Big Four AI CapEx in One Year: $300B–$500B and a Build-Out Bigger Than Most Countries' GDP

Big Four AI CapEx in One Year: Bigger Than Most Countries' GDP

The AI infrastructure spend has crossed into a zone where ordinary comparisons stop working. Amazon, Meta, Microsoft, and Alphabet — the Big Four — are projected to deploy somewhere between $300 billion and $500 billion on AI infrastructure in a single year.

To frame how absurd that scale is: four companies, one technology, one fiscal year, more than the GDP of plenty of mid-sized economies. We have rarely seen capital concentrate this fast into a single category in modern history.

Per-Company Commitments — The Scattered Numbers in One Place

What I want to do here is what most coverage doesn't bother to do — line up the numbers in one frame.

Company	1-Year AI Infra Commitment	Note
Amazon	~$200B	Defended directly by CEO Andy Jassy; consistent with their reinvestment DNA
Meta	$10B for one El Paso facility + $21B with Coreweave	Multi-year figures cited as high as $600B
Microsoft	$80B in a single fiscal year	Tied to OpenAI partnership + Azure
Alphabet	Pacing similar to Microsoft	Google Search + YouTube AI integration

What Is a Data Center, Really

For readers without a technical background — a data center is a giant warehouse stuffed with extremely powerful computers. Every prompt to ChatGPT or Claude, every AI process a business runs, every model training run, all of it happens inside data centers.

That's why the Big Four are racing to build them as fast as they physically can. They believe AI compute demand will keep growing for a long time. Amazon's Jassy has repeated this exact message to shareholders, and his peers are saying the same thing in slightly different words.

When Wall Street's Applause Stopped

For a while, the market loved this spending. More AI infrastructure equaled a wider moat, a stronger competitive position. Every earnings call where a CEO announced higher AI CapEx, the stock would rally. The market was rewarding ambition.

Then the mood shifted. Investors started running a simple calculation. You spent $200 billion. When do you get it back? Fair question. Spending money is easy. Turning that spend into profits that justify the bet — especially in a brand-new technology — is much, much harder.

What I'm Watching — The FCF Story Underneath

CNBC reported recently that Wall Street is genuinely uneasy about the size of these commitments, and the discomfort shows up in free cash flow. Operating cash minus CapEx is falling at all four companies, because so much cash is going out the door to build infrastructure.

That's producing a strange picture in the market right now. Costco and Walmart — businesses with lower growth and lower returns on capital — are trading at higher multiples than Microsoft or Meta. The market is paying a premium for predictable cash today and discounting promises of cash tomorrow.

What to Watch Next

Two things sit at the top of my watchlist.

First, whether demand catches up with supply. All four are building at the same time. Will there be enough AI workload to fill that capacity once everything is online? The fiber optic boom of the late 1990s left the ground full of unused cable, and that comparison keeps tugging at me.

Second, the efficiency curve. We are starting to see examples of powerful AI models built on far less compute than previously assumed. If that holds, some of the data centers being built today may sit underutilized.

Answers to both questions will start arriving over the next two to three years. They will determine which of the Big Four ends up as the real winner — and which one ends up explaining why they overspent.