Big Tech eyes $700B AI spending
- Alphabet, Amazon, Meta, and Microsoft have now put numbers around a 2026 AI buildout that Wall Street pegs near $700 billion combined. - The eye-catcher is the jump rate: analysts see capex up roughly 60% year over year, with Amazon alone guiding about $200 billion. - That matters because AI winners now depend less on apps and more on who can fund chips, power, and data centers.
The story here is not just that Big Tech is spending a lot on AI. It’s that the four companies that matter most to the AI infrastructure race — Amazon, Microsoft, Alphabet, and Meta — are moving from “huge budgets” to something closer to wartime industrial buildout. By early February, Wall Street estimates had their 2026 capital spending approaching $700 billion combined. A few months later, that number was still the frame investors were using, with some estimates drifting even higher. The market is treating that as a real regime change, not a one-quarter burst. (cnbc.com) ### Why is $700 billion such a big deal? Because this is not normal tech capex. These companies already spent at historic levels in 2025, and analysts still expect another leap in 2026. CNBC’s February tally had the four hyperscalers nearing $700 billion combined for the year, up more than 60% from already elevat(cnbc.com)n track to match or exceed the prior three years combined for individual companies. (cnbc.com) ### What are they actually buying? Mostly the unglamorous stuff that makes generative AI work at scale — Nvidia-class GPUs, networking gear, giant data centers, and the power systems to run them. This is why the spending wave keeps flowing into chipmakers, memory suppliers, optical networking names, utilities, an(cnbc.com)ompany.” Basically, it’s “buy the bottleneck.” (cnbc.com) ### Which company is pushing hardest? Amazon is the most eye-popping on the latest numbers. CNBC reported in February that Amazon said it expected to spend about $200 billion in 2026, a figure large enough that analysts at Morgan Stanley and Bank of America were modeling negative free cash flow for the year. Meta(cnbc.com)ion. Alphabet and Microsoft are spending at similarly enormous scale, even when they phrase guidance more cautiously. (cnbc.com) ### Why are investors both excited and nervous? Because the same spending that could lock in AI dominance also crushes near-term cash generation. Last year these four companies generated about $200 billion in combined free cash flow, down from $237 billion in 2024, and the pressure looks worse in 2026 as the buil(cnbc.com)tures demand. The bear case is just as simple — they overbuild, margins get squeezed, and returns arrive much later than promised. (cnbc.com) ### Why does this lift Nvidia and memory stocks? Because a big chunk of every new AI data center budget ends up upstream. More servers mean more accelerators, more high-bandwidth memory, more networking, and more storage. That’s why every fresh hyperscaler capex guide tends to ripple immediately into Nvidia, memo(cnbc.com)budgets as future purchase orders, even if the exact vendor split moves around. That last part is an inference, but it fits the spending mix described in the reporting. (cnbc.com) ### What could break the story? Power. Turns out the hardest bottleneck may not be chips but electricity. The Financial Times has shown that from 2026 onward, several new AI facilities will each need around 1 gigawatt of power, and projected US data-center demand is outrunning near-term grid capacity. If the grid(cnbc.com)how much cash Big Tech throws at it. (ig.ft.com) ### So what’s the real takeaway? AI spending has moved out of the experiment phase. The market is now valuing the sector like an infrastructure arms race — one where the winners may be the companies that can finance, source, and power the biggest buildout first. That’s why the $700 billion figure matters. It tells you AI is no longer just a software story. It’s an industrial one. (cnbc.com)
