The simple version

Amazon, Microsoft, Alphabet, and Meta have each told investors how much they plan to spend on capital projects in 2026. Added together, by a Financial Times compilation of their guidance, it comes to roughly $725 billion, up about 77% from the $410 billion the same four spent in 2025.

Almost all of it is artificial intelligence: chips, data centers, and the power to run them. That is the largest single-year corporate investment cycle in history, and here is the twist that made this week interesting. As the numbers climbed, the stocks fell, because investors stopped asking how big the spending is and started asking when it pays off.

The numbers

• Amazon plans about $200 billion in capital expenditures in 2026, the largest of the group, a plan chief executive Andy Jassy set out on the company's fourth-quarter 2025 earnings call (Amazon, company guidance)
• Microsoft is tracking toward about $190 billion for the calendar year, with roughly $25 billion of that attributed to rising memory and component costs (Microsoft, company guidance)
• Alphabet raised its 2026 capital-spending guidance to $195 billion to $205 billion at its second-quarter results on July 22, up from the $180 billion to $190 billion range it set in April (Alphabet)
• Meta guided to $115 billion to $135 billion, later raised citing higher memory-chip prices (Meta, company guidance)
• Combined, the four plan roughly $725 billion in 2026, up about 77% from about $410 billion in 2025, per a Financial Times compilation of company guidance; that compilation used Alphabet's earlier range, so the raise above pushes the true total higher still (Financial Times)
• Analysts project the combined figure topping $1 trillion in 2027 (as reported)
• The spending goes to graphics chips, custom silicon designed in-house, data centers, and electrical grid capacity (company disclosures)
• Alphabet reported record profit last week and its stock fell as much as 7% intraday before closing down about 6%, partly on free cash flow of negative $5.9 billion, and the Nasdaq slid about 2% on the week as investors scrutinized the spending (our prior coverage; market data)

Where three-quarters of a trillion dollars actually goes

A number this large is hard to feel, so start with a rate: $725 billion across a year is about $2 billion of spending every single day, weekends included. The money buys four things, in rough order of size.

Most of it buys computing hardware: the specialized chips that train and run AI models, both the graphics chips bought from suppliers and the custom silicon the companies design themselves. Then come the buildings to house them, the data centers, which are essentially giant warehouses of those chips.

Then the electricity to power and cool them, which is why these companies are suddenly signing deals for power plants. Underneath all of it sit the memory chips whose price spike we covered earlier this month, now large enough that two of these companies named memory costs as a reason their spending rose.

The justification the companies offer is demand. Alphabet has pointed to a cloud backlog, meaning contracts signed but not yet delivered, of $514 billion. The argument is simple: customers are lining up to rent this computing, so building it is meeting real demand, not guessing at it.

Why the market's mood flipped from excited to nervous

For two years, huge AI spending announcements pushed these stocks up, because investors read them as evidence of ambition and future growth. This week, the same kind of announcement pushed them down. The reason is a shift in the question being asked.

Capital spending does not hit profit all at once. The cost of a data center is spread over years through depreciation, so a company can spend $200 billion and show it slowly, quarter after quarter, for years after the cash goes out. That means the spending is happening now, the charges are starting to arrive, and the revenue that is supposed to justify it has to show up on time, or margins get ground down while everyone waits.

The signal investors watched this week was cash. Alphabet's free cash flow, the money left after spending, went negative by $5.9 billion despite record profit, and Amazon's is projected to turn negative this year too. When the most profitable companies on earth start burning cash to build, the market stops treating the spending as free ambition and starts treating it as a bet with a clock on it.

This is the same lesson our earlier coverage of a single company scaled up to an industry: the spending is real, the demand may be real, and the payoff is a question the reports have not yet answered. Nobody knows whether $725 billion a year produces returns to match.

The companies are betting it does. The market spent this week pricing the doubt.

The Real Cost lens on a bet you probably hold

This is not a distant corporate story if you have a retirement account, because these four companies are among the largest holdings in the index funds most Americans own.

• Amazon, Microsoft, Alphabet, and Meta together make up a large share of the S&P 500 by value, so an S&P 500 index fund holds a meaningful slice of this bet automatically
• That means the $725 billion wager is, in part, being made with money sitting in ordinary retirement accounts, without any account holder choosing it directly
• If the spending pays off, index holders share the gains; if it disappoints, they share the disappointment, which is what a week like this one looks like in a 401(k) statement
• None of that is a reason to act. It is a reason to understand what you own: index investing means holding the market's biggest bets, and this is currently the biggest one in the world

The point is ownership literacy, not a trade. When you read that Big Tech is spending unthinkable sums on AI, the honest translation for most people is: my retirement account is helping pay for it, and its value will rise or fall with whether the bet works.

What this means

The reports coming next week from Microsoft, Meta, and Apple will be read through this lens. The headline revenue will matter less than three things: how much each is spending, whether the cloud and AI revenue is growing fast enough to justify it, and what is happening to free cash flow. That is the scorecard the market has decided to use.

The broader lesson outlasts this cycle. Any company making enormous capital bets runs the same race between spending now and earning later, and the market's patience with that race is not infinite. This week it got shorter.

What this is NOT

This is not a prediction of these companies' results, their stocks, or whether the AI buildout pays off, which nobody knows. This is not advice to buy, sell, or hold any of these companies, any index fund, or any security, and it is not a recommendation for or against index investing. This is not a claim that the spending is wise or wasteful, which this article takes no position on. The $725 billion aggregate is a Financial Times compilation of company guidance, and each company's figure is its own stated plan, subject to revision. This is not investment advice of any kind.

Sources

• Alphabet Inc., Form 8-K, Q2 2026 results, Exhibit 99.1, SEC EDGAR: https://www.sec.gov/Archives/edgar/data/1652044/000165204426000066/googexhibit991q22026.htm
• Amazon.com, Inc., Q4 2025 earnings and 2026 capital expenditure guidance, as reported: https://www.cnbc.com/2026/02/05/amazon-amzn-q4-earnings-report-2025.html
• Amazon.com, Inc., SEC filings index, EDGAR: https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=AMZN&type=8-K
• Financial Times, compilation of Big Tech 2026 capital expenditure guidance (as reported)
• Microsoft Corporation, investor relations: https://www.microsoft.com/en-us/investor
• Meta Platforms, Inc., investor relations: https://investor.atmeta.com/