Amazon, Google, Meta, and Microsoft have spent more than a trillion dollars on AI infrastructure, including data centers, the chips inside them, and the power needed to run the facilities, since 2023. The Financial Times said that these four big companies have already hit $1.1 trillion in capital expenditure based on their latest earnings reports, and that an additional $745 billion is expected to be added to this figure just this year.
“There is basically no end in sight for the growth in capex,” RBC Capital analyst Rishi Jaluria told the publication. “Investors need these companies to toe the tight line between investing in AI and not compromising the things that have made them successful.” This massive investment has upended several other industries — namely electricity prices and memory and storage chips. The massive power demand that data centers have put on the power grid has forced many U.S. utility companies to spend billions of dollars to upgrade their respective infrastructure, which they then passed on to all consumers, not just the big ones that forced the upgrade.
This, alongside other environmental issues, has caused many Americans to push back against data center projects near their communities. The White House instituted the “ratepayer protection pledge” and made AI hyperscalers, utility operators, data center companies, and individual states promise that they will protect the average consumer from electricity cost increases. But so far, no state has taken a step to codify this pledge into law. Oregon actually enacted the POWER Act, which resulted in a 30% increase in the power bill of users that consumed more than 20MW while slashing the bills of residents by 1.3%, but the state did this in 2025, way before President Donald Trump called the tech giants into the White House and told them to “pay their own way.”
The mountains of cash that these tech giants are pouring into AI are also affecting the memory and storage chip industry. Since these AI hyperscalers have a lot of liquidity from investors, they are willing to pay top dollar for the HBM they need to run their data centers. Because of this, it made sense for Micron, Samsung, and SK hynix to prioritize them over DRAM, especially as they can charge a premium for these chips and there are customers who are willing to pay at those prices. This resulted in a shortage of consumer memory that started in 2025 — while this initially affected PC builders and enthusiasts, it has started to affect other industries that require memory as well, including cars and smartphones. Even Apple, which historically had huge sway over its suppliers, was forced to increase prices because of the shortages.
Aside from skewing other industries, the massive CAPEX the big four are going into is alarming some experts, warning that the promises and contracts they’re making are leading to “hidden debt” not listed in their balance sheets. The amount, worth around $1.65 trillion, is annotated in their quarterly financial statements as future obligations that will only come into play as the related asset or service comes online. The current value is 122% of the actual debt reflected on their balance sheets, which could give investors the wrong impression that they have fewer obligations than they actually have.
While the amount of money that the big four are spending on AI might seem dizzyingly high, we must note that these companies are raking in massive amounts of cash quarterly themselves. Microsoft’s latest quarterly revenue is $90 billion, while Meta made $60 billion in the same period. Alphabet (Google) announced revenue of nearly $120 billion, while Amazon made $200 billion. That is a total of nearly $470 billion for these companies in just the last quarter.
Still, that does not mean that they can just keep on spending on AI. For example, even though Google’s cloud business had a revenue of $11 billion last year, its price dropped after it announced that it spent more than it made last quarter — the first time this happened in the 20 years since it went public. Meta is also planning to rent out its AI compute, apparently following in the footsteps of Amazon, Google, and Microsoft, which have growing cloud businesses. However, this announcement caused a drop in its stock price. “They are a bit all over the place,” SLC Management managing director Dec Mullarkey told FT. “For investors it’s no longer growth at any cost; they want to see the spending flowing through to results, like at the Big Three.”
Follow Tom's Hardware on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.
