AI Capex in 2026: $828 Billion of Spending and a Cash Flow Problem

AI capex is the number the market is now trading on. Amazon, Alphabet, Microsoft, Meta and SpaceX, which owns xAI, have outlined as much as $828 billion of capital spending for 2026, according to Fortune, most of it for data centers, chips and the power to run them. Without SpaceX, the four cloud giants alone account for about $740 billion. On October 30 the market finally asked how all of it gets paid for, and tech stocks lost more than $1 trillion in a day.

This piece lays out the spending, the cash flow behind it and the shift to borrowing that happened this year. For what happened in the market that day, see our news report on the selloff.

How much each company plans to spend

Amazon leads with about $220 billion, a figure it reaffirmed this week while saying it expects to spend meaningfully more in 2027. Alphabet has guided up to $205 billion and also expects spending to rise next year. Microsoft is around $170 billion, Meta up to $145 billion and SpaceX up to $88 billion. Several banks, including Evercore and Bank of America, expect AI-related capex to reach $1 trillion in 2027, Fortune reported.

The spending is front-loaded by design. Andy Jassy, Amazon's chief executive, told analysts that it might be more convenient to spread the investment over 25 years, but that is not the opportunity in front of the company.

The cash flow problem

Capital spending comes out of operating cash flow first. What is left is free cash flow, the money a company can use for buybacks, dividends or a rainy day. For most of the AI boom, the giants had plenty. This quarter, that cushion disappeared at most of them.

Alphabet's free cash flow was negative $9.1 billion, its second negative quarter in a row, after $51.4 billion of capital spending in three months. Meta posted negative $2.1 billion, against a positive $10.6 billion a year earlier and its first negative quarter since 2012. Amazon's free cash flow over the past twelve months was negative $42.4 billion. Microsoft is the exception at a positive $18.6 billion, helped by spending $45.2 billion in the quarter, less than the $50 billion it had projected.

-$42.4B

Amazon free cash flow, last 12 months

+$18.6B

Microsoft, the only one still positive

$326B

Debt raised by five AI spenders in 2026

3x

That debt versus 2025

Borrowing fills the gap

When cash flow cannot cover the plan, companies borrow. Microsoft, Alphabet, Amazon, Meta and Oracle raised a combined $326 billion of debt in 2026, roughly triple the amount they raised last year, according to S&P Global Market Intelligence data cited by TechCrunch. Microsoft priced a $25 billion bond this week, its first bond sale in more than nine years, and said the money could go to capital spending. Meta announced a plan to raise up to $95 billion by issuing new shares, Fortune reported, and Oracle does not expect to return to positive free cash flow until the second half of the fiscal year that ends in May 2028.

Debt is not a problem in itself. These are some of the strongest balance sheets in the world, and borrowing to build long-lived assets is normal. What changes is the risk. A company funding data centers from spare cash can slow down quietly if demand disappoints. A company funding them with bonds and new shares has made promises to lenders and shareholders that assume the demand shows up.

Is the revenue keeping up?

The bull case is that the spending is chasing real demand, and the cloud numbers support it. All three big cloud businesses grew quickly in the latest quarter.

The trouble is that cloud revenue is growing from a much smaller base than the spending. And the businesses that fund the build-out are slowing in places. Google's search advertising growth fell to 12%, its weakest in five quarters, while Meta's ad revenue still grew 26%. Investors are weighing fast-growing revenue that is small against slow-growing revenue that pays the bills, and on October 30 they decided the gap was too wide.

How the market has repriced the AI basket

The AI stocks the AI Bubble Index tracks have already moved a long way. Over the past month, the chip makers and the companies with the widest gap between spending and cash flow fell hardest, while Microsoft, the one positive cash flow story, held up.

Even so, the index itself sits in its Neutral band at 48.6. Valuation, the most stretched component at 85, still says these stocks are expensive relative to their earnings. Market concentration, at 9, says the gains are not piled into a handful of names. The full method is on the methodology page.

Does this look like 2000?

The closest historical rhyme is not the dot-com stocks themselves but the telecom build-out underneath them. In the late 1990s, carriers borrowed heavily to lay fiber for traffic that was supposed to arrive soon. It did arrive, years later, after many of the companies that built the network had gone bust. Our dot-com bubble timeline walks through how that played out.

The differences matter just as much. The companies spending today are among the most profitable ever, and most of them are funding the build-out from businesses that still grow. Microsoft's positive cash flow shows it is possible to invest heavily and stay self-funding. The risk is not that these companies fail. It is that the return on $828 billion a year turns out lower than prices assume, and that the market reprices that gap all at once, the way it started to on October 30. Our guide to AI investment risks covers how to think about that exposure.