Since prior coverage documented the grid stress driving the NextEra-Dominion merger and private credit contagion tightening buyout financing, the AI spending machine that underlies all of it has only accelerated — and the Q1 2026 earnings numbers put a hard figure on it.

$130 Billion in Three Months

Amazon, Alphabet, Microsoft, and Meta collectively spent $130.65 billion on capital expenditures in the first quarter of 2026, according to the New York Times. That is 71 percent higher than the same quarter one year earlier. It costs more than three times what the Manhattan Project spent developing nuclear weapons, adjusted for today's dollars.

All four companies said they would spend significantly more in the year ahead.

Alphabet Goes Bigger

Alphabet upsized its full-year 2026 capital expenditure target to at least $75 billion, according to Bloomberg. The company added that spending would be "significantly" higher in 2027. Meta CEO Mark Zuckerberg raised his company's 2026 forecast to between $125 billion and $145 billion, up from a prior range of $115–135 billion. "Every sign that we're seeing in our own work and across the industry gives us confidence in this investment," Zuckerberg told investors on an earnings call, according to the Times.

The Full Five-Company Picture

Zoom out to the five largest U.S. cloud and AI infrastructure players — Microsoft, Alphabet, Amazon, Meta, and Oracle — and the combined 2026 capex commitment sits at $660–690 billion, according to Futurum Group analyst Nick Patience in a February 12, 2026 analysis. Amazon alone projected roughly $200 billion. Microsoft is tracking toward $120 billion or more. Oracle is targeting $50 billion.

Fortune's AI reporter Sharon Goldman put the combined four-hyperscaler figure above $700 billion for the full year, up sharply from approximately $410 billion in 2025.

The Revenue Side Is Real — But Nowhere Near This Scale

On the AI model side, OpenAI ended 2025 with approximately $20 billion in annual recurring revenue — a threefold increase from the prior year. Anthropic's revenue run rate crossed $9 billion in January 2026, up from roughly $1 billion at the end of 2024, per Futurum. Those are strong growth numbers.

The five major infrastructure spenders are committing $660–700 billion this year. OpenAI and Anthropic combined are generating roughly $30 billion in annualized revenue. The gap between spending and current AI software revenue is substantial — unless the hyperscalers themselves are generating the returns through cloud, advertising, and enterprise services that subsidize the infrastructure bet.

They are. The four biggest companies generated $431 billion in combined sales and $151 billion in profit in Q1 2026 alone, according to the Times. They can absorb losses for some time. The question is how long the margins hold if monetization lags.

The Power Problem Is Getting Worse

Bloomberg reported Duke Energy CEO Lynn Good saying AI is fueling power demand growth at 10 times the historic pace in the company's service territory right now.

This connects directly to the $66.8 billion NextEra-Dominion merger covered in prior reporting — the grid cannot keep up with data center demand at this pace. AI capex is no longer just a tech-sector story. The infrastructure strain is spreading.

The Muni Market Is Now Involved

Bloomberg reported that the AI funding boom has reached the municipal bond market, with a Google-tied deal bringing AI infrastructure financing into a corner of finance that normally funds schools and highways. When AI capex starts getting structured through muni deals, the exposure spreads beyond institutional investors and hyperscaler balance sheets into instruments held by pension funds and retail investors.

Private credit is stressed. Private equity is gating funds. The muni market is being recruited to help finance data centers. The financing web keeps expanding.

What the Coverage Is Missing

Most mainstream outlets — including the Times and Fortune — are framing this as a triumphant tech story about bold bets on the future. The numbers are real, the growth is real, and the demand for compute is real.

The hyperscalers are profitable enough to absorb a bad year or two. But the secondary effects — power grid strain, private credit tightening, muni market exposure, and supply chains already stressed from chip shortages — persist regardless.

Regular people are not buying NVIDIA stock. But they are paying electric bills. When Duke's grid is being pushed at 10 times historic growth rates to power data centers in North Carolina, ratepayers absorb some of that cost.

The AI build is real. The demand is real. The money is real. The risks are also real — and they are dispersing across the economy in ways that quarterly earnings calls do not fully capture.