# data center revenue — X 热门讨论 (2026-09-25 03:38 UTC)

## @LeoNelissen (Leo) · 09-24 22:44 · ♥55 ↻16 💬6 I don’t think investors fully understand the scale of what is happening here.

👉 Based on a WSJ story

The AI buildout is on track to become the biggest economic bet in U.S. history. Not the internet. Not highways. Not electrification. Not even the railroad boom.

AI.

The WSJ estimates that U.S. data-center and AI infrastructure investment could reach $10.3 TRILLION between 2025 and 2032. That works out to roughly 3.6% of GDP per year.

For perspective, the railroad boom averaged about 2.2% of GDP. Highways were roughly 1.1%. Telecom and fiber were also around 1.1%.

That alone is incredible.

But I think the charts underneath the headline are even more important.

The five hyperscalers at the center of this buildout - $GOOGL, $AMZN, $META, $MSFT and $ORCL - are expected to spend roughly $4.2 trillion in the four years ending in 2029.

That is no longer normal corporate CapEx.

It is large enough to reshape the economy around it.

And you can already see that happening in construction.

Through July, private U.S. data-center construction spending was running about $9 billion above the same period last year.

Meanwhile, private construction spending on basically everything else - houses, apartments, shopping centers and more - was about $46 billion BELOW year-ago levels.

That chart is remarkable.

Data centers keep going up while almost everything else rolls over.

To me, that is where this becomes much more than an AI story.

Hyperscalers are now competing with the rest of the economy for electricians, construction workers, transformers, turbines, land, natural gas, water and power.

The WSJ gives a perfect example.

Mississippi was in the running for an aluminum smelter that could have created roughly 1,000 permanent jobs. Then a data center was announced near one of the proposed sites and tied up electricity the smelter needed.

The smelter went to Oklahoma instead.

That is what crowding out looks like.

And it is not just power.

Data centers are also pushing up land costs, pulling skilled labor into the buildout and creating shortages across parts of the equipment supply chain.

Then there is inflation.

Import prices for computers, peripherals and semiconductors were roughly 20% higher year-over-year in August.

That matters because the AI boom is not happening in some isolated corner of the economy.

It is bidding up the price of labor, equipment, electricity and capital at the same time.

And then there is the wealth effect.

U.S. households now own roughly $63 trillion of stocks and mutual funds, nearly double the amount at the end of 2022.

Think about the feedback loop here.

$GOOGL, $AMZN, $META, $MSFT and $ORCL spend trillions on AI infrastructure.

That spending boosts construction, wages and demand for equipment.

AI-related earnings expectations push stocks higher.

Higher stocks increase household wealth.

That wealth supports consumption.

Meanwhile, hyperscalers keep borrowing and spending to build even more infrastructure.

That is an incredibly powerful cycle.

But it also creates a risk that I think the market may be underestimating.

The more the U.S. economy depends on one massive investment cycle, the more painful it becomes if that cycle ever slows.

And a growing share of this buildout is being financed with debt.

That is why I think this is much bigger than an AI bubble debate.

The question is no longer just whether AI generates enough revenue to justify Nvidia chips or data-center leases.

The question is what happens to the broader economy if the biggest infrastructure boom in U.S. history suddenly loses momentum.

We are not watching another software cycle.

We are watching the construction of a new layer of the U.S. economy.

And increasingly, I think one of the biggest mistakes investors can make is treating AI as just another sector.

AI is becoming the economy. https://x.com/LeoNelissen/status/2103254288030183768