We heard from Google on Q2 last week.

They spent $44.9 billion on capital projects in three months, produced negative free cash flow of $5.9 billion, and raised the 2026 spending plan to $195 to $205 billion.

This week we hear from Microsoft and Meta on Wednesday. Amazon and Apple on Thursday.

Together with Google, these four are on track to spend about $700 billion this year, against roughly $410 billion last year. Wall Street expects the number to approach a trillion dollars in 2027.

This shouldn't be news to anyone who has been paying attention.

Remember, last October, the man who supplies the most advanced AI chips in the world said he could see half a trillion dollars of demand on the books through 2026.

By March he had doubled it. At least a trillion dollars through 2027.

And he went further, saying he was certain computing demand would run higher than that. His words on supply were plainer still: "we are going to be short."

Then in the May earnings call the Nvidia CFO projected three to four trillion dollars a year in AI infrastructure spending by the end of this decade.

The numbers from the companies selling access to that compute keep backing it up. Google reported extraordinary demand last week. Cloud revenue grew 82%. Contracted backlog reached $514 billion.

So why is Wall Street wringing its hands about capital spending?

Keep in mind, these are companies producing tens of billions of dollars in operating cash flow, quarter after quarter, and putting that cash into capacity that is already sold.

But what about the debt raises?  Not only are they spending their operating cash flow, now they're borrowing. 

When demand is growing faster than you can fill it, and the operating profit on that capacity covers the debt service roughly nine times over, levering the balance sheet is exactly what a shareholder should want.

And that kind of operating leverage is normally rewarded.

But Wall Street seems worried about the spending. Worried about the borrowing.

The only reason to worry about either is if you are worried about demand.

But they don't seem to be worried about that.

Why?

Because agentic AI means model usage is multiplying by the day, and has been since February, when the agentic moment arrived.

Now add the proliferation of open source models, now at or near the level of the best in the world — and demand for compute goes up, not down.

More models, running in more places, doing more work. That is the justification for more capacity, arriving just as the companies best equipped to build and supply it are spending more and planning to spend more — as they should be.