

Corporations. The most likely payer. But Gartner puts the entire global software market at $1.44 trillion for 2026. Every license, every SaaS seat, every database on earth. The AI bill is bigger than all of it combined.

Advertisers. The Google and Meta playbook. But Dentsu says total global ad spend crosses $1 trillion for the first time this year. Every billboard, every Super Bowl spot, every search ad on the planet. AI would need to swallow that market twice.

Consumers. The smallest wallet by far. The entire global video streaming industry, every Netflix and Disney+, every subscription in existence, runs about $150 billion a year. Consumer AI would need to be many multiples of all streaming combined just to carry a fraction of the load.

For scale: $2 trillion is roughly 60 percent of what every American household combined spends on housing in a year. That is the size of the revenue stream this buildout requires. Annually.

There Is No Line Item Big Enough

Which leaves exactly two paths, and both of them matter to your portfolio.

Path one: cannibalization. AI eats existing budgets. Sounds orderly. It is not. Cannibalized revenue is not new money. It is revenue torn out of someone else’s income statement. SaaS companies. Media companies. IT services firms. Ad agencies. And markets do not price lost revenue dollar for dollar. They capitalize it. Software trades at six to eight times sales. Redirect $2 trillion of revenue and you vaporize north of $10 trillion in market cap from the losers. Many of those losers sit in the same index as the hyperscalers doing the spending. In this path, AI does not lift the market. It hollows it out from the inside, and the crater is bigger than the prize.

Path two: creation. Productivity gains so large that companies mint entirely new budgets. That is the bull case, and it might happen. But understand what you are pricing. Not a product cycle. The largest productivity miracle in human history, on a four-year deadline.  I don’t know about you, but this sounds about like my neighbor saying he is going to rebuild the Great Pyramids over the weekend.

What This Means for Allocators

I am not calling a top. Capex is someone else’s revenue, and the spending could keep markets happy for years.

But ask yourself one question. If you own the index, you own both sides of this trade. The companies writing the checks and the companies whose revenue funds them. Which path is your portfolio pricing? Because one path requires a miracle, and the other one detonates inside your own equity book.

These figures are published estimates and revise often. The caution cuts both ways.

But the ledger is on the page. Somebody has to balance it.