Amazon is preparing to spend well over $100 billion on capital expenditures in 2025, and the company says AI will take the central role in that plan.
The signal from Amazon is not an isolated one. Across Big Tech, companies are continuing to commit huge sums to AI infrastructure even after recent debate over whether cheaper AI systems could reduce the need for such aggressive investment.
Amazon’s 2025 Plan Puts AWS At The Center
Amazon CEO Andy Jassy said during Amazon’s fourth-quarter earnings call Thursday that the “vast majority” of the company’s expected 2025 capital expenditures will go toward AI capabilities for AWS, Amazon’s cloud division.
The spending level is based on Amazon’s Q4 2024 capex figure of $26.3 billion. Jassy said that amount “is reasonably representative” of what to expect on an annualized basis in 2025. Multiplied across four quarters, that points to $105.2 billion.
That would be a major increase from the $78 billion in capex that Amazon spent in 2024. The company is effectively telling investors that the next phase of AI demand is large enough to justify a bigger infrastructure push, not a retreat.
The focus on AWS matters because Amazon’s AI strategy is closely tied to cloud demand. The source article notes that AWS has many AI offerings, and Jassy argued that cheaper AI would expand use rather than reduce overall revenue opportunity.
Why Cheaper AI Is Not Slowing The Spending
The recent DeepSeek discussion raised a question for the industry: if AI becomes less expensive to run, will Big Tech need to spend less on infrastructure?
Amazon’s answer, at least for now, is no. Jassy brushed aside concerns that falling AI costs would hurt the company’s revenue. His argument is that lower costs tend to make technology more useful and more widely adopted.
“Sometimes people make the assumption that if you’re able to decrease the cost of any type of technology component … that somehow it leads to less total spend in technology. We’ve never seen that to be the case,”
Jassy compared the expected AI demand pattern with the early days of the internet and cloud. The logic is straightforward: when a technology becomes cheaper and easier to use, more people and businesses may find reasons to use it.
That does not prove the bet will work. It does explain why Amazon is not interpreting lower AI costs as a reason to pull back. Instead, it is treating efficiency as a possible demand accelerator.
Big Tech Is Making The Same Bet
Amazon is not alone in this view. Other major technology companies are also using this earnings season to defend large AI spending plans.
Meta CEO Mark Zuckerberg said last week that the company would spend “hundreds of billions” on AI in the long term. The source article says he cited rising inference demand across Meta’s billions of users. Meta is slated to spend at least $60 billion on capex in 2025, mostly on AI.
Alphabet also increased its capex plan for 2025 by 42% to $75 billion. CEO Sundar Pichai justified that spending by saying lower AI costs “will make more use cases feasible.”
Microsoft announced last month that it would spend $80 billion on AI data centers in 2025 alone. Microsoft CEO Satya Nadella also pointed to Jevons Paradox as the DeepSeek discussion was heating up, tweeting the Wikipedia page for the concept.
In that tweet, Nadella wrote that as AI becomes more efficient and accessible, its use will rise sharply. The point matches the argument now being made across the sector: cheaper AI may mean more AI, not less AI.
The Stakes Behind The AI Capex Race
The shared message from Amazon, Meta, Alphabet and Microsoft is that AI demand is expected to keep growing. Their spending plans suggest that the largest technology companies see infrastructure as a competitive requirement, even as investors question the returns on rapidly rising AI expenses.
For Amazon, the key figure is not only the expected total of well over $100 billion. It is also the shift from $78 billion in 2024 to a 2025 run rate that could reach $105.2 billion based on Q4 2024 spending.
That gap shows how much more aggressively Amazon expects to invest. It also shows that AWS remains central to the company’s AI ambitions, because the “vast majority” of the spending is expected to support AI capabilities in the cloud division.
For the broader market, the important question is whether the Jevons Paradox argument will hold. If lower AI costs unlock more usage, Big Tech’s spending plans may look like preparation for a much larger wave of demand. If demand does not grow enough, the pressure around returns on AI investment will intensify.
For now, there is no sign of an AI spending slowdown. The companies with the largest AI platforms are not reacting to cheaper AI by reducing their budgets. They are treating it as another reason to build more capacity.