Big tech plans to spend $735 billion on AI next year. Nobody can prove it pays off

2.6%.

That is how far the Nasdaq fell on 22 October 2026, its steepest one-day drop since June. The S&P 500 slid 1.2% the same day, its worst since late July. The Nasdaq is now about 5.8% below the all-time high it set just two weeks earlier, on 6 October. Behind both falls was one question investors keep coming back to. Are Microsoft, Amazon, Alphabet and Meta spending more on artificial intelligence than they can ever earn back?

The sell-off started with one company's earnings. Meta Platforms raised its own 2026 spending plan to between $135 billion and $145 billion, and posted its first quarterly cash outflow in more than ten years, meaning it spent more cash than its business brought in. To help pay for the build-out, Meta also announced a two-for-one split of its shares into a new non-voting class, a way to raise fresh capital without giving up board control. Investors did not like what they saw. Meta's shares fell almost 7% the next day, the worst drop among the big technology names.

Amazon fell 4.6%, Nvidia 4.4%, Tesla and Alphabet both 3.1%, and Microsoft 2.2%. Apple was the exception, up 1.2%. Intel and AMD each dropped about 8%, and the index that tracks US chip makers had its worst single day on record, down 6.3%.

Photo: BalticServers.com / Wikimedia Commons, CC BY-SA 3.0

Why does one company's earnings move the whole stock market?

Because seven companies now carry more weight than almost all the others put together. Apple, Microsoft, Nvidia, Amazon, Alphabet, Meta and Tesla made up 34.9% of the entire value of the S&P 500 as of 14 October 2026, according to market tracker MacroMicro. That is roughly a third of America's main stock index sitting inside seven names. When one of them has a bad week, millions of index funds and retirement accounts feel it, even though hundreds of other companies on the index did not move at all that day.

How much is Big Tech actually spending on AI, and is it making the money back?

Each company's own earnings release gives the numbers. Lined up against 2025, they show the size of the jump and, where disclosed, what AI income has come in to match it so far.

Source: Microsoft, Alphabet, Amazon and Meta earnings releases, 2025 full-year actuals against 2026 guidance given through July and October 2026.

Read nextWhy Netflix keeps raising prices, and what every plan costs in 5 countries

Add the four together and they plan to spend about $735 billion on AI infrastructure in 2026, up from roughly $380 billion in 2025. That is close to double in a single year. None of the four has disclosed AI income anywhere near that size. Microsoft's AI business, the most transparent of the four, brings in about $37 billion a year, roughly a fifth of what the company plans to spend building the infrastructure behind it. Amazon and Alphabet are growing their cloud and AI income faster in percentage terms, but still nowhere near the size of their new capital spending.

380.00$bn2025

735.00$bn2026

2025	380.00$bn
2026	735.00$bn

Source: Microsoft, Alphabet, Amazon and Meta earnings releases, 2025 actuals vs 2026 guidance as of Oct 2026

Does my pension or 401k already own a piece of this bet?

Probably, even if nobody asked you. Target date funds, the kind that pick an investment mix for you based on when you plan to retire, are now the default option in most US workplace retirement plans. Vanguard's own data, reported by Fortune, shows 7 in 10 of its retirement plan investors keep their entire balance in a single target date fund, up from just 1 in 10 two decades ago.

Those funds are mostly built from index funds that track the whole US stock market or the S&P 500. Because the seven AI spenders make up over a third of that index, most retirement savers already own a slice of the bet, whether they ever chose tech stocks or not. Put another way, for every $10,000 sitting in a plain S&P 500 index fund, roughly $3,490 of it currently rides on those seven companies and the AI spending decisions they are making this year.

South African readers are not shielded from this either. Retirement annuities and unit trusts that hold global or offshore equity funds, a common way local savers chase growth outside a shrinking local economy, mostly track the same US indices. South African retirement funds face a similar squeeze from a different direction too, where only one in ten workers are currently on track to retire comfortably.

Photo: Nielsoncaetanosalmeron / Wikimedia Commons, CC BY 4.0

Is anyone warning this could go wrong?

Michael Burry, the investor who became known for calling the 2008 housing crash, has been warning about this for weeks. In a blog post on 25 September 2026, he said the AI spending boom could lead to massive write-offs by 2028 or 2029, and compared it to the dot-com bubble of the early 2000s. He estimates Microsoft, Amazon, Alphabet, Meta and Oracle have taken on close to $3 trillion combined in leases, loan guarantees and other commitments tied to AI data centres.

Trending nowBootlegger data leak: The locked dashboard that was never locked

Burry also pointed at an accounting change. Microsoft recently extended the useful life it assumes for its data centres and office buildings from 15 years to 25 years. That change lowers the yearly depreciation charge the company books, which makes its profits look healthier for longer, even though the underlying spending has not changed.

Burry is not alone. Ratings agency S&P Global expects five of the six biggest cloud companies to run out of free cash by 2027, once combined AI spending across the sector reaches $1.3 trillion a year. Spending across the sector is growing about 70% a year. Operating cash flow, the cash these companies actually bring in from running their businesses, is growing about 23% a year. That gap between the two numbers is most of the story.

So is the AI spending bubble about to burst?

Not everyone agrees it is a bubble at all. Jay Goldberg, a senior analyst at Seaport Research Partners, said the sharp swings this week say more about investor mood than about the companies themselves. "There are legitimate concerns about the AI trade, but this level of shift in the markets is driven much more by investor sentiment than any tangible reason," he said. He added that everyone he speaks to blames a different cause, calling it a giant wall of worry.

Both things can be true at once. The spending in the table above is real. The gap between what is spent and what has been earned back so far is real too. Whether that gap closes as AI products catch on, or turns into the write-offs Burry expects, is not something anyone on Wall Street can answer yet with certainty.

Some investors use weeks like this to look at safer corners of their money. Savings account rates have already started moving after the Fed's latest rate cut, and gold has drawn fresh attention as a hedge against stock market swings. None of that is a suggestion for what anyone should do with their own money. It is simply where some attention has gone this week.

What is certain is the size of the bet already placed. Four companies now plan to spend $735 billion next year on a technology that, by their own disclosed numbers, earns back only a fraction of that so far. If the bet pays off, it could define the next decade of the US economy. If it does not, the bill will not stay inside Silicon Valley boardrooms. It will show up inside millions of retirement accounts that never chose to make that bet in the first place, including some that belong to people who have never typed the word AI into a search bar.